QUESTIONS REGARDING SAFE

• Document Type: CREST
• Collection: General CIA Records
• Document Number (FOIA) /ESDN (CREST): CIA-RDP84-00933R000500120001-5
• Release Decision: RIPPUB
• Original Classification: K
• Document Page Count: 85
• Document Creation Date: December 9, 2016
• Document Release Date: March 26, 2001
• Sequence Number: 1
• Publication Date: March 18, 1980
• Content Type: MF

Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
OFFICIAL USE ONLY
APPENDIX I AND II (T9 srPle
7
Approved For Release 2001/07f2J:cCtITT86,-\0,0933R000500120001-5
?
14A rITT,.?.ilag, fill
Approved For Release 2tQl/t0702,.:CIA*4RIDF'88i-uou3ixo
' elDENDIX I.
500120001-
DIRECTOR OF CENTRAL INTELLIGENCE
Science and Technology Advisory Panel
?
?
18 MAR 1983
MEMORANDUM FOR: Director of Central Intelligence
.?
Deputy Director of Central Intelligence .
STATINTL FROM:
STATI NTL
JU Lflalrman
SUBJECT: Questions Regarding SAFE
1. In response to your request of March 14, 1980,
--
for comments on the current status of SAFE, I attach a
list of eight questions and brief comments that STAP
believes should be addressed.
?/
ft
STAP is continuing its analysis of the SAFE
problem and will prepare an options paper within the next
two weeks for your consideration.
WI ?
Attachments:
.As Stated
Vir
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
L.?-)1-
I I I
01EP oxrf:nH.,,it. USE alio
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
1.0 What steps are being taken to ensure that the Agency,
rather than the contractor is in control of the technical
aspects of the design of the system?
The geographic remoteness of the contractor, and the
lack of continuing contractor-user interaction can lead to
a situation in which the builder of the system also becomes
the architect. The absence of reciprocal technical
representation, like resident engineers, delays every
routine decision and makes larger ones unresponsive
Agency/community needs or technological constraints. Striang
Agency technical management is absolutely essential if the
system is to satisfy real and evolving agency needs and if
it is to be integrated with other Agency and community
resources.
2.0 How is SAFE management ensuring that a final working
system has been developed from the continuing evolution of
an operationally valid pilot system? Now has SAFE take'
advantage of the experience of similar, very large systems
in their:
1)%System architecture,
2r'communication and control, and
3) changing performance requirements?
How is SAFE management ensuring that the system will:
1) make available data on operation and usage of the
pilot system to guide development;
2)..be able to modify both system functions and
interaction capabilities so as to meet changing and evolving
requirements; and
3) be able to add new functions and interactions so as
to meet new requirements?
Relevant Experience -- It is well agreed that an
information handling system cannot be achieved by a simple
Design-Build-Use cycle, no matter how brillant the design
or faithful the building. Consider three (out of many)
currently operating very large nets with requirements at
least comparable in size and complexity with the Agency/
community's:
1) the ARPA net,
2) the airlines reservation system, and
3) the IBM in-house computing net.
Approved For Release 2001/07/12 : CIA-RDP84-00933R0001500120001-5
001.,
Approved For Releas
1/%717121), VA-R
044000500120001-5
All of these systems evolved - that is, they started
as soon as possible with operating pilot systems, so that
there was always an operational- evaluation of effectiveness.
Furthermore, in each case, a primary purpose of the original
plans was drastically altered as experience was gained. Even
so, developments in concept, hardware, software, and practice
are continuing now in greater volume than ever.
2.1 The ARPA Net
The ARPA net was originally conceived by
at the ARPA IPO (Information Processing Office) as a means
of netting research computers in order to do distributed
computing. The message facility was then a minor function.
As Message usage rose, packet switching became a powerful
tool with wide application elsewhere in the technology. The
separate centers in the ARPA net have much independence,
subject to some fairly strict requirements for communications
protocols and access. This allows competitive development
and a common evaluation of new technological developments -
e.g., the intelligent terminal is being subjected to widespread
experimentation and development.
4,
2.2 Airlines Reservation System
STATI NTL
The first automated electronic airfare reservations 'and
ticketing systems were disasters - e.g., SABER, the
American Airlines/IBM effort was predesigned and built to
an apparently reasonable set of specifications that turned
out not at all to match the operational needs. The first
successful ones were ad hoc temporary devices (e.g., UNITED)
that worked just well enough to be improved.
Note that the requirement for fast and accurate inter-
line communications, backed up by automatic'commitments for
seats and equipment was a much later development; it is, however,
by now one of the most valuable and cost-effective facilities.
2.3 The IBM In-House Computing Net
Virtually all the on-line computers at IBM company
.installations world-wide are netted together by communication
-2-
g
rntt
Approved For Release 2001/0t/A : ORpR434.100:532R000500120001-5
srensermirreemr.m.? _ ? .
I I I
???????.1?11111~....."1"61"110110.161110.00010011.0.11
?
CA ---------"---Approved-Forikeiease-2001/0711-2-telMRDP64-009-35R0005001-200-----
'
. ,
PULL UL
1
facilities, making them, we believe, the largest net in
the world. Admirable new functions exhibited by their
systems are: ,
1) a modifiable macro command language,
2), a consistent, speedy and flexible data transmission/
translation scheme.
3.0 The SAFE user community consists of Intelligence
Community analysts covering the full spectrum of research
into foreign political, military, economic, scientific,
and technological activity. Their effective use of this
system and, ultimately, the quality of intelligence they
produce rest on whether their real needs can be identified
and satisfied by the system. To that end, what actions
will be taken to ensure that:
1) all elements of the intended user community
are actually involved in the system's continuing development,
?
2) the broadest of the present analytical requirements
are ideiatified,
3) these present requirements can be validated by a
consistent method,
*.
4) the validated present requirements will be met,
5) modifications and new requirements can be accepted
as they,are identified,
6) all analytical users will acquire the necessary
skills and familiarization with SAFE on an interim basis so
they are ready to begin broad utilization when system IOC is
reached,
7) :this interim SAFE test-phase acquires continuing
comprehensive experimental data on user experience with the
system, and
8) .the acquired data on user experience is actually
utilized in the architecture and development of the system?
-3-
A IHLY
--"'""rx4tj;'Piiefwet"?O,i44a614iass.2aO1/07l12 CIA-RDP84-00933R000500120001-5
Approved For Release 201,074ig7tPtitiS4-094)/1110500120001-5
4.0 How can the Agency make a reasonable evaluation of the
current status of SAFE with major portions of the proposed
operational capabilities either unspecified or uncommunicated
to the Agency? For example,
1). the user command language and its parsing,
2) the user programming languages,
3) the user editing languages, and
4) procedures for backup, including regeneration
derived files lost in crashes.
Note the above have to be ari2Laypes capable of continuing
responsive evolution, rather than final imposed prescriptions.
5.0 What actions are under way to insure that the Intelligence
Community has access to CIA SAFE and that CIA SAFE has access
to DIA SAFE as well as such systems as COINS and SOLIS?
GIL
The absence of appropriate linkages with other IC systems
makes it highly probable that duplicate facilities and files
will be acquired and constructed with higher costs and lessened
capabiLity for the total IC system. Provisions for such
linkages should be built into SAFE from the start, otherwise
itwill be difficult if not impossible to.backfit these
linkages. STATSPEC
p.
6.0 How will SAFE deal with open source material? Will
material, either finished publications or field reporting,
be made-available to the analyst through SAFE? How will SAFE
deal wiEh current newspaper and journal entries?
Several of the offices that will use SAFE, in particular
OPA and OSWR, make extensive use of open source material.
Their analytical efforts will be seriously hampered if their
files do not include open source materials.
7.0 What steps are being taken to ensure that SAFE will be
designed,to allow collaborative usage?
Examination of other similar systems, such as Stanford
University's SUMEX system or the internal IBM system, shows
Approved For Release 2001Va12d iA i:C084-1!; 31461Cid00120001-5
I I I
? ? ? ?? ? - " .
CA NTEIINAL USE LY
that user-to-user interactions comprise a significant
fraction of the total use of these systems and greatly
enhance the overall analytical capability of its users.
8.0 What steps' are under way to ensure that in the
procure.ment of major hardware items, these items will be
compatible with existing Agency systems?
The Agency has made very substantial investment
ADP equipment that currently serves a wide variety of.
users. This investment should be capitalized on in order
to enhance the future capabilities and particularly the
flexibility of SAFE. If the interoperability of SAFE and
existing ODP hardware is going to be dependent on software,
then provisions should be made:
1) for the development of the needed
it will be a major undertaking; and
2) for the establishment of evolving
protocols for interconnection.
,
N.
-5-
software since
standards and
CAA INTEM.IA.1 USE ONLY ?
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
I I I
ApprOV-e-d For Release 2001/07%12 : CIA-RDP84-00-93-311-600-5-0012-6001-5
MANAGING VM/CMS SYSTEMS FOR USER EFFECTIVENESS
APPENDIX II
Walter J. Doherty and Richard P. Kelisky, IBM Thomas J. Watson Research Center,
Yorktown Heights, New York. USA
Abstract
Management of the VM/CMS systems at IBM's Thomas J. Watson Research Center is based on assumptions
that the user's time and work product are valuable, and the system being used should be managed so that the
user's ability to work is enhanced with least inconvenience to the user. Within the financial limits on
Computing Center hardware and personnel it is possible to provide user-effective interactive computing
services which approach these objectives by placing highest priority on system reliability and availability
followed by performance improvements and introduction of new functions. Because our studies show that
user productivity is well correlated with system response we are convinced, first, that it is essential to maintain
a management commitment to the goal of good service to the laboratory, and second, that published service
schedules must be adhered to. At the same time,, it is necessary to realize that users will have spezial
problems, and management must be willing to work out solutions for these users. The goal of the computer -
operator is to keep the system up, and system programmers must introduce necessary system changes in a way
which does not jeopardize stability. We realize that the computing system is a tool which enhances both the
usei's memory and reasoning power. We have, therefore, assumed responsibility for preserving our users'
data. We have also developed data migration mechanisms to aid the user in managing data.
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
'vERstotsT 53 --LEvr 32?DOC R5cr-TDS- lines - -18
S
ApprO-ved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
Introduction
This paper describes the evolution of VM/CMS interactive computing services at the IBM Thomas J. Watson
Research Center, and the necessary changes in viewpoint needed to manage interactive computing services
effectively for the user community. Benefits deriving from alternative strategies have been evaluated against
the following criteria:
(1) The computer user's time and work product are valuable. Therefore, the interactive systems on which
the user works should be managed so that his ability to work is enhanced with least inconvenience to
him.
In fact, for most interactive computing the aggregate user time is more costly than the computer time. Figure
1 shows the cumulative percentage of people using computers at Yorktown vs. the cumulative CPU cycles
consumed by those people. Similar curves characterize the distribution of computing usage in many other
installations. (I) In general, about five percent of the people consume about seventy five percent of the
computing resource. If we focus on the top five percent of the computer users we find that a small fraction of
their interactions dominates their demand for computing. The computer user's technical management, not the
Computing Center, must determine whether or not the machine intensive computing to be done is technically
justified. Machine intensive computing occurs for about 5 percent or less of the interactions. The remainder
of the interactions involve immediate, ongoing communications between people and the computer. If we use a
figure of $800 per hour for the cost of a large computer today, the 25 percent used for 95 percent of the
interactions costs $200 per hour. One hundred simultaneous users, whose own time might be worth at least
$20 per hour, cost $2,000 per hour. These figures are only for illustration. In fact,. our computer costs are
less, our users' time is worth more, and we normally have many more than 100 simultaneous users on each of
two VM/370 systems at Yorktown. Therefore, for more than 95 percent of the computer interactions, the
user's time is much more costly than the computer's. Furthermore, these costs are diverging. The Computing
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
!VERSION- 53 -11:WEL 52 ?0C3C.# 8859 PD5linen 19 -
Approved For Release 2001/07112 : CIA-RDP84-00933R000500120001-5
Center at IBM's Thomas J. Watson Research Center attempts to provide an environment in which the
computer user can then make effective use of his time, and efficient use of the computing resources provided.
Figure 2 shows the growth in the interactive use of computing E. Yorktown from 1973 through 1977. It is
clear that the aggregate cost of the users' time is greater than our computer costs, and this difference
continues to grow. Therefore, the Computing Center seeks maintenance and enhancement strategies which,
within budget limits, help the user to get work done and which tend to give highest priority to the user's
requirements for service. This means that in so far as possible we must isolate changes from each other and
allow both old and new versions of functions to coexist so that the user is not forced to change his ways of
working.
An essential tool for our maintenance and enhancement strategy is the virtual machine in VM/370. In virtual
machines we are able to test new versions of most subsystems or applications without disturbing existing
versions. We can isolate many special functions to separate virtual machines which gives us the ability to
tailor the appearance of that function and simplify the user interface without fear of "contamination" from
other non-related function. When CMS is the operating system in the virtual machine, we find that we often
have least system overhead for performing that function.
(2) Complexity in the man-machine interface is wasteful of users' time, but complexity can be reduced by
the judicious use of procedures (2).
In April, 1978, the Yorktown Computing Center processed more than 10 million interactions. An interaction
is any user input and its accompanying system response. If the information content per interaction is too
O.
small, people could be decreasing the rate at which they work rather than increasing it. Or, viewed another
way if we can decrease the amount of work the user has to do to derive, a useful result from an interaction,
then we have made the system more effective for that user.
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
VERSEOW 53 LV[ 52 -DOC-..4 tig59 PDS -lines 42 -
v? ?
6
TF DOHERFY YFL V006 WJDRPK
Approved-Fac--Release-20
PAGE 4 HIS I. Oti/l/// 15 11:11
.2-:, e:? -es:
We find that often our computer users will take great care to determine how to use some combination of
computing functions for a task performed repetitively. Once they have done this the combination of functions
is given a name and stored on secondary storage. After the computer has stored this information, the person
need not repeat those thought processes again. These procedures (EXEC's) may include useful default values
for parameters as well as logic to make the computer adapt to the user's environment, rather than requiring
the person to adapt to the machine.. Once a user has constructed an EXEC, others can, and do take
advantage of it. For example, EXEC's exist to simplify the business of using IBM's internal computer
network. The user normally doesn't have to concern himself with the kind of data involved, the protocols, or
the ways of addressing the person or group to whom data is sent or from whom it is received. Good default
actions have been determined by one of our users, and the EXEC's which he developed are now routinely
used by most of our network users at Yorktown. The naive user need not even be aware of the existence of
the commands or parameters inside these EXEC's. By this process complexity is reduced, typing errors are
avoided, and the user-effectiveness of the interaction increases.
From examining twenty percent of the users on one of the VM/CMS systems at Yorktown, we find that there
are twice as many EXEC's as all other source programs in conventional programming languages combined. (3)
In 1977, the Stanford Linear Accelerator Center (SLAC) installation observed an average of twenty five
commands executed per command typed at a terminal (4), when using the Wylbur interface. Wylbur is an
editing interface for interactive computing. Wylbur was about ten years old at that time. In 1971, we found
at Yorktown that the number of commands executed per person, counted at execution time after procedure
expansion, doubled in just five months. 'Thus, our experience indicates that there is continued growth in
function as users learn to adapt the system to their requirements. We can think of this growth as arising from
captured intelligence. In the early 60's the computing industry speculated on the notion of artificial
gence whereby the computer would heuristically determine the solution to some problem. In practice, we find
the concept of captured intelligence to be extremely useful for raising the man-machine interface to a
meaningful level for many people.
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
'UtluCtriTZ- r rvtly 1II5 itS lines 4 -
TTF DOHERTY YFL V006 WJDRPK PAGE 5 HIST 08/07/78 17:17
Approm_erilor Releas- 2001/07/12 ? CIA-RD7114=011913130.0.1510.120.01t5____
(3) Display terminals are able to provide a user more information in a given time than typewriters. They
can be especially valuable to the user as an aid to how to use the system.
Where justified by the user's work in the view of the user's management, it is our objective to place an IBM
' 3277 display terminal on each user's desk. In general, we find terminal usage of an hour or more per day is a
reasonable prerequisite to installation of a terminal in a user's office. An example of the value of the display
terminal in accelerating the user's work is provided by our documentation practices on EXEC's. It has
become customary, at Yorktown, to include comments on how to use any EXEC directly inline with the
EXEC code. Thus, each user can determine what functions the EXEC does for him, what its syntax is, and
what the default values of parameters are by simply typing the name of the EXEC followed by a "?". By
having the comments included inline with the EXEC, these comments are usually changed whenever the
EXEC is changed, which maintains currency between the documentation and the code, and they are there
whenever needed. Our Yorktown user community has told us that these "self-defining" commands are
regarded very much more favorably than all other forms of documentation or education. This facility was
simply not feasible with the slower data rates of typewriter terminals.
(4) System programming is costly; therefore, management seeks solutions to user problems which do not
lead to growing commitments in system programming time.
System programmers are a scarce and costly resource. Strategies which increase system programming
requirements are to,be,avoided. Imgeneral, a local change to an operating systemcarries with it an obligation
to reevaluate that change everytime there is another change to that part of the operating system, e.g., PLC,
SU, PTF, etc. Such local changes, then, represent a kind of promissory note on which one pays interest
-?
(system programming time) indefinitely. The Yorktown Computing Center seeks strategies to reduce this
cost. For example, the Computing Center selected a strategy to attach new devices which avoids repeated
reprogramming as the underlying operating system changes. This subsystem, called the Advanced Terminal
Subsystem, attaches non-IBM terminals to an IBM System/7 which has been programmed so that the
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
TFF DON ERTY YFL V006 WIDRPK PAGE 6 HIST 08/07/78 17:17
Apprbved For Release 2001/07/12 . CIA-R
Ap-;-,,- 'Iui ii uii
underlying operating systems (VM/CMS, MVS/TSO, and formerly, TSS) on the 370/168 are unchanged.
Change is localized to the System/7, and new terminal types are introduced by means of table entry rather
than by modifications to MVS or VM.
. We have also distributed the responsibility for change to the people who most need it. By giving them special
virtual machines, and access to whatever source code they needed, together with good tools for debugging
?
changes and for finding out who else might be interestetk in any given function, we provide an environment in
which changes take place but reliability remains high. Special libraries exist to allow users to make their
changes available to others. By simply typing the command OWNER followed by a command name, any
person can determine who the current owner of that version of the function is, and communicate directly with
that person through the system for describing problems or new ideas for enhancement. Once such functions
have proven to have broad value to many people, they can readily be incorporated into the central Library
with little effort by the systems programmers.
(5) The development of new computing services is costly; therefore installation management develops new
services incrementally in order to evaluate these services before the next step is taken.
We avoid large development proi.,ccts which must span several years before their benefits are available to the z
users. Not only are such projectr. costly, but there is the great risk that the problem to be solved will change
or disappear during the lengthy development process. Because program development is a very complex
process, we find that feedback is required -continually throughout the development cycle. By bringing the
computer to the person who has the greatest knowledge of the problem to be solved, and providing that
person with adequate computing resource and an appropriate set of tools, we find that he can build a
prototype solution, try it personally, get others to try it, and iterate ma; times in this fashion until the
prototype is perfected to yield the required function. We find that this is the fastest way to produce high
quality function for the least cost. We also find that very skillful programmers absolutely require such
feedback throughout the development cycle. There is accumulated evidence to show that most errors have
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
' tVrki:tt)1,4-I IVI?K, rine, -44 "Ile.
eie-a=rairmirasalas mamma tosis Was
1TF DOfiERTY YFL V006 WIDRPK PAGE 7 HIST 08/07/78 17:17
Apprpveri For Release 2001/07/12 ? CIA-RDP84-00933R000500120001-5
???????
?1
occurred in the design stage (5), and that individual skills are the major factor in determining the quality of
programming projects of less than 50,000 lines of executable source code in size.
By developing new services in this iterative fashion, we are more likely to be sure that we are solving the right
problem, that the perceived solution is indeed a solution, and that the function can be readily changed as
future needs dictate. VM/CMS and the virtual machine available under VM give us the basic environment we
need to work in this fashion. Our experience provides evidence that this iterative method of program
development is an improvement over the traditional programming development process.
(6) User need for asynchronous computing grows proportionally to interactive computing; therefore the
availability of well-planned distributed function is increasingly important.
The load distribution curve of Figure 1 is the same for interactive computing as it is for traditional batch
processing. Because of the growth that has continued for the past ten years in interactive computing, there is
pressure to off-load that part of people's work which is large in computing resource demand, or which has
reached some logical stage of completion. If this were not done, long running processes would act as a block
at the terminal preventing a person from accessing his or her other data while that long running process
completed. By packaging special functions in special purpose virtual machines, it becomes easier to off-load
work whenever appropriate. These special purpose virtual machines need not run on the same real machine,
and in many cases, do not. However, to facilitate communication between such special purpose virtual
machines, A. N. Chandra developed an experimental mechanism which greatly facilitates communication
between different virtual machines running on the same real machine. We have employed this in our text
processing work, our laboratory automation work (6), our MSS data migration work, etc. The Virtual
Machine Communication Facility (VMCF) portion of VM/CMS is an outgrowth of Chandra's work. The
Network Job Entry and Network Job Interface software products have given us the necessary communication
facilities between different real machines. By being able to communicate with other machines, whether virtual
or real, we can permit new and old functions to coexist. Our users communicate with old function as needed
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
p.
tr L/l./1111: K. 1 I 1 ri. vu,o VVJ Littr
r atJC 111.11 1.10/ VI/ 10 11.11
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
rather than convert old functions to new environments. By insisting on having most of our terminals locally
attached, we minimize response time delays, reduce security problems associated with remote access, and
reduce the number of separate components required to be available simultaneously. Thus we find that
networking data rather than people is a more effective way to promote collaboration and avoid replication of
work.
?
In the following sections, management actions will be discussed as they affected the Growth of Interactive
Computing, Availability and Reliability, Response Time, Expansion Factors, and Visibility of Service, Data
Management, and Distributed Function. ?
The Growth of Interactive Computing at the Watson Research Center
The computer user at the IBM Thomas J. Watson Research Center has available a variety of computing
services. VM/370 is provided on two (six megabyte and seven, megabyte) 370/168'5 and OS/VS2 release
3.7 (MVS) together with TSO is provided on an eight megabyte 370/168-3. With their manager's concur-
rence, employees at the laboratory are urged to employ these computing facilities in order to do creative and
innovative work. Usage of the computer is not restricted to the scientist; secretaries and other non-technical
people use computers if management determines that it is cost-effective for them to do so. Since nearly all
computing services are accessed interactively from terminals in user's offices, terminal rooms or, in some cases
home terminals, we have attempted to develop management policies and practices which make interactive
computing services effective for the user. Some of these policies will be described in the following sections.
In what follows, prime shift means the period 08:00 to 18:00 Monday through Friday excluding holidays.
Second shift is 18:00 to 24:00 Monday through Friday and 08:00 to 18:00 Saturday, Sunday, and .holidays.
Third shift is all remaining time.
Interactive computing began, as a service, at the Thomas J. Watson Research Center with the introduction of
APL in 1965. To the computer user APL presented a fundamentally new interface different from the
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
, 1CILRSION? 53 I r VET 51-1)0C.# f0.131) FDS- lines To I - 181
TiF DOHERI Y Y1 L VUL/O WJURP1( PACA:, 9 IllS1 05/W/ 1: /
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
interface of batch computing. Clearly, most of the strong user acceptance of APL resulted from the APL
language and the APL system, but these positive characteristics might not have sufficed for user acceptance if
at the same time the APL group had not recognized that this new way of computing required a new way of
managing computing services. Simply stated, they recognized that interactive computing must be available in
order to be useful. The user must be able to turn on his terminal at any time he is ready to work and, with a
high degree of confidence, find the interactive system ready to serve him. APL service was provided on an
IBM 360/50 essentially twenty-four hours a day, seven days a week. Preventive maintenance of the 360/50
was limited to a few hours every three peeks, and because of the intrinsic reliability of the APL system and
the machine, the computer user was able to assume that APL would be available whenever he was ready to
work: in the evening, in the middle of the night, on weekends, holidays, etc. While Computing Center
management may not have realized it at the time, criteria for user effectiveness, system availability and system
reliability were being established for interactive computing services. These criteria have strongly influenced
subsequent management practices of the Watson Research Computing Center.
In 1967, TSS/360 was introduced at the Watson Research Center on an IBM 360/67. On TSS we began to
experiment with and to understand the effect on user productivity of variations in response time, the value of
indicating to the user a measure of system load prior to his logging on, the importance of scheduling controls
in order to distribute system resources equitably, and the need for "transparent" management of a growing
on-line user data base. TSS has been terminated at the Watson Research Center, but the experience gained
from TSS suggested new extensions to VM/370 and TS() so that these systems might become more effective
for the user.
In 1969, CP/67, the predecessor of VM/370 was introduced on a second 360/67 to serve as a software
development facility for operating systems. The irresistible user demand for CMS, which was not offered at
first to the computer users, showed the importance of providing an easy-to-use system for the non-specialist.
The increased load on the system resulting from demand for CMS showed that it was necessary to give the
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
VERSION-53- /.EVEL-52 -D005I-8859-113Slifies VW- -207?
I It? ULIIALK 1 Y Y L V UM" WJUKI-IC..
PAUL 1U 1-11S1 08/1.3// /8 11:1/
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
programmers, who had responsibility to maintain CP/67 (and later VM/370 which replace CP/67 in 1972)
toots to enable them to understand what is taking place inside the system so that a few users could not usurp
system resources to the detriment of system performance for the entire user community. In particular, we
learned that interactive system users are reluctant to accept the loss of a useful function or facility which they
used on an earlier interactive system, e.g., something equivalent to the ease of data management provided on
TSS must be made available on VM/370.
In the following sections we will examine in more detail results of our experience with managing interactive
systems.
Availability and Reliability
Computing services at the Watson Research Center are available, insofar as possible, twenty-four hours a
days wad seven days a week. Clearly, one must ask if such a schedule is accessary. Does the Computing
Center have cost-justified requirements from computer users for such a schedule? Technical management has
authorized computing which requires more than prime shift, but there are other reasons why such a schedule
was established: the requirement for computer availability by the experimentalists, the necessity that system
maintenance be done after prime shift, and the relatively low incremental cost to provide these added services.
There are, however, impediments to such a schedule: unscheduled software or hardware failures, scheduled
software changes or tests, scheduled hardware shutdowns for engineering changes, preventive hardware
maintenance, hardware relocation, and finally, the unavailability of personnel to operate the systems on
weekends and holidays.
On the question whether such extreme availability is necessary we note that researchers (computer scientists,
chemists, physicists, engineers, etc.) expect and require a degree of flexibility in working conditions.
Experiments cannot be constrained to the normal eight hour working day, and there are experimentalists who
may extend or shift their working hours to coincide with the requirements of their experiments. There are
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
. ? ./r.iksiotsr f EVEC 52 DOC.# 89-1W lines 20g - ?De
1TF DOHERTY YFI, V006 WJDRPK PAGE II HIST 08/07/78 17:17
ApprovecFFor-Release 2001/07/12-:--et-A-Repaiveo93-3-Reees- e01 2Oeff1=5-
also scientists who work more than eight hours, or are simply more productive if given access to the tools they
need in the middle of the night, on weekends and on holidays. Finally, we realized that if users are to entrust
all of their programs, procedures, documents, and data to the interactive systems then that information must
be available at least in the same way as from a locked file in the user's office. To limit the user's access to his
files is simply not satisfactory from his point of view. Consequently, Computing Center management made
the decision to offer computer services, insofar as possible, around the clock. This policy is at times
incompatible with the needs of the Computing Center to test new software at night, modify or repair
hardware or carry out necessary housekeeping tasks such as making copies of the on-line data base as a
safeguard against destruction. But, given a commitment to provide service around the clock, compromises can
be worked out so that many hardware changes are postponed to third shift or to weekends. Together with
our users we have evolved a strategy for software testing so that changes to the software system are
introduced on pre-scheduled days, usually during a shutdown at 18:00 lasting 15 to 20 minutes. The period
from 18:00 to 08:00 the next day is then designated as a User Risk Session which informs our users that new
and possibly unstable software is being tested. The Computing Center is then allowed two system crashes.
After the second crash, the standard prime shift system is restarted, and the users know that no more testing
will be done that night. Housekeeping tasks can usually be designed to run in a time-shared mode so that
back-up of the data base can be done after 18:00 even though users are on the systems. Such tasks take
longer to complete when executed time shared rather than stand-alone, but stand-alone use of the computer is
employed very rarely because it limits user access to the systems.
When new software changes to the operating systems have been tested for about a month on second and third
shifts, they are then introduced on prime shift, usually on a Friday so that we have the weekend to solve new
problems which may appear. In general, after one prime shift failure of the new software we return to the
older standard system. If the new software runs for a week without prime shift failures it becomes the new
standard system.
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
" -fie'Ser as e A 1,i^ser - 571-1 ott
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
Unscheduled hardware failures normally call for immediate repair in many computing centers. In general, our
policy is to restart the system and maintain service even if it is degraded. Only if the failures become so
frequent that the user cannot proceed with his work, or if we cannot restart the system, will we give up the
system for service on prime shift. Relocation of hardware and the addition of engineering changes is
constrained to second and third shift periods or on weekends, excluding 08:00 to 18:00 on Saturday if
possible.
Despite our intentions to provide around the clock service, a major difficulty is that of operator availability.
Because we always operate under manpower limitations, we were led to seek new ways to increase operator
availability with a fixed number of computer operators. We realized that we could not operate each machine
from its own console and continue to mount tapes and disks on demand with the operations staff we have.
We, therefore, designed a central ''operations bridge" from which all three 370/168's could be operated by
means of IBM 7412's and 3277's, and made changes to the operating system where necessary so that systems
could be started and controlled remotely. Since all of the operators must be trained to operate all of the
systems, it is much more difficult to develop operators highly skilled on all systems. Consequently, we have
an MVS/TSO/JES3 operator team with a lead operator and a VM/370 operator team with its lead operator.
Members of one team can operate the other system, but do not necessarily hav? specialized knowledge of
their secondary system. This strategy does allow the Computing Center to operate on weekdays with six
operators on prime shift, five operators on second shift and three operators on third shift. On weekends and
holidays there is ione operator on each of two 12 hour periods per day. On certain holidays, such as
Christmas, New Year's Day and special IBM work holidays the interactive systems run unattended with the
understanding that private tapes and disks cannot be mounted. Nevertheless, there are people who come into
??
the laboratory on holidays to use the computers, or who, in a limited number of cases, dial in from home
terminals which represent yet another way to extend system availability.
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
\TRSION- 33 -LEVEL- 52- - DOC.#. g859 Siiiie253---- 277
TF DOHERTY YFL V006 WJDRPK PAGE 13 HIST 08/07/78 1717
Approved Fur Release 2001/07/12 . CIA-RDP84-00933R000 00120001 5
Another important aspect of system avails blility is that of management flexibility. Computing Center
management must be willing to make compromises in schedules in order to serve special needs of users.
When Computing Center schedules are published each week, or when sudden changes to those schedules must
be published by means of logon messages to the users, we invite users with special needs to contact us. If a
user is attempting to finish a paper to meet a deadline, or simply needs to get an urgent piece of work done
by a certain time, it is important that the Computing Center try to adapt its needs to the user rather than
insisting that users always adapt to the Computing Center. On the other hand, it is important that schedules
promising service at specified hours be adhered to. It is extremely frustrating for a computer user to come
into the laboratory on a weekend or holiday expecting to use the computer and find that the Computing
Center has changed the schedule even though the reasons may be good ones. We have also established a
special telephone number from which the user can obtain a recorded message on the status of the systems.
This message is changed and time-stamped whenever a system failure or other emergency changes the service
schedule. The main disadvantage of the recorded message is that it is just not accurate enough for brief
system failures although it is useful for failures lasting more than five minutes. Our users tell us it is accurate
no more than 70 percent of the time, and we are seeking a better way of informing users of the status of the
systems.
Performance Management and Visibility of Service
System response time is the time measured from the user's signal to the system that there is work to be done
to the point at which the system begins to present those results to the user. It is time the user had to wait for
results and is important to his ability to continue effectively with his work. Figure 3 shows a picture of the
-?
response time to a single small program which ran every 20 seconds in the spring of 1971. Each 20 seconds, it
awakened, and recorded the number of logged on users and the response time that it received from the
system. Figure 3 is really a 3 dimensional graph. Each plot position is a number indicating the number of
times a specific response time occurred.
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
1-11- 1)011111-CI Y Y1 L V006 WIDRPK
PAGE 14 HIST 08/07/78 17:17
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
Notice that as the number of simultaneous users increases, the response time steadily and sharply degrades.
Thus, if we simply try to maximize the number of simultaneous users, we find that we are maximizing the
number of people to whom we are presenting a sharply degraded picture of service. This means that the
number of simultaneous system users is a misleading measure of system load. The system expansion facto...,
rather than the number of simultaneous users is a meaningful measure to the Computing Center and the users.
It is defined to be the ratio of the actual time to do a unit of computer-limited work to the minimum time to
do that work in a stand-alone environment. We will discuss the expansion factor in more detail below.
We find that significant performance improvements lead to a reduction in the number of simultaneous system
users, perhaps because the user gets done with his planned work sooner. This is accompanied by an increase
in the number of interactions processed, but the overhead needed to do that work tends to decrease.
Figure 4 shows the profound influence that system response time degradation can have on user behavior. If
we break an interaction into two parts, the system response time (SRT) during which the system is processing
a request for the user, and a user response time (URT) during which a user is keying-in his or her next
request to the system; then each second of system response degradation leads to a similar degradation added
to the user's time for the following request.
This phenomenon seems to be related to an individual's attention span. The traditional model of a person
thinking after each system response appears to be inaccurate. Instead, people seem to have a sequence of
actions in mind contained in a short term memory buffer . Increases in SRT seem to disrupt their thought
processes, and this may result in a time consuming reloading or alteration orthe short term memory buffer.
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
? ? 'VERSION 53 invEu 52 Doc# 88s9-rns (Ines 302- - 32C -
1 11- 1iIM1.12.1{ Y YIL V000 NV.IIJKYK.
FACJI:: 13 1-1131 U25/0// /8 11:1
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
This phenomenon was first discovered by S.J. Boies (7) at Yorktown in 1971 While doing normative studies of
interactive computing. We had recorded every interaction on TSS for three years, and this data base included
a wide range of human tasks. Boles' phenomenon was detected during most types of work.
. It is interesting to note that controlled behavior studies going back to 1938 (8) showed that three things
happened to people in a stimulus-response situation when the stimulus became both tong and erratic. First,
they slowed down in their work, second, they became emotionally upset; third, they made more mistakes.
If we were to tolerate a system responce time as long as two seconds in our laboratory and assume Boles'
phenomenon occurs, it would cost us a minimum of 36 million seconds per month of lost human time. That is
10,000 man-hours, or 60 people lost full time for the month. It is our objective to keep 90 percent of our
interactions to a response time of .5 seconds or less. Subsecond response time is an important human
requirement.
Figure 5 shows the distribution of user response time (URT). Others have called this think time. The mode
of this distribution is 2 seconds. The median is 8.5 seconds and the mean is 12 to 15 seconds, depending on
when one terminates the data points. If all end points, such as the two hour lunch break, or the time when
users forgot to log off, are included then nearly any average is conceivable. In the extreme case what is really
being measured is how long the system stayed up. But, by truncating after two minutes we find an average
URT of 12 to 15 seconds, and we have observed that average for the past several years. The high frequency
of points in the two second range may reflect the fact that people scan far more information than they read,
and read more than they write.
Figure 6 highlights the impact of scheduling on the man-machine interface. It shows the impact of the
Resource Management PRPQ for VM/370 in a benchmark environment as it would affect users. The
columns titled RM PRPQ show the results of running with the PRPQ in an heavily loaded environment. The
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
? 'VERSION- 53 -LEVEL 52- DOC.0-8859 Mg lines 321 -
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
1
columns titled RELEASE 3 show that same heavily loaded environment without the PRPQ. Notice that even
in this heavily loaded environment 50 percent more interactions can be handled by the system with the PRPQ.
And the response time at the 90th percentile is 20 times better. Notice also, that the base system handles
more interactions with 60 people than with 80. By allowing those extra 20 users onto the system in that
overloaded environment we observe that we have effectively lost the 20 people and slowed the other 60
down. In other words, without a knowledge of the expansion factor the number of simultaneous users being
served is a poor measure of the load on the system.
Figure 7 shows computer cost on the vertical axis and the expansion factor on the horizontal axis. If the
machine is underused, then most of its cost may be lost dollars. As the demand for computing services grow,
the expansion factor also grows. At some range of expansion factors, the system is being optimally used in
the sense that as the load builds beyond that to higher expansion factors a significant portion of the hardware
usage goes into unproductive work.
Figure 8 shows the lost hardware dollars per hour due to underload and overload on one of the VM/CMS
systems at Yorktown. The V shaped curve is an envelope containing the actual operating points. The
expansion factor used here is an actual elapsed time expansion factor. It is the actual time to do some typical
unit of computer-limited work divided by the minimum time to do that work in a stand-alone environment.
Figure 9 shows the cost of the lost end user .time per hour on the vertical axis. The horizontal axis shows the
expansion factor. The expansion factor grows together with the number of simultaneous users, and the
number of people being slowed down by the delivery of poorer service to each also grows.
Figure 10 shows the combination of the hardware operating points and the end user costs as a combined loss.
Note that there is indeed an optimal operating range, having expansion factors in the range of 4 to 6. This
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
. VERSION ?53 LEVEL 52 DOC:4 8/459 EDS fines ?341 - 361
TTF DOI IERTY YR_ V006 WJDRPK
PAGE 17 HIST 08/07/78 17:17
Approved For Retedbe 2001/07t12-.-er-A-Riape4-e0930005 1
optimal operating range is dependent on both the operating system and the machine speed. The cost of
overload is far greater than the cost of underload when the end user's time is considered.
We use VM Monitor to monitor the daily use of our VM/CMS systems in Yorktown, and we attempt to
maintain the operating point very close to an expansion factor of 5. Note that this definition of an expansion
factor, using elapsed time to stand-alone time, differs by about a factor of three from that reported via the
IND command in VM/370. The IND command normally reports a number that considers all users in a single
server environment.
W. H. Tetzlaff has developed a very sharp set of analysis programs to analyze the service each user receives
each day. These programs are run each night, and give us a clear indication of the quality of service, what
caused problems, and what the relative pressure is on each system resource. His programs calculate the
expansion factor for each hour of the day. These programs make the business of tuning, capacity planning,
and service analysis a fairly straightforward task. They are described in his paper 'State sampling of
Interactive computer users'.
W. J. Doherty [9] has studied these issues extensively and has defined the performance of an entity (a
function, a system, etc.) as the degree to which the behavior that entity meets the observer's expectations.
More precisely, performance is the difference between the observed behavior of an entity and the observer's
expectations of behavior. We have found that if we can give the computer user some notion of the current
state of the system he is planning to use, his expectations are modified accordingly and he can make sensible
decisions in planning his work. For example, he may elect not to log on to the loaded system but .to do his
computing later when the expansion factor is lower. This decision on his part avoids making unsatisfactory
service even worse. In TSS we introduced a numerical measure called the THI (based on the U. S. Weather
Bureau's Temperature-Humidity Index of human comfort) which was an averaged value of the time needed to
perform a standard task. The TI-Il was available to the user before he logged-on, and he interpreted that
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
? ? ? ...eva &LEN,. 1 itt. V VV,I VYJLflfl AfltiC. 10 1-113 I U0/ 111/10 11;11
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
number in terms of what he was planning to do, e.g., editing would get reasonable response even with a
relatively high THI, but a PL/1 compilation and execution might be unacceptably slow under the same
conditions.
' It was reasonable to carry to VM/370 our experience with the THI but it was also desirable to give VM users
more information about the status of VM than the simple THI of TSS. Accordingly, P. H. Callaway [10]
implemented a THI for VM. The user must logon to VM/370 to obtain this information, but VM logon is
usually very rapid and consumes very little resource. The components of this THI are available to both the
control program and a user program executing in virtual storage. Thus, certain maintenance tasks such as
data base back-up or data migration can check for low system activity before proceeding, which tends to
minimize their impact on the user community. We have also found that by giving the operator information
about system overloads derived from this THI, he can ask dominating users to postpone or moderate their
activities, and he has been able to detect program looping not apparent to the user. The Watson Research
Center version of the VM/370 THI together with other information collecting facilities we had implemented
were made part of VM/370 Release 2 as the VM/370 Measurement Facility. VM/370 THI is now known as
the Load Indicator.
Data Management
In 1965, APL on the Yorktown Computing Center's 360/50 imposed rigid constraints on the size of the
user's data base: 36K bytes of working space of which about 32K were available to the user, who constantly
protested against those limits. At that time the only way to increase work space size in APL was to provide
more main memory and/or reduce the number of workspaces in main memory simultaneously (thereby usually
increasing response time). Until APL/CMS became available, APL work space limitation was an annoyance
to many users. On TSS/360 the situation was radically different because to its users TSS appeared to be a
one level store. To the limits of allocated disk space, the TSS user could create data sets and file them and
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
?Ci.funct,,?Io tit tor neNr 4 00411- Une Arlo
masa tobakv.A.:
TTF DOHERTY YR. V006 WJDRPK
Appro
1 1 1 1 1 1 1
PAGE 19 HIST 08/07/78 17:17
84-00933R000500120001-5_
retrieve them by data set name. The system managed all direct access storage space for the user. We could
not continue to add on-line disk storage to TSS, but we also wanted to avoid requiring the user to spend time
moving data sets to and from private tape or disk volumes. We developed a data set migration facility for
TSS which became operational in April 1969, and had the following characteristics:
(1) Data sets were marked with a date last referred to.
(2) Data sets not referred to for more than N days (the value of N determined to keep a safe amount of
on-line space available) would be compressed and migrated from the on-line disk volumes to
demountable disk volumes stored off-line.
(3) If the user referred to a migrated data set, the system informed him it was migrated and told him how
to get it back.
(4) The user could list the names of all migrated data sets, voluntarily migrate a data set not needed for a
while, and erase unneeded data sets.
TSS Data Migration met many of the user's data management requirements but had some limitations. There
was no mechanism by which a user could specify that any migrated data set to which he referred be restored
automatically without additional action. This was an annoyance even though the user was given a message
containing the name of the migrated data set. If the user knew that he must restore a migrated data set, he
still was required to initiate the restoration 'process and to wait until restoration was complete. Most users
preferred that restoration proceed without locking the terminal so that other work could be done. We have
learned that for future data migration systems, when restoration is completed the user should be informed.
With the introduction of CP/67 and later VM/370 at the Watson Research Computing Center, the interactive
system user found that the burden of on-line space management was transferred back from the system to the
User.
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
D011ER1 Y YFL V006 WJDRPK PAGE 20 HIST 08/07/78 17:17
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
(1) Because he is given a fixed allocation of disk space, the user might have to stop in the middle of his
work to decide which files to erase or move to tape in order to make room for new files being created.
(2) As his files grow, the user must try to persuade installation management to give him more space; on
the other hand he is pressed to give up space he is not actually using.
(3) When the user is given more 'space, his previous allocation must be copied by the Computing Center
into the new space. This is an overnight operation, and usually requires that many users' space
allocations be copied even if the}, are not changed in size.
By late 1972, the Computing Center had more requests to add new users than we had available disk space.
Since user disk space on VM is frequently only partially used, and since on-line disk space is only fractionally
active at any point in time, rather than invest heavily in additional disk storage devices, management explored
two strategies. First, and least acceptable to the user community, we imposed an external schedule which
distributed the available system time among the users by mounting only certain subsets of users' disk files at
certain times. This not only constrained each user to specific hours, but, frequently made it impossible for him
to access files of a colleague if those files were only available at a different time.
Second, we began a one man development effort to determine whether or not user files could be managed by
the system so that only "active" files occupied on-line disk space. This design was very promising and by
1973, we decided to carry our experience from TSS to the VM minidisk concept in VM. The most convenient
unit of space on VM is not the user's file (although we subsequently implemented file migration) but his
minidisk, a specific number of contiguous cylinders of on-line disk space. VM Data Migration as implemented,
at the Watson Research Computing Center moves and compresses an .ntire inactive minidisk on to a
demountable disk, thereby leaving an "empty slot" on-line into which an active minidisk of that size can be
moved on demand. As in the case of TSS, the decision to migrate a minidisk is made on the basis of a
specified number of days elapsed since the minidisk was last accessed. The user whose minidisk has been
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
ttirtiVahr 1TVvrr-cl -rm
'TTF DOHERTY YFL V006 WJDRPK PAGE 21 HIST 08/07/78 17:17
ApprovedFOr Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
migrated experiences a longer than normal delay during his first logon while the minidisk is located, decom-
pressed and moved into an empty slot on-line. Subsequent logons will proceed normally unless the user again
fails to logon within the specified number of days.
? Since the minidisk migration task must be available at all times, we implemented an AUTOLOGON facility
which automatically logs on specified tasks after a system restart. AUTOLOGON has had value far beyond
its role in minidisk migration; as a result of demands for asynchronous processing mentioned earlier, we find
there are many special tasks which carry out important services for our users, e.g., the NETWORK task
?
enabling users to send work between systems, and which required manual restart by the operator. Now they
are automatically started at system load time. If the operator failed to restart these special tasks, the user
found these important functions unavailable.
The file migration capability which we have added to 'VM/CMS enables the user to migrate a file by file name
from his minidisk to a demountable disk. In this way he can save and retrieve CMS files.
A valuable consequence of VM migration is the back up capability it affords the installation. On third shift,
copies are made of all minidisks accessed on the previous day, and are stored on a second set of on-line disks
which are copied to tape. If a user inadvertently erases his files, within four hours the Computing Center is
able to produce a copy of his minidisk which is no more than twenty-four hours old. The Computing Center
has very rarely lost a user's data; nearly all data losses take place when the user erases or inadvertently writes
over his own files. We believe that the Computing Center has a continuing responsibility to safeguard the
user's data against both accidental and unauthorized access or destruction.
Another major benefit of the data migration facility is the reduction in cost for managing on-line data. Just
before the introduction of these facilities for VM/CMS in 1973, there were about 350 tape mounts per day
and 200 disk mounts per day across the whole installation. By early 1978, when the use of computing had
grown by a factor of 8 times, we were processing 30 tape mounts per day and 5 disk mounts per day. Thus
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
If ? ? gal " ??? a 15, ? "Oi
TTF DOHERTY YFL V006 WIDRPK PAGE 22 HIST 08/07/78 17:17
Approved For Release 2001/07/12 . CIA=Riapazroonancro5oatmert-
- -5-
an effective reduction of at least a factor of 80 times was achieved in the frequency of tape and disk
mounting.
Distributed Function
A user's CMS vitrual machine in VM is well isolated from other CMS virtual machines in the sense that a user
may cause his CMS machine to "crash", but does not affect others. Therefore it was our belief that the CMS
machine would be an excellent vehicle for the development of special functions which could then be tailored
for ease of use, ease of maintenance, and good performance. The strength of VM lies mainly in the strong
isolation provided by virtual machines.
Because we could use CMS to isolate a particular function from all other functions, the user need learn only
that function to use that function. Thus, complexity in the end user interface is lessened. As a function
evolves over time, tradeoffs in efficiency and ease of use can evolve together without being clouded by other,
non-related issues.
Because the function is isolated to a virtual machine, it becomes easy to add new versions either as copies or
replacements. This greatly reduces the sensitivity of the whole system to changes. In addition, the VM/370
system automatically re-IPLs itself in most cases if it should fail. It then automatically re-IPLs all such special
purpose virtual machines. This is normally done without operator intervention and typically takes about 30
seconds of elapsed time.
Because the users of interactive systems grow in what they use the system to do, we find that interactive
growth results in a proportional requirement for asynchronous processing, i.e., the ability to do computing
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
TTF DOHERTY YFL V006 WJDRPK
Approved --For
PAGE 23 HIST 08/07/78 17:17
-iii iv 'Vie
which is separated in time from the user's interaction with the system. This had often been called batch
processing in the past. By grouping special purpose functions in special machines, we can apply better
controls to those functions. i This includes special scheduling controls, grouping of requests to reduce
overhead of initialization, and tailoring of the function over time in accordance with the user's evolving needs.
Also, by giving the user the ability to submit work to asynchronous virtual machines, we remove bottlenecks
from the user's interface to the system and his data. It is important to note that these special purpose
functions need not run on the same real machine as the user is currently using.
Thus, the virtual machine concept directly applies to distributed processing. The gains to be had for load
distribution are real, but they are small in comparison to the gains in reduced complexity, ease of mainte-
nance, and improved service to the end user.
Our experience with distributed function leads us to the conclusion that we have now seen three distinct
stages of computing center activities. The first occurred in the early sixties, when the computing center
provided computing services and served as a pool of applications programmers allocated to help scientists
develop their programs to use the computer to solve problems that were machine limited. This was because
computing was about 100 times more costly at that time, and so only a few applications were cost justifiable.
It was also due to the lack of data management and primitive user interface then in existence.
The second stage was that of the system programmer, when most computing centers grew rapidly in size and
the systems did also. Highly skilled people were required and change management grew increasingly complex.
The computing center personnel were devoted to make the systems more manageable. However, their time no
??
longer went directly to the end user. He became a direct user of the systems.
The third stage is our current environment in which the systems personnel are primarily involved in some
basic system changes and the management of tools for distributing he responsibility of change control to the
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
TTF DOI IERTY YR. V006 WIDRPK PAGE 24 HIST 08/07/78 17:17
Approved For Release 2001/117/12 : GIA-RDP84-00933R000500120001-5
users who have the strongest need for the changes. In this way the end users themselves can better control
their own environment. Our networking capability allows our users to communicate freely between one
another and send new tunctions directly to each other in machine readable form. Our users send an average
of the 3,000 files a day in and out of Yorktown computing systems. The typical file size is 50,000 characters,
which is the equivalent of a 20 page paper. That is a measure of the communication among users on
physically different real machines. In fact, the communication among users of the same physical machine, but
different virtual machines is probably far greater.
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
? a ir-ar? Soi ? un tic
?
DOHERTY YFL V006 WIDRPK
Approved ForRetease-200-tfOttt2-.
Summary
PAGE 25 HIST 08/07/78 17:17
? 11
III
Appropriate management actions can significantly enhance the effectiveness of an interactive system from the
user's point of view. During more than ten years experience with interactive systems at the IBM Thomas J.
? Watson Research Center, we have made management decisions which extend the capability of these systems
to work for the user by enabling him to make sensible decisions about how he will use his time, and by
selectively adding to the systems facilities which enable him to spend less time compensating for system
limitations and more of his time on his problem. Increased emphasis has been placed on using the computer
as a tool to extend the users' memory as well as their reasoning power. Our computing systems are adapted
by experienced users, via EXEC's, so that the effectiveness of man-machine communication is increased with
time. This is of special benefit to the inexperienced user who then takes advantage of the experienced user's
evolutionary growth. Issues of availability, distributed change management, data management, inline
documentation, response time, expansion factors, distributed function, and effective user to user communica-
tion have been key to our success. VM/370 and CMS are the primary vehicles that our users have found to
be effective for our rapidly evolving interactive environment.
Acknowledgements
The managers of VM/CMS at Yorktown Heights have contributed much to the IBM product in addition to
effectively providing excellent service for the Yorktown users. These include W. M. Buco, A. N. Chandra, B.
Lie, and N. J. Pass. W. H. Tetzlaff and P. H. Callaway have contributed much to the performance manage-
ment of VM/CMS and their work is gratefully acknowledged. L. H. Wheeler has developed scheduling and
resource management strategies that have sharply increased our ability to provide interactive computing
service to a large number of people at Yorktown.
REFERENCES
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
? ? ???? 11--F4 IL, f?CONT, 01b en 1-es0 11? - ?/ ? A
Approved For Release 2001/07%12 : CIA-RDP84-00933R000500120001-5
[1] A.W. Luehrmann and J.M. Nevison, Computer Use under a Free-Access Policy, Science 184, 957-961
(1974)
[2] W.E. Daniels and R.W. Ryniker, EXEC 2, A Computer Language for Word Programming, IBM
Research Report RC 6292, November 17, 1976
[3] W.J. Doherty, Human Factors: Impact on Interactive Computing, Proceedings of SHARE 50, vol. 2,
1244-1266, March 1978.
[4] Private communication between t. Johnston (SLAC) and W. J. Doherty, Feb. 1978.
[6] T. R. Bell,
[6] A. Guido and J. P. Considine NCC paper on TPVM
[7] S. L Boles and J. D. Gould, User Performance in an Interactive Computer System, Proceedings of the
Fifth Annual Conference on Information Sciences and Systems, Pg. 122, 1971
[8] R. S. Woodworth, Experimental Psychology, Henry Holt and Co., Inc., 1938
[9] W. J. Doherty, Measurement and Management of Interactive Computing, Proceedings of SHARE
XLIV, vol. 3, 1587-1598, March 1975.
[10] P. H. Callaway, A Performance Measurement Approach for VM/370, IBM Research Report RC
4666, January 4, 1974.
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
VERSION-53- LEvEL 52 ---DOC:if rDS lines 540 - 334-
- - Approved For-Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
100
80
0-0 70
60
>50
-I 40
2
F., 30
YORKTOWN. eiCZ
20
10
romMIIMI?
1???11/11011
*111111111110
???????Ir
100
90
80
60
50
40
30
2.0
10
1 1 I 1 I f
0 10 20 30 40 50 60 70 80 .90 (00
CUMULATIVE % - USERS*80,' (1)&00
-,
'
FIGURE JO
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
0.0.0 ????? ????? mn?
 :-1,,... , ?
-I ...
7:
. ? I . '
.
' 'APO
Ved
.
Ftit:ROteate
..-
20611
7112: '
; , ,
CIA:-FiDP,84
? 1 ' ? ?
0206[12496)05001.12:b
'0.1i-5 ;
....... .......
I
' ; : ;I: ; ?
.....
-;;;;::.
.
.
..
I
, ? ?
. ' ....
..
1
.
.
.... ; . .... ....
? ' : : 1: : " I 1
:: ..... ::
.;::.-
, . , , .
-?-:+:-:-:
" ' ' ? "
:::1::".
- ; ? ?
. .1..
.?. ... ::
?
: -
1
...., .
?
....... ?
77:7 : . ? -- I 7: 7'1 :
"?"...' ......
.,
T
'
:..i::::
::::: . ;,..:;?;..;
: :. :-. :1-7,1
, : ? : ?
'.. ........... ' :
,i-:. - - I' -71
' ? :1'
_I ...
'7: ' 17 ' -
? ? , . ? ? ?
... t .
i ''-- '
? . ? ? , ....
.........
7 7771: 7 ::
.... : ....
... '
I
? I' ? I
1
1
,
'
.
1
.,. ..
1
.. ; .............. .
? I
, ... ,....
1
- 1" "
.,
99,
- .....
,..., .. ?
099::?:
1
???
.........
::.:::,::::
I
. ..
.
....,....
::::i::::
'
.-71-71-77 ..
:::1::::;::::1::::
....
1
::.....,!:?::-:
...........:-
' -
..
'
'
-.-
?
.....
??i,
.......................
'
? ? ' ? ' '
f
l? ?
.
7
?
..r. . :
-
,
?
? . ::ti:
...
.. ? :
? t. ..
. : :1
-
.... ... _
MONTH
. .. .
Y .: ......... 'I ? ?? ?
... .
...
? ? ? ;:- ?: i i ?
. ......
? ?? / /: '?
.......
.. ......
?????-??
....... ??
.... ; ....
. .
???.:::_.
I .1.
.. . .. .
.
:::: ::::
.. . ...
.. ... ...
.... .
. .... . ......
.-.:_:.:::.: ........
? - ? -
. .. .
,.
' CO tI1JECT
s'
?.
=:_:.:?:.:-:_:1::...-::
.80,000---
-?-????--
-- ...... ?
,
??::???: : . :-::?-
?-- .....
-----
t ?
: 7 .... ? ?
. I..". .. -
.1.. 1
.. - OURS
- .......
:.. . ........ .
....... ... ::: i
.. . ..... ..... _
1,:. . . .
. ...
.....
...
...; ....
....1......
.
.. -t::::
--,---7:..:::
..:: _: . : ..
....
:_ : I:77.7:71:7: '
: ::-..-:-.::::-tzr: ::: ? -
..t...-.:::.
- -4.
? - --- -.:.r=::::::.:
r ?
- if.-_,?_-- . .? :
i- -'
,
- -7J-:_:::::::::::
:::.:.:_::::::..1-=___
_

-- ? --- ? --
---:::::_t:-_:-_,-:-.7.z.
t -::71:-.:.::"7:::
_-_?-:_-_--...;___---::::_::_:_-
4.._._.---_:,:_,- _ .
" 1;-.
-
70; 000.7-.7-
.7_,:-_:_:-::-...-..-_- : t :-.:_--:
_ . . :.--....- - -
' -

-
-_ . ....... ...,....
.4 ----:::==----7-=7---.-..=
-
.-"-7.--..t.7---':--1-
. ......
? -----`
.. .. -- ..
-
..:...1.:...:_:'
+--
. - .......
. -
-- - --
----- =....-- --
-1.
-= : _' =
----*---- - --
T

--- ----1-:::-.--
-_-_-z...4...
--
,-1
-- ---
,
*
--- '' --- . ---
. !"
__
-__
:-=:: . _
........ ...
____.
_____ _ _ _ _ ..,__
,
?
60 ;00 a.?
f ______.
.,.........
,..'1{OUR3--- r
-E-----
.__.I.
1 1 '1
--,
4.
.=--- .
-{.--
- . i......::.. YEAR
,
:
? 50; 000. MONTH --
-__ ____.
., ...-
._t___________
-
i----
MON 1-tL.T__-_.=_.
71 -...::.7. : :::: -_-?.=_-_
_, _ - _ - _- . _ .
.------=.'=--t-P'-----
--.:-. ?-=
1
7771.tSER
,
.___.,...... _ . _
-
t
140.; 000_
?_,
i1
- .--'--
- --? -71
.I
_
,
_,,,..
.
-- - -- -
---
-^''
4
__
--.,.
'?--- -
- --
-----,
?
_
----"------'
------
--:----*-- -
---4------?-- -t
. 30"; 000?
_
47.-...*..-
I. 4
I
- - =t
. '...
L . ''-.
-44...4.44
444.44.4..4, .4...4.
4-.4444I
-4.-44.44.4.44..44=7.:
-.-......-4.: . ...-.-
.
T
? --
....
z.: =:=:-...= -
:_..-=::: 7 : :
=7_1 : ==-...==
.7" ...-_:=1--.._ _7.
=17. 4.:::: : : . ?
' 0,i 00
0
-T
I
, L
.-,-.- ..
----7.=--
. .-
'
.
D

-- .-
---
__
, _:.:TOTAL
_
- -
- - __..._
?
...7..: -"...: : : : i:
F-
:::::-.? ::-.---.7.:-.:=..
T
?
-.I.-----* ,
" ::::_:#=-._;.:'__
'

_:
- ? -
-?
DP HARDWARE.
.S.
?.......--.1"4
10,000E:
11_
?
, ,__
?..-
.:
i.:.:
__. ...... .
.....eir.......-:.,
..
... ..
- :._:?- 4-_::::::-----
_ :-.I. _,
- -
---.:-.:-?
...........
_
I
--
...
. .. , , .

......
1
_
- ..-
...
---:7--.......:
. - .
'''''. ..:*-:..

? ......... : .....
. -.
- .. -
.....-
7 ---;.:Z
-7.1-7:::?.7 1 1: - :--.77 -7.=l :-
...t-,..... .
. ? .
:t.. ..
,...:17:_. ,
: 7: 71 7:--7
7:7.11.7-4.7:
-..:' ? 1 ' - ? --..7 1
-:-. '''
":-:.:7'..:7"
. ' ,
19 3 : . .... 1..9 ' 1915 ? ? ? 19'i6 " V ' 1977
.... ...., ....
--
.
- '''''
:::::.:-::
1;:.:1-:.?- .. :: : : :::1::::
:::.....,....z..-1:... ._....-
"" ?? 7:
:::7::::::.
...... ?? ?? ? ? -?? - -:.?1::?: r ?? ?'
::t:. .. ? .... ...... ...
r
,
.......
7 ...............
: .. ....1?.-...
- - -.. ........... . - .
I 1 , . , ...... .
,
..-- ::.1:77.::: .... . .. . ? '
? - . IBM.: T .J. WATSON COMPUTING CENTER:::.
-.v.:::
,
..... -
' '' " ? i
,:t::::
--..
' -"" ? - ? ....... : : 1 : '
::::1:::;
,- ,
?-.1.'
. ,....
.... : .
...., 7".
?
: . '.'. ?
...
........ :.. .? r ? ? ................
? ?--i? .-
..:... .1 ? . .. ..1.....: .. : .
.. ,...
.' .
....;
....
COMPLIT-I?NG-COST-S--srs-r-USER-?
, ; ,
1 1 ,
COSTS
. t .
. .
...
L.....4.1:: ? t- ? ???
" "
.... .. 1 . ; .........
:????1;:?? ..... ?
..:
... ....
??:::::-::
..................
?
_...
-
1
. 4. 4 4
___
* ,*
-r
,
......... I ...... . . . ! . 1 '
r
? ? 1 ...
I I'
-
.J...J..... ... : ....
? *, '.
. .
... . . l? ..
... .: . ... . ....:.1.: ... :I.::: ??.
1
? ? : ? ?
I
: : :
.. ? ..
. ? -
. C :
... . . .....
..........:-.:::?
......r.::?
? ? W
' '
i...
?1 7 ....
.:::,...;
!.
1.?;:',. .1:.::',....
y.....i:"..... ... .. .....
? .
....1:.:..7"---::;:..?....,
... ...
.
..... ....
..-
...I.
? ? 1. ' .. ....
.....
1.
,
,: .
NO. OF SIMULTANEOUS USERS
Approved For Release 2001/07/(12 : CIA-RDP84-00933R000500120001-5
13k- 68A25111 1 I
- 8A3262123
-136535331
,, - 3536432
1 f - 1272523
r 125 1211
21111
4 1
1
2 11 1112 1
2
2 121111
11
2 21121
2
- 13125 41132 1 2 I 2 1 11 I
- 143121 3 2 I 1
,,,,,, - 1213 21 431 1221 311 I
LL - 241 311321 1 2 II 2 1 1 11 "
L1 131326231 12 22 33 1 1 1 1
-11111331 111314331 2 I
- 11332333411123241 2 3 II 1
- 12 23 136231 1112313 1 1 1 I I
27- 12542 3 III 211 1 1 1 1 1
- 1421114 1541213 6121 2 2 1 ? 1
- 442213323132323132121 23 I 1 1
- 1224 23442134213 41531 1 1 1 I
- 171334184222455546116 1 ? 11 2 1
32- 11322 22414324572332 51 11 1 11 1
- 1331 341223323335313346111 211 I 11 I
- 111122 343 5731363425 1121 1 1 1 1 1
- 22 2133 1 2423C68142221331 II 1
? - 212243132224541923 212 1 2 2 1 1 1
Of - 11112 1112463624A69A25234 1 31 1 $22 121 1
22541 1 22316383754382 22423 2111 I
- 2 223322377655453 2113131121 1 1 I
332132377324985342241 12 1 II 1 1 i
2 113 221433647331352 12 12 1 1 111
42"- I I I 212458534 335241 14 12 221 11
1 1 52 441273333752415114111 2 1111 4 1
3 3212213412 112113 1221 23 21' 1 1
l-- 1 2213812544244311231413 21 2
1
1
1.
? 1 II 1 1
1 1 1 I
1 II 1
1
1 1
2
2 ? 1
Ill
I
I I I
21 1
II
1 111 4112 22 31122221111211 1 1
47r- 1 11 11 1. 11 2 2 1 I
52
0 1.2 3 4 5 6 7
:RESPONSE TIME DISTRIBUTION (SECONDS) ?
1
I 1 i . I II
1 1 I II I 2 1 1
I I
1 1 21 1 i I I I 11
1
1 1
1 1
1
II I
?
?
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
??????
?????
USER RESPONSE TWE IS DIRECTLY INFLUENCED
BY SYSTEll RESPONSE TINE (SRT)
cn
30
25
20
15
10
5
10 15 SECONDS
SRI
FIGURE .0- zi
? ?
Approygd For Release. 2001/07/12 : CIA-RDP84-00933R000500120001-5
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
FREQUENCY
FREQUENCY OF USER RESPONSE TIME
TIME (SEC.)
FIGURE 46, trs
??
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
? 158 CAMBRIDGE BENCHMARKS (40 MIN,)
?
WHEELER RELEASE 3 -
MEMORY 2*MEG 1 MEG 2 MEG. I MEG
RESPONSE 0.146 0.624 2,2 0,537 1.6 0.601
TRANSACTIONS 12,374. 11,149 7,921 9,176 7,319 8,062
PROBLEM CPU 60,4% 50.3% 33.6% " 37.1% '23,8% 43.5'L
TOTAL CPU 95.5%. 88.8% 66.6% 97.4% 51.2%
?
DRUM I/O 95% 95% 65%
DISK I/O 570 57, 35% ------------- ...
r?
;
USERS .49 ' . 80 80 60 80 60
FIGURE tla!
?
6.?
??
?
a?-?
Approved For Release 2001/07/12 : CIA-RDP84-0093313200001200015,
?
Cosr
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
4o.
+11011.011.11111?00
OVCIVialo
4:9 si a Ai
4 cop
FIGURE Cr .7
??
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
rp 6c6- 7
itL.4,1434,4*
CpJ
/6
App
14a0
Lor
SYSTM7o
ved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
?
/0.
wt.
????????
. ? :
? i
.- ? I ... _
R itt . /c2
? .?
Mi??????
7 ?'
-
Y
FIGURE *CM
??
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
ApifiRel
/400
? /30 0
? /ZOO
_bra'
_ .
- ? - (76o
Lo5 T Tca
0...7aC)
600
I- .7 76-75-c
?
. ?4100
Hooie
. Joe)
?-?11::16.0
For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
- . -
????
?:????? ? ? ? ??????? ? owns ? ???? ? ???? ? ???? ??? ? ? ? ??
------- 10 g c.c. 1k 11
? ?????
? ?????????????? ??? ??? +so ??????????b? ? ????? ? ? ? ? ????? .m.00 ?
??
???...
.tsy sr./k1 .
??????? ??????? ???? ? ? ??????? ?? ? ?I? ???? ? ???????? ?1111 ????????
? ??? ? NI ?????? ? ? ? ? ??? ? ? P
FIGURE
? ?
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
ed For Release 2001107/12 : CIA-RDP84-00933R000500120001-5
. . .
???? ? ?Ii? ????????? 4??? ???? ???? 41.? *IMO ????????:?????????????M????????????? ???????Mi ?
?????????? CNN.,
??????
???????? ???
a.
S /4 /V- . 16 . ow 074.7 g/74 (24
? ???? ? Y. 1=????????? ? ? ? .?? ??? ????????? ?? ? ww. ?? ? ? ??? ????????? ?????;? ?????? ? ? 0. ? ? y ? ?? ? ? ? ?
SYSTEM EXPANS ION FAcTOR I
???????? ????? ?? mm?? ? ?? ????? ???? .??. ? ? ? 0..40 ? ? 0 ?
? ?'FIGURE ?
?
? ? .
. ? -
? ?
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
-Approved For Release 2001/07/12 : CIA-RbP84-00933R0005001200011
HUMAN FACTORS: IMPACT ON INTERACTIVE COMPUTING
APPENDIX II
5 Page
W.J. Doherty, IBM Research, Box 218, Yorktown Hgts, N.Y., 10598, U.S.A.
This presentation will describe the human side of interactive computing as seen in the Comput-
ing Center at IBM's T. J. Watson Research Laboratory in Yorktown Heights, N. Y. This
computing center has been a SHARE member (PK) since 1958.
Figure 1 describes the mission of this computing center. We find that complexity is the single
biggest problem facing the end user today. Since 1968, we have looped on the last two items in
our mission. We made the decision at that time to stand at the man-machine interface and
look outward to study the impact on people of the computing function which has evolved over
time. We learned of a number of technical and administrative barriers which got in people's
way. We worked hard to remove those barriers.
Upon removal of these barriers, we can make the following observations:
The value of today's computing systems to their end users is in their DATA, PROGRAMS,
and PROCEDURES. Libraries of thousands of programs exist in all major installations. The
user interface is simplified by combining programs together and burying default parameter
specifications inside PROCEDURES. PROCEDURES outnumber all ? programs written in
conventional programming languages.
The issue of integrating the DATA RESOURCE into the enterprise, (building bridges to tie
the different subsystems together) will probably be the dominant driving force in computing
over the next ten years. (See Dick Nolan's paper 'Thoughts about the fifth stage'.)
Figure 3 shows the current picture of the computing configurations at Yorktown. Notice that
all systems can communicate with each other and use the MSS as a common back end.
Terminal growth has been rapid, especially since the advent of the IBM 3270. Laboratory
automation, program development, and text processing have all grown rapidly as well.
VM/CMS is the interactive interface which is the most popular in our laboratory.
At Yorktown, the computing systems have moved from being specific tools to solve specific
problems to a general resource which extends people's memory and reasoning power across the
entire spectrum of tasks they do. This is a NEW DIMENSION of computing use. It requires
the removal of the barriers we encountered to achieve the levels of productivity we have today.
Figure 3A slows the growth in end user time relative to computing costs from 1973 through
1977 at Yorktown. The uppermost line shows the growth by a factor of 8 in monthly 'connect'
hours. This includes dormant end user time, active end user time and special function virtual
machine time. The next two lines down show the growth in end user time at the terminal.
The uppermost tine shows a growth from 8,000 to 64,000 hours per month. It is a reasonable
upper limit on the amount of time that the interactive systems were influencing our people in
their work. The lower line, showing a,growth from 5,000 to 38,000 hours per month, is an
estimated lower bound on the amount of time that our interactive systems had a direct
influence on people's work. The next two lines down show the total computing cost relative to
the end user cost, and the hardware cost relative to the end user cost.
Thus, small errors in DP equipment planning can result in sharp losses of end user time due to
overload.
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
Page 2
Figure 4 shows how complete the growth of interactive computing has been in penetrating the
different departments we have in Yorktown. It shows the number of people in each department
who spent many hours of their time working interactively in Nov. 1973, and again in Novem-
ber, 1976. For example, the first column shows that department 41X, our administration
department, had three people in 1973 who spent from 20 to 40 hours per month of their time
working interactively. It also shows that there were 38 people in that department who spent
from 20 to 40 hours per month of their time working interactively in November, 1976.
Yorktown processed at least 7,000,000 and at most 11,500,000 interactions per month in
November, 1976. By November, 1977 that range had grown to at least 9,000,000 and at
most, 15,000,000 interactions per month.
Figure 5 shows the number of people in the Laboratory who spent different amounts of their
time working interactively in the month of November 1976. Today, over 25 percent of our
Laboratory is working interactively at each instant throughout the day. Over 75 percent of
our Laboratory works interactively in each week.
Figure 6 shows the difference in intensity between an interactive and a batch environment.
Because interactive computing is so much more intimate in its impact on the end user, we have
learned that functional availability must be targeted for 24 hours per day and 7 days per week.
If information is not available from a computing system when it is needed, then people do not
keep all their data online. This also implies that maintenance must be done time-shared.
In 1971, we saw that the number of commands per person doubled in just 5 months, when
counted after procedure expansion.
At the SLAC (Stanford Linear Accelerator Center) installation today, using the Wylbur
interface, there are an average of 25 commands executed for each command typed at the
terminal.
People learn how to use a computer to perform some function for them. They then ask the
computer to remember how to do that. As people grow in experience with such a system, they
combine functions to raise' the information processing. bandwidth between themselves and the
computer. It is hard to forecast how long this.growth will, continue. It seems evident that the
quality of the command procedure facility,'Which is the software glue to help this growth, is
critical to user growth.
Figure 8 shows the. number of different file types on one of the VM/CMS systems at York-
town. Of the 800 people who use this system, this chart shows the files belonging to 164 of
them. In some sense, that's about 20 percent of the users of this system at Yorktown. Notice
that 162 of le 164 users had a total of 6074 EXEC riles. If we look at all other forms of
programming' language we get a total of 3000 programs in source form. EXECs are used' to
raise the level of the man-machine interface by combining programs together They reduce
complexity by allowing many parameter' values to be decided automatically. They redute
errors by eliminating the need for people to reenter the same commands and parameters time
and again.
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
Page 3
As people learn how to do more and more things with computers, EXECs help that growth.
EXECs are probably the single most important facility in the system to aid user growth.
Figure 9 show.s the importance of the software component known as the editor. Editors are
the single system component where the computer user spends the most time. Estimates are
that 50 to 80 percent of people's time is spent editing text, data, or programs.
Editing programs have historically not received much attention as formal computer software
products. However, we find ourselves with a great proliferation of EDITORS today. Carol
Thompson's USER'S GUIDE TO THE RESEARCH CONTEXT EDITOR--REDIT, IBM
Research report No. RA 28A, contains an excellent history of the evolution of one such editor
program.
It seems that the human factors associated with editor programs are therefore of great
importance.
Figure 10 shows a picture of the response time to a single small program which ran every 20
seconds in the spring of 1971. Each 20 seconds, it woke up, looked around at the other users
on the system, and recorded the response time that it received from the system. This is really
a 3 dimensional graph. Each plot position is a number indicating the number of times that
operating condition was found to exist.
Notice that as the number of simultaneous users increases, the response time steadily, and
sharply degrades. Thus, if we simply try to maximize the number of simultaneous users, we
find that we are maximizing the number of people to whom we are presenting a sharply
degrading picture of service.
Remembering how rapidly people grow in their use of programs, it is clear that a given number
of people will seldom ask for the same work to be done.
If we recall that user growth results in a sharply increasing number of programs invoked per
person, and that service degrades beyond some load threshold, we should realize that the
number of simultaneous system users is a misleading measure. It is indeed a measure of
potential liability. It is not clear what else it measures.
We find that most genuine performance improvements lead to a direct reduction in the number
of simultaneous system users. This is accompanied by an increase in the number of interac-
tions processed. There is also an increase in the number of people served over a longer time
such as a day or week. The simultaneous congestion has decreased.
Figure 12 shows the profound influence that system response time degradation can have on
user behavior. If we break an interaction into two parts, the system response time (SRT)
where the system is processing a request for the user, and a user response time (URT) where a
user is inputting his or her next request to the system; we find that each second of system
response degradation leads to a similar degradation added onto the user's time for the follow-
ing request.
This phenomenon seems to come from people's attention span. The traditional model of a
person thinking after each system response appears to be inaccurate. Instead, people seem to
have a whole series of actions in mind, contained in a short term buffer in their heads. In fact,
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
Page 4
they are continually thinking. Disruptions to their thought processes are generated by delays
in SRT. This results in their short term memory buffer being reset.
It was first discovered by S.J. Boies at Yorktown in 1971 while doing normative studies of
interactive computing. We had recorded every interaction on TSS for 3 years. It included a
wide range of human tasks. The phenomenon was consistently repeatable across most types of
work.
In 1936, controlled behavioral studies showed that three things happened to people in a
stimulus response situation when the stimulus became both long and erratic.
First, they slowed down in their work.
Second, they became emotionally upset.
Third, they made more mistakes.
If we were to tolerate a system response time as long as two seconds in our Laboratory today,
it would cost us a minimum of 36 million seconds per month of lost people time. That is
10,000 man-hours, or 60 people lost full time for the month. It is our objective to keep 90
percent of our interactions to a response time of .5 seconds or less. Subsecond response time
is an important human requirement.
Figure 13 shows the distribution of user response time (URT). Others have called this MINK
time. The mode of this distribution is 2 seconds. The median is 8.5 seconds and the mean is
12 to 15 seconds, depending on when you terminate the data points. If all end points, such as
the two hour lunch break, or the time when people forgot to log off, are included then any
average is conceivable. What is really being measured then however is how long the system
stayed up. By truncating after two minutes we find an average URT of 12 to 15 seconds.
This distribution has held up for twelve years now. The high frequency of points in the two
second range reflect the fact that people scan far more information than they read, and read
more than they write. There is a real need for computing system response times on the order
of one or two tenths of a second.
Figures 14 and 15 show the load on a S/360 Model 67 from running a few programs. They
are a conservative two dimensional picture of load. CPU time is on the horizontal axis and
main storage requirements in 4096 byte pages are on the vertical axis. The shaded area is
supervisor state time, and the clear area is problem state time. The area under each rectangle
is a conservative measure of load. It is conservative because the I/O dimension and the paging
overhead dimension is missing. The difference in area varies by five orders of magnitude from
the smallest to the largest. Almost all the variation in load comes from the amount of data
being processed and not from the kind of processing.
A classic problem in our industry has been the problem of benchmarking. This has arisen as a
problem by the confusion between load and function measures. For example, one installation
said that they had characterized their load. They had a beautiful functional distribution of
their work, but had only accounted for 2 percent of their load.
Load, that is, demand on the CPU, I/O, and main storage, arises mainly from the amount of
data being processed rather than from the nature of the processing. Variations in load on the
order of five orders of magnitude are regularly seen for editing as well as compiling.
Measures of function are useful for productivity studies. They are of very Little use in capacity
planning.
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
Page 5
Figure 16 shows service measures for two classes of work as seen on a S/370 Model 158 in
one afternoon in July 1974 at Yorktown.
The horizontal axis shows the level of multiprogramming. It is the sum of the ELIGIBLE and
DISPATCHABLE lists in VM. It is the parallelism in the input stream of requests. The
vertical axis shows two different classes of service measures, depending on the nature of the
work being done. For those interactions which could be completed within 200 ms. of 158
CPU time, the service measure is RESPONSE TIME, and it ranges from 0 to 2 seconds. For
those interactions which were bigger, we use the EXPANSION FACTOR as the service
measure. This is a multiplier that measures how much longer the work takes due to the sharing
of the machine than it would in a stand-alone environment. The numbers across the top are
the number of interactions in each class of work at each level of multiprogramming.
Notice that most of the interactions took place at a level of multiprogramming between 15 and
21. There is more simultaneity in the input stream than in the hardware.
Notice also that 97 to 98 percent of the interactions fell in the light load class of work. That
is 97 to 98 percent of the people time, the most expensive commodity. The remaining two to
three percent of the interactions accounted for 80 percent of the total demand on the CPU,
I/O, and main storage.
By selectively detecting and scheduling the few large tasks, we can assure the remaining 98
percent of the interactions will have excellent response.
This has been done in the Resource Management PRPQ in VM/CMS.
Figures 17 and 18 show the distribution of CPU time over the user population at Dartmouth
and in a pure batch environment at Yorktown. The horizontal axis shows the cumulative
percent of the people using each system. The vertical axis shows the cumulative percent of
CPU consumed by those users over a month. Notice that 95 percent of the users use a total of
25 percent of the resource. If we further examine the top 5 percent of the users we find that
5 percent of the programs and data they run account for 85 percent of their load. Again, a
very small percent of the interactions account for the overwhelming majority of the load.
By putting a small effort into tuning these few users' large programs, we find large service
gains can be achieved in short time periods. This seldom results in less overall load since you
are taking a machine limited activity and making it less so. Most people simply, run their
programs more then. However, their service is improved sharply, and the granularity of the
load they impose on others is smaller thus giving others better service as well.
It turns out that I/O has a similar distribution while main storage is slightly flatter.
Again, scheduling can be, and is effective in taking advantage of this load distribution.
I have seen many different systems in many different environments. All of them, except ACP,
have the same load distribution. This is a reliable and repeatable load curve distribution.
Figure 19 highlights the impact of scheduling on the man-machine interface. It shows the
impact of the Resource Management PRPQ for VM/370 in a benchmark environment as it
would apply to people. The columns titled WHEELER show the results of running with the
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
Page 6
PRPQ in an overloaded environment. The columns titled RELEASE 3 show that same
overloaded environment without the PRPQ. Notice that 50 percent more interactions can be
handled by the PRPQ in this overloaded environment. And the response time at the 90th
percentile is 20 times better. Notice also, that the base system handles more interactions with
60 people than with 80. By allowing those extra 20 users onto the system in that overloaded
environment we observe that we have effectively lost the 20 people and slowed the other 60
down.
Since people currently cost more than computers do, it seems that such measures should be
incorporated into evaluations of systems.
Figure 19a shows computer cost on the vertical axis and the service measure called the
expansion factor on the horizontal axis. If the machine is underused, then most of its cost may
be lost dollars. As the demand for computing services grow, the expansion factor also grows.
At some point, the system is being optimally used. As the load builds beyond that point, much
of the hardware usage goes into unproductive work.
The next foil shows the lost hardware dollars per hour due to underload and overload on one
of the VM/CMS systems at Yorktown. The V shaped curve is an envelope containing the
actual operating points. The expansion factor used here is an actual elapsed time expansion
factor. It is the actual time to do some typical unit of computer-limited work divided by the
minimum time to do that work in a stand-alone environment.
Foil 19c shows the cost of the lost end user time per hour on the vertical axis. The horizontal
axis shows the expansion factor. As the expansion factor grows the number of simultaneous
users is also growing, and the number of simultaneous people being simultaneously slowed
down by a slowly decreasing service to each also grows.
Figure 19d shows the combination of the hardware operating points and the end user costs as
a combined loss. Note that there is indeed an optimal operating point, having expansion factors
in the range of 4 to 6. The cost of overload is far greater than the cost of underload when the
end user's time is considered.
We use measures like these to monitor the daily use of our VM/CMS systems in Yorktown on
our two 1688. We find that we can normally maintain the operating point very close to an
expansion factor of 5 for most days.
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
Page 7
Figures 20 and 21 show the five layers of storage in today's systems and the problems
addressed by managing each boundary in a better way.
By observation we have learned that data lives usefully in the cache for about 10 ms., in main
storage for about 10 secs., on drum or fixed head file for about 10 minutes, on disk for about
10 days, and on archival storage, such as MSS, for many years.
The cache-main storage boundary has to do with CPU speed. Differences ranging from 2 tol
up to 5 to 1 are achieved by proper cache management. This is because the locality of
reference of most programs is such that a very small part of the storage receives 95 percent or
more of the action.
The main storage - drum or fixed head file boundary has to do mainly with system response
time. The issues here are the working set characteristics of programs to be run, the page
replacement algorithms of the operating system, the parachor characteristics of the combina-
tion of these two, and the management of the paging device channel program. Differences
greater than 10 to 1 in interactive response time can be achieved by proper combinations of
these four functions.
The drum - disk boundary has to do with the management of virtual storage. If there are 200
simultaneous users of an operating system, and each one has 2 megabytes of virtual storage,
that is 600 megabytes of virtual storage. The 2305-2 fixed head files (drums) hold 11 mega-
bytes of virtual storage on each drum. It clearly doesn't all fit on even 4 drums. However, the
same locality of reference properties apply here as in all other levels of this hierarchy, only the
time scale is on the order of ten minutes. Once again differences ranging from 2 to 1 up to 5
to 1 are achievable by a good drum-disk migration scheme.
Finally, the boundary between online disk storage and online archival storage is the key to user
growth and user productivity savings. Again the locality of reference properties apply and we
find that 95 percent of online disk storage is never referenced for periods of time greater than
a day. Furthermore, it has been found that users who fill their allocated online disk space
spend a fair amount of time being data space managers from then on. The gating factor is not
the amount of disk space they have. It is the users themselves. They inevitably free up just
enough disk space to do what they have to each day. This is counter productive on their part.
However, the most important problem of all is the issue of user growth. By automating the
disk - archive boundary an open ended secondary store effect is created. This opens up many
new kinds of applications and a whole new way of using computers. At Yorktown the
computing systems are treated as open ended extensions of peoples' memory and reasoning
power. This applies across the entire spectrum of tasks they do. It includes a wide spread of
people.
The previous Figures provide insight to the barriers and problems we have found and correct-
ed. Motivated by what we have learned in terms of placing people productivity ahead of
machine productivity we have realized that the need for asynchronous processing becomes very
much stronger. If people are to grow in the use of computers to extend their memory and
reasoning power, then obstacles in the path of that growth must be removed.
At Yorktown we use 5100s and the IBM 7406 for many Lab Automation problems. We built
TPVM to interface System 7s for bigger Lab Automation problems. Computer Networking is
used extensively to interconnect many different computing systems within IBM. In this way we
can sharply reduce most conversion problems since we can nearly always find a site to run old
programs. The message of COMMUNICATE DON'T CONVERT is powerful.
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
Page 8
We have developed many special function virtual machines to address those functions which
we found to be very common. In this way the special function evolves in an environment
which is uncontaminated by other non related function. This sharply reduces complexity while
simultaneously allowing the user to do other work at his or her terminal While the asynchro-
nous processing of the special function goes on.
Some examples of this special purpose function are listed on Figure 22.
The rate at which our users increase their ability to work in this environment is most impres-
sive. They first show improved quality of what they do. Second they show improved quantity.
In general the quantity shows an increase in the neighborhood of a factor of 10 in Lab
automation, document production, and program development.
Figures 23, 24, and 25 show a picture of the Research Device Coupler and several ways in
which it can be interfaced to the systems. The RDC. which acts with, or as, a terminal, brings
digital and analog input and output facilities to interactive terminal-oriented systems or to
stand-alone systems such as the IBM 5100 Portable Computer. It is designed to provide those
experiments with modest requirements access to high level language facilities, such as APL,
FORTRAN, BASIC, PL/1, etc., for direct experiment control. This accounts for 70 to 80
percent of lab experiments. It eliminates the complexities of learning other operating systems.
Figure 26 shows some of the distributed function which we have in place today.
The lab scientist, using Labs/7 (EDX) as the operating system in his System 7 can control
multiple experiments. When the need arises to use a more powerful computing system either
for processing or for accessing a data base, a simple mechanism called TPVM provides remote
intelligent sub-systems with the ability to access and utilize the power of the IBM VM/370
environment. The scientist can directly access and edit data in VM/CMS from his terminal on
the System 7. He can also initiate asynchronous processing of data either totally independent
from what he is doing, or in a way whereby he can monitor the progress of the number
crunching while doing other things. The speed of the interfaces are such that we have seen
response times ranging from 7 to 40 ms. for the user at the System 7 terminal.
As the need arose to interconnect the different computing systems with different operating
systems within IBM, the Subsystems Unified Network (SUN) was developed. This has
recently become a series of IBM products, known collectively as NJ! and NJE. We now have
the ability for users of more than 190 internal IBM systems to communicate data to each
other, run jobs remotely, and disseminate the results to as many diverse systems and terminals
as they need to.
By having locally attached display devices as terminals, our users can normally achieve
sub-second response time for almost all that they do. At the same time they can communicate
results, interchange technical documents, and freely communicate with colleagues all over the
world. This communication takes place asynchronously from whatever they are doing at the
terminal. A special function virtual machine manages the networking in our VM/CMS
systems. Currently, VM/CMS systems constitute just over 2/3 of the systems in our SUN
network. At Yorktown, about 500 people use the network to send and receive about 17,000
files amongst themselves and their colleagues on other systems each week.
The message COMMUNICATE DATA, NOT PEOPLE is powerful for interactive computing
where the sub-second response time requirements are so strong.
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
Page 9
Figures 27 and 28 show the early growth of the SUN network.
We have seen in earlier studies that editing accounts for 50 to 80 percent of what people do at
terminals. One of the uses for editing is text processing. A special purpose virtual machine
called TTF handles our text processing requirements at Yorktown. It provides the user with
many useful facilities for composing documents. SCRIPT/370 is the primary formatting
language used by our community. We have a compiler called STC (SCRIPT to TERMTEXT
converter) which gives the user access to the more powerful TERMTEXT facilities, including
over 400 separate fonts at the Yorktown computing center. Figures 29 and 30 show the sharp
growth of the TTF service machine at Yorktown over the past year.
In summary, we have realized that the user at the terminal needs distributed function. In some
cases this means special purpose virtual machines. In others it means interconnected computing
systems. While in still others it means independently operating computing systems which may
only be interconnected occasionally.
We have learned that users steadily and rapidly increase the number of diverse functions they
perform and that the requirements for better asynchronous processing grow proportionally to
the interactive growth.
For the past ten years we have studied the way our people used interactive systems and
observed the barriers they encountered. They have been described in this presentation. The
resulting working environment we have created for our Yorktown Research laboratory is highly
productive and exciting.
Computer availability must be targeted for 24 hours per day and 7 days per week. If informa-
tion is not available from a computing system when it is needed, then people do not keep all
their data online. This implies that maintenance will have to be done time-shared.
Computer response time must usually be sub-second. This directly impacts people perform-
ance. Response time delays can lead to people delays, irritation, and more mistakes by people.
Sub-second computer response time also opens up broad new areas of applications.
Data space management barriers, especially across main-drum, drum-disk, and disk-archive
must be minimized. They are major barriers to economies of service and limits to user growth.
Complexity is the major problem facing the user today. By the use of EXECs we can sharply
reduce complexity while increasing function. Also, by using virtual machines for special
purpose function, we minimize contamination of the function, thus reducing complexity.
The VM/CMS systems at Yorktown are the major interactive interface to all forms of
computing for the Yorktown user community.
Today, over 25 percent of our laboratory is working interactively at any time throughout the
day. Over 75 percent of our laboratory works interactively each week. There has been, aL
factor of 8 growth in the time our people spend working interactively in the past four years
alone. Over 98,000 hours of 'connect' time is measured in a month today. The users are
present, and subject to system influence at least 38,000 hours per month, and at most 64,000
hours per month. Most of the remainder is for the special function virtual machines.
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
MISS 1 oN
PROVIDE ADVANCED COMPUTER SERVICES
ENCOURAGE INNOVATIVE USES OF COMPUTERS
MA.KE COMPUTERS EASY TO USE
EDUCATE USERS:
WHAT IS AVAILABLE
HOW IT CAN BE USED
UNDERTAZZ SELECTED DEVELOPMENT PROJECTS
MEASURE, ANALYZE, AND EVALUATE COMPUTER SERVICES
IN ORDER TO IMPROVE AND EXTEND THEM
FIGURE 1
ENVIRONMENT
INTERACTIVE USAGE GROWS RAPIDLY
MULTIPLE SERVICES AND LANGUAGES ARE ROUTINELY- REQUIRED
REQUIRED FUNCTIONS MAY RESIDE IN DIFFERENT. SYSTEMS_
.FIGURE 2
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
THOMAS J. WATSON RESEARCH CENTER
COMPUTING SYSTEMS DEPAP:INEr
TERMNALS ?
3277s,2741' s
OTHER TERMINAL TYPES
,--,--,--------
I 370/!68-3 , I 370466
'
OS/\'S2 ?1:2=d VM/370
ISO MS
JES 3 VS APL
1:11
)!
RUE (FRMOTE)
4 1 ; 3'0 c.
?
EXPgiiMENTALi
PRINTER
ti?S?C:4020
3 A68
VW371)
OtS
- VS AR.
RUE (LOCAL)
2922
2/1977
MSS
3850 82
DEVICE
COUPLERS
TIME /NDEPENCENT
EXPERINENTS,..
TIME DEPENDENT
EXPERIMENTS
FIGURE 3
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
a 11.
14-
" / 1
i
, . .
. ? : : : ? ::
. . .
..... '..
? : : t : : ?
,
?
.
,,,,
. : I '
..)/00:.Qpr,.IS'Eplp4?e
. ?
: . . 1 ::: :: ?
? . ? ... .
gP01.1.
...
7/12:
.C.I.AFkDF'84,2pig34Rooso01:AogiZ
?
. ,
, . .
?
:
.
:::-
.
-?,.;.:.:
to
......
7.1::::-,::

.......
, .
... ; .
. ::::::
..... . .. !::::
::::::::: ::::;!:::
I ? ..
.:;:i1:::"
:I" ::
_Li ...1,,...: :
:' ::I .
i .::; : : :
. .....
.; : . ?
....; ...
' ; ..1...
.. : .:
' .....
: . :;::: .
-?.'7 ''::?
'I,
' ?
:: .
'7?
..... ::::
::::! .
? i ...
::: . ::::
? ....
? ? ? 1" ? ?
:::t:::;
???1??"..
.? .? .? .? ., ...? ?.
..-...... _
... ?
....I ...
? ....
:::: . ::::
....
::::
.........
......:-? 7:7'77

7.7.7.
:?: 7:.1:." :?:
": ?:.
1?
....:,.:.{....;:
-T?

Il
i:.-
:?"

.:: . :::
.....
":. .
:....
...
::::;::::
.........
:::::::::
;::::: . : .
:::!::::
???1:??:
:::::-::::
:.:?;?:??
...... .. !"??
... : . : . i?
....
90,000;:::
,.....
,. ? 1 -
-?-?
?:::t:????
?-: -
?:...:
. :: ;:??
.
:.: .-:. :_.
:.7:..' ,::.
.:.?.. :." :..:.- I:.. :.-:. :. :.

: .: :- : :. .:.: :.
2. . :? 4..;...:._:...:..
..:.:.. .. f,
.? ... . I

:?:?:::.: : ........I
:.:.:.:.::.7:..
.....
.....
...... .?
::..
..:...::..
7
-:
MONTHLY
-2.::.
....

.2...

,...4.....
.. ...
.. .
:.......
. .
..
.
.. ..
' CONNECT
.":---
? - - ?
...
. ? . ' . ; : .:
; 1 !". ?:: ..........
:1 ,
....
? ? ...:::::.:1
? ....
.: .. ..?
.. .,.
?????,??????4
? .. ? ? ? ......? ?
? ..1.
I : .
?:1:1.:::. ..;
...
........:2:. *
.
4.1.1.:???:.?
. : ' ...
1.0:??:??????
30,000 ... .... :::.:1 ,
? . . : . :::
:
:7. :.: :7....:
? 1. . ?
.." "-"'"'...".
....
???? ...????? o.p...
? ..? ? ? . ?
-
. .
. . ? "
tiOURS-
.,
.... .
. .....
-
... . . .. .
. ..... .
:::::
.. . ..
:::-.1:::: . : . ?????4
..... ? ..
".......?417...7. ?
..4.?????.
.
:
... :. :.
... ......
..... ....... .
.... ...?
..............
..,-.........., ......
:: .1.7????.**.
..?
i'... I...:...-I...
i ????:. .
. .....-:. ror
. ????????????*
..... .................
.::Z7..=:.

.:
.... ...
-
.. ..
::.... ::: :.::::f:
..... . ..m.7.L.?
:::::.:-1.77.--.::i.--...........-F"'"
...2:E..._.....
...../.........4...........r.
.. .1 . ? ... .
L: i. ...- a: : :. - i Z.. :. . i . 7. ? .?? . i . ..?. ?.. ? . ? i :
,...-......
/
...
..
.
.-.
. i
--... r
-'
. .
:- . " ..
*70r.o.
. . - -
-.
.
. .. .
' -:
. --'?-?
-: :-::::-.:7-. -
_ ....
.1
..:7 7. "7. :..
0 a T'r li 7="
:
7. ; i ' : - :-
77' A . - . ?
1 " .
1 ?"?-? ..r.=.
.
.. ..
' '-''
........
----.-...?_...._.:.......I.T.r...74:.:::
...,...-
?-??+ i
a
,....
1-.
I
4-..........
4
.1
I
.
'
\ -
.4
...-7-..- . - t ..:::: ::-
- ,77--'
- " .=.....
r
?
--.......... ...
7-. !.
?????-.-
-.a- I
? r
..1........-........... .L.
1
.-.7---
,-_-60.000-_-_,
n .
-,...=?}
-,-.--
,
,.
I
.................
''.._ ..,.........
..-?
.......,.????????
____-_-1---
.? ' Hourts-----

_. . .__.
,___...
1t
?
1.?
.__. PER
r
, .
-
PER-,.-
L*
?-?
- ? -..rz.,_______
7----7--....---:

YEAR.
- -
-r----
- -- -T.
0; ooa MONTH
:
..,--
=--....-
........-,---.
...

-.
.

.....-,
t I'
-
-....
'..."'It
4.........
-,
...................
,...,..,
--,.............
7'
?
?
-,
-----
HOU
40, GOO
.,--...._ ,
-..--,...
.....
4
.0=????4.
?...1.do?
??????+.1.1....
?
I...".
.......?."?
....,=?...:::
\.
........??.????
????.????1??1??
own?????
?
....mi...
1=:iii
.1
.............t....
50., i
....
...
.. rr. t. i? , A
.00
m?.._..
11MIENEMS.
=Er....
i
......
1..........
..."...........
...E...
y
izT
"...0......
..
I
::=.........
L.
...... NIIIIIMMIIIIIIIIIIMOMMINIII
....
-77'.....7.-.1...-....-.....
?
=
lemoloom?
4...
. ?
27,0001
11111111--111511'
....-..
..-_..........._.............._..
::
--.........
'
TO TA fO
P
...--,-.....
- ? ,o,..........?_,
----
?????,...?????
_......
..........,....,..
..: Li-
-----.
...
. ? .1?
. ....
-.74'..... .7....... .... .
............. . . ... -
- "
'' I.
.
...........*.
..... Amin
....
................. ? v.
........? ? we... ...me ? ..... ....................-..
...........,.....

? .:17...17...:::?;w
P
HARDWARE
r
.,.."."'t41:....-`4.
no 001
...- ,..........-=
. ..... - .....
...... I ,.. ?...
1.?
moms. ? ? ??? ?
...1.11..-..
. .....
1....1 :Z. ?
. .
..
. ....
......
...,...,........_?_?...
. ........
?=7.........;:
...
...?=...:.
...........
.......: .......
.......
.????????????......?
-,
...............-
.........???...
?????.:
.:. =o: .?
.....:==:::.?
? :
:.... 4.::.:
......... ..
*
7.,....-........L=i,--,..--
- --. ..--
,--
&
,--
. ...L.::: . : .. .
. ,.......-
-r= :L........
?.........7..1..=
.:7.-..........
---.
--:::t
zzz.
:::::::
=.:::::.
,..... _

.
: .......... ::::.' :I= : : : : : :.
--"':': :::-: I 7: -:: . " '
. : 7.7.:::?:= ..:......:::::=
... ! r ' ? ' t ' ' ? ' ...
=:..........
""*"''''''''''""
???........
"' - ' ' ? -
/
''. I
=4....
' '
7::i ::: :
...I
::-:.r.
........;- ..
=:-.7......
...........-.......
1400.o?
...
19 3 :.: 197k --.'74:?
' ? ? 19
.
: -... t:: ? 19
.... . ..... ?

..--F-1::-.77.??=1.:*
r.-- - riz. i :-----.-. !:- - -
:-----1-:::- '-'-
:
*---"?*^-17::::"::;
--*-.:?:-1::::::::
:2.3 -.:::::2-.?-_,-:
-?
?.? .....
--2".:::: . 2 :::1--:?:44-.7.:.:
..- ? - - "----::
:-.-.:::::...2:-: _=?-.4Z-2--?
t 2 ** ' :2 --- -
" ? .... ? . ? ? ? -? - "-? '
- IBM? : ? T .'? J.
... ? ? ? ? ? '
WATSON. COMPbT I N
..-.
CENTER ; :::
. ......
;!: :I ;-:
..,.
:F:1. F. :7?17-:_117:
ME_
..... ..
.:.....
----

....
..... 4
. , . ...... .. .......
...'". ?
. '`" ? ' ? ? ? . : . ' !" ' ? ? ::::: :
"
....
...... .. . . .
C0t4PU1ti
NC COS
-TS '4,
.. VSER-O0S-T
....? mil
Effiglos
? ..
rr,::::
, --
---:1:--- - -
---.L.-
. :'.-"1::.?:::
:::11::::
.:.::
.
.:-_-:--.J.:::?.1:::::::.::::
::::n:::
i -
L::::::::+::?!:;;:
..1
?
::. .. .... :.FF,...:.:?.::.::::
F.:.:..i::, ? -. .
, ?
: ri'-'.. i : : ?.
' ' .. .. : .. ?
;:.:: . .
.
.
.
; ; : ; :
..? :,:...
; ? : : t : :1: ?
.
.. ' .. ':'1 ... :1:.
1 ?
i?i: 1 .:. ''" ? : ; . ' . .
: : : I: !:
?
? .....
I
, .... ..
.., .... .-.4.
...4.1221; ::4.2.
.... , ? : . , .....
-.
42.2;
???
: .: I : .
...... "?2.?
1
fFP_uved.
.. ...
."".1".
"
'Y i - R
... ..
.6ase,
,.. ,
200
? , ?
- le - ? '
? .1
_.......
..... ..
b. boa
............. o.v.
....I.."
...ht.". .1"
..,..
I
till
?
flit
WI
-
IIII
11
:if
!lit
'1
;.11
fr.
1. '
:::
,.
-.
'......:
'.
,
; .'
.
'
H
.
, .
i
.1' .
.
:
i -
"i?
: ..:
. ...,,
?I'
. .
.--
.,
Ii.
fli,
" r,
1
COMPUTER'r.
5C1 Et
I

C ?
.
11 .
,
!. '
Ifr.
1:-
!II'.
-
.
' !I
': . ".
. f!
PilYSIOA
-.H..,"
St 1 E
.
,, .
":;'
?---;
CE
:,
1
.:
.
1.1,
'''
.i '
im?
GENE
.-.1.-.
SC
-lilts
. .
AL
:?
ENCE
'?
;L:
Is
_
-
.
?
..-.11-::
,
H!
...,
!
1; ,:?,
:.1::::
.....
..
::1
. ...
7
.. ..
-'

MATH.
SC1
E
--:
? :
CF '
fPA:LIED
JIESEARCli
..
:':
i
COMP
:.:.1:
SYST
TING.
MS!.?
.
.
.
- :::::-
.
? .
::.
...
...
...
..; ,;--
.. ....
:-.? ,- ..
.... . ...
",. .
:.? --.?

... t .... ,.. -
........ .
........ .
'' ,
4,,.
;;.::
.i..
I: .
:iti
,,,,
:!1!
:::
I.,.
:::
..;
..-.
?,. :
i,,,

"
,.
? 1
111
::
r.
::
.,
tit
!! 1
Li
i ''
;
1.: !
?
i 1.
':
'
:
.
?
1
,,
t
:
;
. I-1
!
:
?
-
i
:::.
P.::
`,-*-1.
.1, ;
., ;
I '
..i
iiii
' '
;
.
...
,.
1
:11
I,:
'';
All
::'-'
-
1,
. :
,, :
:
*
:4;
I:''
:i; .'
.-::-.
.
'
....
1,,.
.
,
'.
. 1::
..
..
I
. ..
? ?
. .
.
.
1
.
...
.
....,.
--""
-:
......1::..
.:..,....
... ; ... : ..
........ -
. ....... I:.
:::::::::::-
? a.
1 :
:I!,
PI
1 ,
i
I
.
1
11
I,
11,,
1 1.
i 1.1
.
:
: t
:
!
.t! t
1
i
III!
4
--i-
:I
li
I
1 III!
! 111!
. ,
'I
it
if
1 fr; !Iv
I IIII
: ? t'
1.i:

.... ....
I::::
..,,...r.-7.77
,! T
...,
.
. .
.. 1
? Ai
iT
t iti; ,..
W.
P.':
....
........
....
::::I::::
.................
.. .......
..
.. ....... ...
'
;
'...t..
.1:1:
.0';
.1:1:.
il
.::.
A
ili
. 1
II
-.-ii
1M
' 1
.
:tl
1 ji?
1.11
i li '111
i t
'
1 ,
, i
4 :
:
i
!
;
1
i
j
Ili:
t'
tft;
..1
:I
!
..
.1
.1
I ill,
' Il.:
f
?
?
,.
'..
:. 1 1
.
1: !
71 -:-
If :
.i ,
t!li 1
I ,
'Of i -
1
'
..:.
!t;:
::::
?
1:-
?
?-?
...
.
.
-
???
...
....,
-
....
-.
...L.,-
....................
.........
..... -
........ ?
- ---- --- - --
... ; .... I ..
...,.... ...
t."
......?_t.i
r-,-
ir
'
r
1 I ..
I.
,
.
.
if
I. .
'III
:
1 i
1
.
lir
.
I:.
:
I
_
1 1
/
i.
.
li ;
.. 1
ti
II f
.
-41-
f 1 :i.
1 fif
..
1 :

li
: ii
"
,??
tt,
I.::
17?.
fft'
'7'....7:

.
?
.
.
..
,.
.. . ..
?
. .,
4_4dt.;*
1111
:III
::::
i_ti_i
III,
...
LI'.
-.11!
. I
if
1, I
'
ft 1 .
,
+
ri.i;o
I/
!,
_.
/
-I
.
a '
.
.
?
,
.
:;:,....
: .?
,
,.
i .
1
,
.
:
1
i
1
.
, .
.:
1 0
.. ''
11
'
if,..,
t
' I
'I,. :::
.
::::
1.',::
-:-1'....:.:7
::
....
,
' .
. .
. ..
. -?
: ... ; ... .
...::::.:
....t ..-
: . ::
,1...._,...
? _ __.
........
. ,...
.. .
?
.
i *
??
III
I
.!
I
I ,
?
1
,.
.
.
:
:
1
,
:It
I,
'T.
,
.
;
' -.
:
,
1-ff,
,I :
.,
_....,
.":
.?, ..
:!!! :
.
,
?
. ?
. .
...... .
.".""

NUMB
?
OF
::.
-
:
_ 1
::!
11.
-
;
I
4 li
; . III
0
..
1
.
'
I
1 II':
1.
:
1.
,
.
, 1
1
I
.
i
i
)
,
0,
;
1.4:
'
'
? '
I?
,.
17
.1
.1
:,
?
I
1
,
.
il ..i
..I1
.,,
II!
. 11
11
.
11
411
i i..
. 1,
.,
''''t,
:;. I
,
'.1:1-i
.1
. 1
.
il'i
li .1
it.:
.-.I
lip
1..41
''',
;:!:
7 7
li '
i47 '
;..1?
1,1
'i,,
7??
:::
...L>T20
...
,...7
:
_____......
7"
:;',..;??-...14-0
;NON
TY:
777
ZERO
:
-HOURS
tIME
-
HOURS
...
- -
PER
. . ...
PER
.
....... .
. . ... ....
MONTH 1.::.;
.. _
.1.. . .
MONTH -:?----
.....
..
.. ......
........-
7.:.::::LH
. ......
- ----.--
7 -
..
--
pcort.
-.
77
-...
?
'
I?
tii
:tt
.
t
' it ii.
i, 77i
1?,-...
'rU "'
::.. ::-
.- .
'
II
ir
"
1!
,.
,.
[
,
it
''
7
.1 r,
; 11:1
''.T-
: .
.
1
i.:t::::.;
till
:.I.:
. '
?
....;
it
Li
,,? ?
?
iil
1
.
.
1
I'
_11-1-1.
i,
i.
?
,
r
1
f
;t.
,
I
1;
,-
?
;
,
:
1'1-
, ,
?,?
,
I
i
,.
, ?
?
.
;
.
:I.;
..t;
::
.
..
;:-..
:.
...;.-..-1
.
'
. . , ...
.,.
--
..; . . .
'Li
iili
:it
:'
.
?
?
-.
:
::
.
.::
..
..
.
.
?
...., i

11:1
tr;
1:!:
::;!
i::
.:1 .
?
? .
r
.
1
,ii
1,'
? -
i '
"0
Il.
'
1
..
?
1
.
,.
,'I
11
,,
, I
?
? '
,
,'
..
.
' i'
..1 -
1
;
:11.4111111-I:.
.
,. ,,
1
fil:'
?
'':
ifff
:::
::I:
:::: ,!:
.i::
;
.......
--?!.. ::'':
, _
:
1#..
..
.: ?
?
1
' 1:
?
.... ,i.
Ifi 1,',
-
?
1 I iii
; ? ,ri
. I.:
06
.,..
WI
iiii
..,.
I,-
.".
ii
.
.
--
?
. . .
....?...
........ ,
...
.......
.. -
-_,
7-?
: ?
7
'
7.
-. .
;7:
..
7
3
.
7 :7 7:
1 3....
'4
7 t--:
RACT
:7.
6
ti
:1
7
-7
T:
yE'-cOMPU
.
1.
11
71
3.11
-1'
?
.
i7
..
.
7.
. ? 1
NO
7
1 .
1
?
,
T ip
P Ft:
,
;
1
'..7. r
,
?EPA
i
7.
'
7
3
r
-MEN
i
'
!
1 1
T-1.
t
.
I
ll
rr
i,
1
,
1
7,
'
7 ?
:
,
.
11
1. 7 T.,
,1 . ?
:
?,4-..?",-,41
. ? -...,:
.1 ; : '
:::.? ?I
.,,, .-I.
hi: .4.1
.
r:7
....
--..--1 -
.... ..
.

.
,,,
i:1:
" 7
: i:
,
it
:7!
..:
kil'f.':
.
...-
ROWTH
..
.:
OF
. :
-.r. T
? ?
1 i :T..
r ? 1 1 ii
4r,
' ? '. .11
.1 . , :'
:t : :IT!
. 't-i-t-tril.
: 1IN
: :: ....
:11: .:::: ::::
!t:-:. F.::
Ott : ?
..:
::.
-
-
- '
- .
....
..
11:1
..,
.
....
...1-.
1:-::-:.

I.:
1-:-..
,
! .i
I....
11!
MON
".'t
Ti
1
'"
r
I 1
?I
tr,4
ilt1
1.tr
II!!
cil:
itil
"II
liti
11,;
'J4
III!
IT
1.! 1
!1i,
.-;7"
.t:
14.17
;1.97
77"..i.i
i II It
-
197
II

--
D.'
fc
-
I
11.
LL
1
;
'
,_ ,_
I
1
III,
ii
i
t.:,...
I
,
1
f
il : :: ' '
. ,.
1.1 i 14_, -
i
:I ! :tit
!, . 7,1,
hi-,.. ii2_:_,31.:?.....,_-
ii : Hi!,
:I- "- ' ?
,t,, 7.77 ?
',1:i '1i: ;:::
-
--?!----
.. .... .
-.

--,-
-'-'
.
?
?-?!? ??1--7777-
.. ',..
,,..
......
:;.:i
"-'3
:!li
'''.
.,
? ?
ili_.4,_::
,-,
?-:',
t:2,.:'
_t_:_?,'
IP'
..!!
i
11
1 i
.1:
' 1.1
1
t..t?-i
--i"
11
li
.
i 1iii
1111
1 1T1!
"
-'-!-1------1-?---'4.-=--'
1:ir
.
: ,
1
t
1 i
! I
h ,
4'
"( i
1
91--, i
i_
, ! TT1:
:0, 1 . I 1
'
11'
hrt
ii 1 -711--i,-io
ii 1 If
' ' .::
II1rott
.,..1
:..-
00 1111 ,::
r-4 i-: -
ititiT. ::
0;1 : :: ..
,
:
.. .
-.
-
?
-
::
--'
::I.t....
::::.
: . ..... -
-.7-i-..--:-..
.....
III
n
I.,.
0 I
101
II 1 II
1
I
I
I
1 :t
t
I I t ..
t. . ttsr
t.t. ;
. .
???
? rO4VPd
4017 ? !O.!! ? ???1?;; ?
? ; ? ???::
? .42,
? ? i. ?
?
? ? ?????????
? ?
??? ? . ??
?
? / JO SO.
1 0:
".,
? I ?
? . ?
+am ??????????..I.... ? .
.. ..:*.t. ',*t?*.
....I::: ::
i?!
?-? "-:. epiimi
. (,(ti..
? ;
Fort Rel
?-? ;??!.?
? .1, ? : ? :
;
84-0093iR000500120001-5
,
. .
I
/4/eftori v.,
1 .
? 0
. 1.7:1 ? le:.041 .. I a
/ ;
? I. ? ..? / ? . '.;' i
???????...14... ...... IL:1 .'...1: . .
. .
. . . :
.. ' I ?
1
.. ?1 ? ? . ' .;''
I ' a
.
t
? ?
a. !...11.++,1, ? ? ? :JI... +of . ? ?... r.?
I ? i I
? ?
?
? ?????????.11.
:::.11:
? I e
m ? ?
????????r? 4:?40 ????
? ? ? ? ? ;
? I. I ?
4?Mi.
bib
tvi? 50 '77/1/i/-
?
? 1:?''I.
?
: '? .1 1
!... ::: '
'. .1.41 ::.. .
. ? ?
I
:* r:.tt
? ? ? ? ?
+ ? +ma
? ? ?
:i;;
4.44
.??. ':'!
1...
I.rt,;
Ti^ 1"i
:Ilt':
. ...
I 4 ?
!!!.
41:11114
0 4 0
004,0 is.
?0 480.
?*.?
t...
:t.
: .
I . ? 1 ? ?
? ?????? ? ????
60 hfeCileS
; ? . ? . ? ? i
? I
ee
HOMO/
o:':1
.... ? Ovs .; v. (762A/
0 ? . ; ?
? 0 ? ??:4 ? 4
ihiiii "C.1...: t.7.t..1;10?r. .0::..?.*:
;.itc:: . ..!:' 4..,: .i?. ;,:: LI?. '-- t*:. * ...1'... '
....
;
i
'
?
I
? ? ow+ sr +door+ +a .6.00+1++4, oleo ++.0ormr
I ..e
LWO 0 0 -4.- ?
!.
? ;1.0
m
ApproviIR Releasa
e 1/07/12 : A- DP84-0041 0050012 75 t
? FIGURE-- 5-
04. ?????. ow
: ? I: ? ;
? ?
? ? ?
.?
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
INTERACTIVE BATCH
CONTINUOUS
5 HOURS / DAY
EVOLUTIONARY
200 COMMANDS / DAY
LOAD RANGE 100,000 : 1
FIGURE 6
DISCRETE
1/4 HOUR / DAY
STATIC
9 STEPS / DAY
LOAD RANGE 100,000 : 1
THE NUMBER OF COMMANDS
DOUBLED IN FIVE MONTHS
THE NUMBER OF COMMANDS
TERMINAL IS 25 : 1 IN 1
USER GROWTH
EXECUTED PER PERSON
IN 1971 AT YORKTOWN
EXECUTED PER COMMAND TYPED AT A
977 AT STANFORD
COMMAND PROCEDURES (30,000) OUTNUMBER PROGRAMS IN
CONVENTIONAL LANGUAGES (15,000) AT YORKTOWN IN -1-77-7
FIGURE 7
?
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
e
Appattl For Release994/07/1ii Rilfirqr84-oo933R00059o122oo4A-713049
7
SCRIPT
2397
9.322132
1 1s
72.56097
TEXT
1758
6.037008
1 14
69.51219
10
NODULE
728
2.831252
99
60.36584
ASSEMBLE
656
2.551238
73
44.51219-
20
MAIL
261
1.01505
68
41.46341
45
PLI
1496
5.825v47
64
39.02433
16
UEFAUL1S
72
0.280014
? ?
58
35.361,84
127
FORTRAN
723
2.811607
56
34.14033
19
$016ECT,
68
0.2644576
49
'.:9.37604
52
MACLIE
0.320663
44
26.02925
22
VMAPLwS
350
1.361178
4.?
25.60974
95
8E110
198
0.7700385
41
25.
?
15
ARCHIVES
5b
0.2177880
40
24.39024
21
SYNONYM
140
0.1555633
40
24.39024
14
CONSOLE
52
U.202232J
36
23.17073
104
SYSIN
1.73453
36
21.95122
?
193
TERMTEXT
99
0.3650192
30
16.29268
53
TXTLIB
48
0.1e6676
30
18.29268
17
VSAPLwS
14.7
0.5716952
30
1R.29268
148
JCL
253
0.96.1930
e8
17.07317
277
PLIOPT
14u
0.5076061
17.(.7317
23
OLDMAIL
27
0.1050052
25
15.2439
? ?
84
PLS2
272
1.05783
20
12.19512
22o
DIAG
24
0.9333795E-01
16
9.756097
Se?
96
MACRO
177
0.6883677
12
7.317073
25
LISP
104
0.037E:096
9
5.487804
451
PIC
4U
0.1555633
9
5.487804
480
MYS
ou
0.2644576
4.678046
112
OS
35
0.1301179
B
4.8.7d048
,
144
424
PICTURE
GRIN
35
56
0.1361179.
0.2177886
(3
4.878048-
3.658.536
450
LABEL
39
0.1516742
-b
3.618536
18
LISP370
70
0.2955703
b
3.658538
719
KGRAPN
85
G.J305721
4
2.4390F44
UM ? ?
345
KSCRIPT
27
0.1050052 ?
4
2.439024
499
INCLUDE
28
0.1068943
3
1.629260
?? ?????
110
SHS
305
1.18617
3
1.029268
94
BCPL
63
0.2450122
2
1.219512
????? ?
723
CLIST
115
0.4472445
2
1.219512'
743
FTO5F001
94
0.3055730
1.219512
????
24
LISPLIE
31
0.1205615
2
1.2191112
1668
BoaoFF
36
0.140007
1.'2191,12
210
FIsLIB
51
0.1983432
1
0.8097561
1557
Fun
34
0.132228d
1
0.6097561
285
'7
0.272235dE-01
0.
223
.SMONUANS
15
0.5833624E-01
0.
? A
82
0.3189048
O.
123
ACCESS
7
0.2722358E-01
0.
119
AEL
33
0.1283397.
0.
781
.APAR
9
0.3500174E-01
(1.
1669
AREAUESC
52
02-022323
U..
1670
AREADUMP
52
0.2022323
O.
565
.ASM
43
0.1672305
0.
782
ASSEMBSV
25
0.9722704E-01
0.
? IP
.04111.10 ? 1???
?AIA. ?
FIGURE 8
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
EDITORS
75 PERCENT OF COMMANDS TYPED ARE FOR EDITING
MAJOR EDITING PROGRAMS HAVE HISTORICALLY BEEN
NEGLECTED BY SYSTEM DEVELOPMENT GROUPS
MANY EDITORS DEVELOPED AND EVOLVED AT YORKTOWN
FIGURE 9
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
alitIDLE
NO. OF SIMULTANEOUS USERS ON TSS/360
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
13? 0A25111 I 1
- 843262123 21111 1
-136535331 4 I I
...?.- 3536432 2 II 1112 I 2 I
I( - 1272523 2 121111 II I
- 125 1211 2 21121 2 1 I
- $3125 41132 I 2 1 2$ II
- 143121 32 I I
- 1213 21 431 1221 311
- 241 311321 I 2 II 2 1
-I 131326231 12 22 33 I I
-11111331 111314331 2 1
- 11-332333411123241 2 3 II I
' 1223 136231 1112313 I I I
12542 3 III 211 I I I
- 1421114 1541213 6121 2 2 1
442213323132323132121 23 1 I 1
_ 1224 23442134213 41531 1 I I
_ 171334184222455546116 1 - II
11322 22414324572332 51 II I II
- 1331 341223323335313346111 211
- 111122 343 5731363425 1121 I 1
- 22 2133 I 2423C68142221331 11 I
- 212243132224541923 212 I 2 2 I
11112 1112463624A69A25234 I 31 I 122
_ 225411 22316383754382 22423 2111
- 2 223322377655453 2113131121 1
- 332132377324985342241 12 I II I I I
I- 2 113 221433647331352 12 12 I I III
42? I I 1 212458534 335241 14 12 221 11 1 1
I 1 52 441273333752415114111 2 1111 4 I 2
I 2213812544244311231413 21 2 1
- 3 3212213412 112113 1221 23 21' I
1 111 4112 22 31122221111211 1 1 I
- 1 II 11 I. II 2 2 1 1
I I ; .1 II
,... ; I ; Is ; 2 I 1
1 i
a-...
-.I I 21 I I I 1 I II
22
27
32
37
2 I
1 II I
1 1
12
1 1
I II
I I I I
47-
52
0
Ii
1
1
2 3 4 5 6
,RESPONSE TIME DISTRIBUTION (SECONDS)
Approved For Rele. e 2001/07/12 : Cl'A-RDP84-00933R00050010001-5
8 9
Approved For Release'2001/07/12 : CIA-RDP84-00933R000500120001-5
THE NUMBER OF SIMULTANEOUS SYSTEM
USERS IS A MISLEADING MEASURE,
FIGURE 1 1
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
?????
USER RESPONSE TPE IS DIRECTLY INFLUENCED
BY SYSTEll RESPONSE TIME (SRT)
L-3
30
25
cr- 20
15
10
5
41=11?.
10 15 SECONDS
SRI
FIGURE 12
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
FREQUENCY
FREQUENCY OF USER ;it:SPONSE TIME
FIGURE 13
TIME (SEC..)
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
70
? AVERAGE CMS WORKING SET SIZES
CP/67
LCC(5)
(200)
FORTRAN
EDIT LOCATE
FILE
LIST(2)
0
0 10 20
K c?
30 40
ASSEMBLE (600)
50 60 70
CPU TIME (SECS) ?
FIGURE 14
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
I I
90 100
? 4* ??? ? ???? .P0
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
AVERAGE LISP WORKING SET SIZES
CP/67
LARGE ALGEBRAIC SYSTEM (CC..t1PILE CI1RYBD8)
TYPICAL
COMPILATION
LOAD SMALL
LISP
TYPICAL
COMPILATION
GARBAGE LOAD
COLLECTION ALGEBRAICLISP
L.!
ozt
CC
L.L.1
10 20 30 40 50 60 70 80 90 100
CPU TIME (SECS)
FIGURE 15
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
O 0: 04 -. 04 ml tm eg C2 WI OW WI 01 0 0 C, cm cm um u) h. cm um h. h. I.. c) u)
-- 0 04 OA U2 ml M... 01 0) V) 43 42 0) M. CD C) CD -- 0 04 -- -" 04 C) 0) 0) 0 r?
126 ??? U2 WI WI CM 44 er CD C) f4 U2 er 03 CD -- cl 03 rm u, 04 04 (43 pi
??? 04 04 e4 C4 MI er a, r. Cm -- NI 01 el. rm rft ler 0) "
O 0 C) mm 04 1..? 101 -? -? C) C4 /1 M. 04 ?? 0 U) ?? 01 a3 om mr1 cu m1 $1
M. 111 WI 04 -- C2 CY rm er CD 4 CD 04 Cm 0 ?0 o3
ca. C4 ml M1 MI ml ml C4 C11
CD
40
VM/370 (158)
R -RESPONSE TIME (0-2SECS)
C- EXPANSION FACTORS (1-40)
z.)
0
cs
CD ac
cc c
CC
?
C3
0
O LC
U.
(.3. =
0 CC
42 = =i1lltitC111111,111tiltitt
W?CC-CC '
5 10 15 20 25
MULTIPROGRAMMING BANDWI MIL
FIGURE 16
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
30>--
I00
90
80
00.- 70
60
0
". 50
-1 40
.72
8 30
20
10
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
DARTMOUTH
(BASIC)
????????
mommisol
ROMEO
100
90
80
70
60
50
40
30
20
?10
1 1 1 1 1 1 1 I
0 10 20 30 40 50 60 70 80 90 100
CUMULATIVE % - USERS (4000)
FIGURE 17
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
Approved For Release 2001/07/12 : CIA-RDP84:00933R000500120001-5
100
99
80
tolo
YORKTOWN BATCH
(0S/91)
0.110,
INMONS
Minna,
??=1/11,
OW
20
10
100
90
80
60
50
40
30
20
10
o 10 20 30 40 50 60 70 80 SO (00
CUMULATIVE % - USERS (300)
FIGURE 18
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
Approved For Release 2001/07/12: CIA-RDP84-00933R000500120001-5
Am.
?
FEEDBACK TO USERS ENABLES INDIVIDUAL
IMPROVEMENTS BY FACTORS OF 2 TO 20
IN SYSTEM USAGE,
FIGURE 18a
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
- 158 CAMBRIDGE BENCHMARKS (40 MIN,)
WHEELER . RELEASE 3
?
MEMORY 2'11EG 1 MEG 2 MEG 1 MEG
RESPONSE 0,146 0.624 2,2 0.537 1.6 0.60:
TRANSACTIONS 12,374 11,149 7,921 9,176 7,319 8,06e,
? PROBLEM CPU 60.4% 50.3%. 33.6% 37,1% 23,3% 43.5
TOTAL CPU 95% 88.8% 66.6% 97.4% 51. 2x.
DRUM I/0 95% 95% 65%
DISK I/O 57. 5,70 35%
USERS 80 80 60 80 60
FIGURE. 1 9
? ?
?
.1
Approved For Release 2001/07/12 : CIA-RDP84-00933R0005001200U1-5
. . r.
eirp
em,
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
x72,89,usi o Aj
oc
Loot)
FIQUE. 192
Approved For Release 2001/07/12 : 1_
u1A-NuP84-00933R000500120001-5
/6C
Approved For Release 2001/07/12.: CIA-RDP84-00933R000500120001-5
/5Z,
?. f()
?? 1 loi6
.._. TAO_ ???????
-
???????? :OM. 0 ???/1???
? .1?????1
????
???????
??111,0 ?????=??
????
ap? ?????1?????
? ?????? ??????????
????????? ??????????
? r??? ???? 00.14?1, ?I????? ???? ?????????0
MI, 0. ????? ? ?MIO MOM ??10.
???? ? ? ? ? ? ? I ? ? ..1.1?01??????
?? ? ? ?
? 0??????? ?
??? ??????=? ????????
????1?1????????? ????=1. *I/ ???111??????? ?????????????
0?1?????1???????????
??? ??M, ?
???????
?? M. ??????/?/??????????????=...????? 61/M ????? ?? ????? ?
0E1 ? ? ?I? ???????????? ?0 ? ? ? 01 ?????? 01=?? ??=???
? ?????? ? ?? ? ? ? ?
.- ? t -
?
??? ?,????-, ? ??????? ??? ? ...um.*
??
??????
?
?????? ? I ? ??
rgti ? X A/ 5/ Q4/
? ?
FIGURE .1.9-13
Os. ?
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
Itioo
I3oo
. goo
lot
1oo
1_05 Tc
S 700
6o0
C
goo
Ho O1
3400
_ 0200
0 ??????
For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
" ......01??????????? ?
?
*
??
f
? ? .. ? .
' I
FIGURE 19C
? ???
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
ed For Release 2001/07/12 CIA-RDP84-00933R000500120001-5
Pico
IJa7
1A0D
/evo
. ...Mho
LoST 6'O
t'outi .......?????????=w4or??????dik.mr?mm
. 1/4" ?
am. 4?Emr gsmom .mmss 40 410 ow,
? ? ?????? ??? ??? ?????? ??????? ea MD ??? ????????????
????
????????
. - .
740o.
??????
-
??? ?? ? /0 0
???????????? ?????????????????
NIP AM
???? ????????? ?M?
?? ? ?????? ? ????????
?
abo????? ???? ? ?????? ?
- - ? - ? ? ? ? - ? ? ?
?? ??????? ?????? ? rime rep mim????????? ? ??????
??????
? ? An. ?????? ???????/?? ? ...........???????r ???????????? ?????
? ? .1. ????????????????.
. ? ??? .?? ????? ? ?? ? ? ?r ? ?????? r?billior???? ,?????e ? row.. ????=, to.
. tet . r. _ i. Ic6./6 . Jr...ga V/74 02
I
????....???? ??,....00 ? r ? ??????????? 11.????????? ???=. ?
0????? ???????? ? ? '? ;??
, SYSTEM EXPANSION FACTOR
?-- Amor a* ????????? ??????
?
FIGURE '19b
?
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
? ???r, - ? rm. - ? ? ?
??? ???????? r ? r
?
ApLoved Fo1r Release 20(iltib2 : CIALIRDP84-01901:1114-11-
BOLLIOPPY. FRD3LE17 s
CACHE (10ni
MAIN Cio sec)
nh,4 ?mil
)iSiclelacW
. FIGURE 20
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
CeLL
WTE (11. CT V EsPcnis
V 12.-TuAL M GM ofQ,'
MI4 nIA-qe-meATT-
? FIGURE 21
?
uts era 1E0 WTH
LI M 1i-A T1 oil
Approved For Release 2001/07/12 : CIA-R0P84-00933R000500120001-5
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
LAB AUTOMATION 5100 - 7406 (DEVICE COUPLER).
LAB AUTOMATION SYSTEM 7s TPVM
NETWORKING VNET (RSCS)
NETWORKING DATA STAGE
.TEXT PROCESSING TTF
INFORMATION BULLETIN BOARD
INFORMATION NATURAL LANGUAGE INQUIRY
MACHINE TOOLING VS1 - APT
SPEECH FILING SYSTEM SFS
DATA SPACE MANAGEMENT MIGRATE - ARCHIVE
CMS BATCH NEW BATCH. MONITOR
DATA SPACE MANAGEMENT ? CPDIR
LIBRARIES
19D - 19E - 19?
FIGURE 22
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
0
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5 ?
?
?
? ?? ? ? 4' ' .1 . k.
??? ..-4 .4,?Ir i; $ L.-, ? .
' .. ? ...V.,/
.. ;.'s .e. ...!
? t-k?,.;.;
I . ? ? t '
It-
if - ? ;?-;
, %II' .-. ? ,,.. 4
i - f'?? ???? 4 . I ' I .
..t e? ft, :et.
? ......t. 1 ii
? ? ' . ' %.4.114.,V.
. - .? 911.1
:=:.4.'.47.:41?;''' -
....s ,e)
- f .
? ?. -4 -.! l'r
? i, ,,,:p1 ?
RESEARCH DEVICE COUPLER
PIGURE 23
?
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
? ..?
!
I Approved For Release 2001/07/12.: CIA-RDP84-00933R000500120001-5
? ? .
:? ?
135-1200 BAUD
HOST d MODEM
? ?
RESEARCH
DEVICE COUPLER
SDS
PD
A
D
H
C
I
PS
I
IOS
MMMMMIVIMM
? .;
?
?
FIGURE 2L
??
?
?
?
USER
DEVICES
?
?
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
? ..
Approved For'Release 2001/07/12 :*CIA-RDP84-00933R00050012000.1 -5
? ?
?
?
?
?
?4
?
? ?
. . . .
NO
. . -
?
?????
? ? .
?12%
? ?
?
ONO ??111 ???? ????????1
? =rum
11. I
?P ???????1
TAPC CASSETTE
? FIGURE 25
Cd
IZEPt Ectammv4T
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
?
atSG.
5 Y 5TEM/7
PT'S.
V1t1-LA05/7
pprov
ase 2001/07/12 : CIA-RDP84-009
(usGtC.60/
el
12000175
user_ 'V
SYST-0A/1
Vfil -Li1-135(7,
SBak
P TO Ai 5his CAA1'8&-
_ ;41 17/1-cFt 51) )
r
? 1 MP4 E
4,4
.YA;5;1?
LA OSI-15 OtACCE
FOg,TZ-lAt So fC6
NETWOKK
#PAT4 F" I 1,.6.5
STAT us FILE'S
exec FU Es
Loh 1) Avr ("Int ES., sm(4.1eY eicoq go4Pts
ocu g -rn o
4.04*.t...",12".? ?
?
???11.1?1?11110?1???????????
????????????????????
MV5 M/X1-1 id&
fiSs em BL.G
comp(LE
L NM.- EDIT
FoRMAT'
FIGURE 26
...111???????
? ? ?
(vi VS DR-T4 5Ers6A3si
Mg. it...4 LA:r Tull. es,
1,16gA-zz
511 co hi Ficu. eh Tt
(oVE egit Stri
7/0 copl(cicAf./9-rio.
(NI& gKPT.)
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
? Approved Far Release 2001/07/12 : CIA-RDP84-00933R000500120001.-5
41.61111m?
$4 $1
I $11.$$$? I
its ?
11??????.?????????1
tt.
?
t s ?
??=??????????,-
s
??????????????1?????
I
t $ IA
:34:tt? I
_
$ KOLAS I
???????? II
1
1111/1/11:31.
4?????????????????
???????????mm????
??????=?????=??????
??????=????=??
ft4.01
??=??????????????i d?m???????????,?
en II 1141
? ? s ? $ until. ? ?
FOSess I ? 1.1,...0.1? I
11???????pm????=.1 11.????,????????
SON *OWN
poi sisommormutomolt Tit ??????
I S. ??????????????? I
I I
I I t
irmitimmtil?????4 ft?????????????????, 1??????????????????? ????????111,
Ws tot we toss tie I t?S isms vitt
? mosso 4 mom ? 1,1/41:/:3 $???????.$ j
I#.,.$. I $11113?114 tit" 1 ? Irle?1t I lPdep
-
111M11?1??.??=???=1 Im?m????=1?1=1?101 . 11/eNEN?1 thmlnlimi???????????
?????????mommEmIt
8
ommomrssommoi?
OW, SW
?
?
?
?
fp?Pwommbotem????
I ? vs Of ?
MOW SO
titoo?Or
?
CO I 3140
.1111????dinnl?mmti. ??????.?,?????????
i Mi ST, I int
II Org:111$141 ?????^???????????4 i u.osu ? g
Mk $ i ? s I
i?.............. 4..........,...?
IMO
???????????????????????
1$ tuts ??????=mmo4
? $ ? ? ? I
????????????????? ???????=????????????? .
omit.
? ?
?
1.11A?4
e8.611/l4 I
emittmtumm???=???????? ?
?????p???????1,
$
ostimm?mm.
?
4 *woo
11.?????&????????????
11.4 I
??^????1 111=115/ 0????
II I ro
toil ?
?
1?3111,10I rIte ?
$ 14?11?940. I lettloo
I WI ? 1?2/1,1
Illoommoomm,??? Iltmoniontmomm?
?
?
"NIP
?
OPAL IS?WIle
?
,t?
SUN NETWORK APRIL 1 9.75
FIGURE 27
? ' ? ? Ilt?
?
ISO BP ICT
?tt? 1. .? ? - -
? nom** s
? vs 1 ? st
11?1?44
WM/
I I
? mut a
mew.
Immum???????
omoOmmoommo,
'MO 14 Se
trts11
I nine ? I
1$11?0?1?????=?????11=1$
????? .11?Pill? 40 .0.
?
?
'
?
?
? ? ?
? ? ????
?
?
?
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
Ap
,a1 IN?66.6.64
I 1.6 I
? 411MMIVIMONMENNOM111.
ra
I'.J.4?$ I p 6?6?1.666
06 I I I Oft? e? 660
?44.444.???????? 6
11614 1 1 1 1 1, I
IC. 1 ? 4 88480
rdlieritTrizRelease 2001/074+2-.-QA-R14p84-43=3R0CI0i00120001-5
14.4 ???????? I ? 4.410 I ???? ?
118 80 I .
*W:6 6 ?
#616"4/0 I
?????
888
488
4114
I..
O. 1.??????????? 44.4?44.44
I. ? 616?6??? 4
????? tr.. 4............? ......? in.:
8 80.4.44. I
:I. ?./....r...? ?mr...........?
.41
1446
P? i! .
1184
I.1 4
1?14 .
? 4 44................. ..1............?
11141 MOO 1 .01
8844, 1.640 ? I 4 .416004 ?
8,84?40 4 410 1
u. :14.............. ???????????
II. 8444.
004 ..............r. ?
4
14.80 4
?
8840
...*?????
?
^ ??????? ?
?
????=0??? #mmemmemmwom.
I l???
. maps? ? .........? I./. ? ?
??? 491466064 I
????????? .bommommtammea
1 144$ 1.6
I 64.0111
001141??? 1
?
104
4 064.044 ? ?
1.
a.r....r...? .............ra
?
. 8841* ? ????
? ?
???????????????6# .....?????????=.11.#
? 1614 IAA* 6 6464
$.04646 0..........., ??????? ??????
0114/680 i I #41 # 1
*0 ?
I.
#??????????., $1?6????????????=m, vammrimmmo666.# #60.romm?????? a
?????????? ? 8.14 5 /61. NM* ? VIsb? 6
?????? Pr..I *04. I ????? I I 00468 I 111
ww ? or 8 1104.08.4 $ ? 406 8 I'
f#44,44444?4
4.4 4 4 I I 11116166
466............. ....../....., )4144 #1mm?=6?????? # $.#1=.41.?????????? ????????=6.:,
14.80 6664. ?? 1 10.......? , #4#4 1 ???? ???????11 ???611
II... b64$ I 46.66???????4 6641110 #6......mommn$ 04115 ? .......? WO. .......... 11.66 WO 4
?4848 ???????? I $10 WU III 1 ....4. 0..... ??? ? I
Me 41??????????????????? 11,0?0?????
84 114 I 611.0 It 1 'T1117."..
11?811 41.........r. # I 4
4464 41 be MN I
1$444..... .111111MOMIIIIMMIIIMI.P. ,IMMIN????ft.
4.4.1 *Ma .????? 8 14664 P?14 1 1146.1
????? 0414. 4..........0 666.111?4 116?6?????????? ??????? I
MI POP ? PIO I ?????116. I
1$$ #0.1?411* ???=4?4?44444=144
6.11
14?4
to./ 6441101411
SO 1 ??
61:6
4
144 OWN
884
??? es???????????ft
.4. 044
???? 141411154
*44p.4
I ??=ftwmg/.???????
1444
14 ???????????
114
84.8
804, ???????? 1
8+4.4
$4661.????????
???????.????.,
1.4
000000 1
60.
17.:1114140
MOB 14..4.
? 1 1
??????
10.411 4141
???????????????
????? 11.
Mb. ?
? 1.1
??,#I#
?
?
ii............... ,,MIIINIAVIIIIIIIIII. $1.111?61,0111$????????#
I ???4 .44 4 048 8 1461
1 44410 0...........4. 144166 fr.,* ? 04144 4
0???????? 6?.?????????? 6111 II?????? ? 414. ?
.......r........ / ....r........ ? ?
I 8? .................,
4"...... ? I I
? 6 0~ 46.6.?????????
.?????????? I OW 6???????????????? 4 1
1 46644.4446?04-_? 4114.... 1
4 4044.6 4,....4 44080 /
????????? ? 4.? Or ? t4
mr...... ............ 4
4 4 6 a
? $ 1 14
1.14111 ? 84014 114
#11MININ#110?110.1m. #01.1..bilimm##$$ I r.....d. .
1 ????????? ? 86.4 8 444 14
/ ese14141 .............? ,???? 4 1 ..1 wow* 1 1 I.
? ?? 1 44 ? * 8.4 ? 14
............. ...........11 I 1............... .
I
vo? Amil??????????? $41//1 II
I I II
1 v??????????? 14
4.???? ? 4 ????? 11/
"."...... ? es. 4 $ $ $,.
6064 660 I II
?????????????????? $ #
6mmeamm6?????# ?????????6?????? ???????=nonm?????
l 61666666w 0 ????????? ????
4
1 ....? 016 ??????????????? ?????? ?????????? I 60.4 ?
I
8?1 600
????????????? ???????????????
? OMNI. 66.
V
^W*
MON.. 1 14.4
? 1604 I I ????? I
??? 1 ????I
toommemarems? ?mmemengra4
44010
I ????????
twast ???????? I
14.~... ?
Ihmleammol
110
a
?
?
1
4.4.4 ?
1 1 1.4 ?
I 1..4 114511415
Ss
????????????????
?I_ I
? ?
I 4011.1
1 I?441
? 4.00.44 1
1
?????????????
I 0444108
??????
?
441446
.1i1OMIIMINN$
1004
844 14406411 ?
1?11
1114?
1 IMMO 1
1?? ?
oss ?
???????
64010
1
4444.4????44=4?4.
????
?
I 010.?
810 1
??
?
???
44???.m4???????4.
4#410
? ????? ?
I P.A..
???????44
01410
1 1?6I
1 b. ?
????
1,4 ?
?
# ??????
??? 1
?
MN 1
1
$11?06???????=m, $1:??????????mm$
1644 ???? 1 1640
??? W. I
??? ? W.440 ? 1
.1?1???=11..,,11
I
It
#???????????????$ la
1
41461 lin 66
? ???????? I 160
6 IVO I 16
1 .............. ? ?
6 V 84
g On. 41
? i Ii
?????............ I ?
? M. 44
? 64.6144141 $ 11.1
40444104 11. 16
.............? $ ?
10
411
? ? ....
................. II
I 14.04 11
? 1450418$ 1 $4.
???????64 I 11
gle....m.....??????? 0$
6 l?
OS. 41114 41
? II
? /. ..
.......... ? ?
? 4.0,4?414 11
.....4 ?4 44444 4 880
4.24/?44 1 11
*....'".r"...."1. ? 14
........,-......? It
:14 SO gbh 41. II
8 Ili
???=???????? ?????????=.=???? $ $
6 4???????? ? 1???? NAN II
I 16?1106 ? 11????????????11 h. 8.41 ? I 8*
14?11444 1 14
$???????????#!$/$ I,
CO
teen* 4 ?
'I/........... I ?
? ft.. ???? 11
r...6 ?? 644.? ? I 1*.
4.1040.44 V 11$
?imatimompmvarama. 88
I 44444+????????
? Is 11,........A I, 4444
0484 1 ? 1......... II ???? II/I ?
?? 1......4 06..4 I
1............. 1...".........
1 ? ? .4. ?
1 1 4 ?
armor.. ? 14.4
I $.66?????????????
1 1 1644
? 4 0014.411 I
? 41
8,."11114
?
1?14. 64010
?
????????????=6.#
6E64 1646
$46101?66 I
??????
4.1rmr.?????
????? ftddl?
6m.$ ???????? f
4440.4?16 ? $
la
?
?446.
foomaimmomm. 4444444444 444.444,44444444
1 i.4;.04
44.010.1111 I I 06044
4I? 0.r O.
4...???????????..
1
.................? 4
11444 II..
,...4., ...ft...4,
? 1444 1 14.0
? ??????? 1 8 ?11?1611 I
ION l 160 #
114?44444=4444414 44.4444?44,,44m4.4
:15.
4
#1?11?011,11.011???????# fl?MINO.161.1=???##
i 16400.4 1 1111.1
4 4484. ......4.....4 ????? e
....... I 4-. ?
?
NM. m ?
,smows??????66, #.66???????????.66,,
84
04????10 I 6649
I 111411 66641 I I
8.6 I ??? 1
? SUN NETWORK NOVEMBER 1976
FIGURE 28
Approved For Release 2001/07/12 : CIA-RDP84-00933R000500120001-5
?
?
-0
TTF SERVICE MACHINE St'MMARY USAGE sTA risTics laage
1
n
n
Approved For Release 2001/071W 6CIA-RDP84-00933R000500120001-5
August September October November December
January
Total Monthly Job Counts
I
1:.
Total
3054
3494
3354
3652
3724
4354
:
i..
TI EXT
Non-TTEXT
2454
600
2688
306
302$
329
3087
545
293 I
743
3579
775
First Shift
2568
2739
2926
.2854
2999
3694
Normal Workday JOb Counts
Total
132
161
134
176
171
199
TTIEXT
107
. 124-
14$
151
139
163
Non-TTExT
26
37
39
' 24
32
36
First Shift
114
127
141
? 133
139
170
First Shift Small Jobs ? ?
95
102
112
113
1 1 2
134
First Shift Large Jobs . . ?
20
25
.29
2.5
- 27
36
' ? '
Percent Jobs From
..
v . . . .. .
66
67
56
52
46
43
T . . . . . . -
22
23
32
3X .
40
46
NI . . . . ? ... . .
9
8
11
9
14
12
Percent Jobs Run On
T
100
81
92
62
100
100
v.
0
19
8
33
0
0
User Counts
.
Total
152
185
190
193
175
184
TTEXT
128
ISO
154
153
146
156
Non-TTExT
47
51
60
70
' 54
59
Normal Workday Total . .
40
43
54
56
52,
59
Normal Workday Trt:XT . .
35
41
45
50
44
52
Normal Workday Non-TTEXT
6
9
10
X
9
9:
Monthly CPU Minutes
-
Total
403
452
53X
641 -
620
660
TTEXT
379
422
498
611
' 5.32
620
Non-TTEXT
24-
31
39
29
.
33
40
First Shift
331
339
409
456.
479
557
Normal Workday CPU Minutes
Total
1$
21
26
31
19
30
TIFXT
17
20
24
30
27
23
Non-TTENT
1
1
2
?
1
2
2
First Shift
15
16
20
22
22
26
First Shift Small Jobs .
9
9
11
12
12
12
First Shift Large Jobs .
6
7
9
11
11 -
13
TurnaroUnd Minutes
First Shift Small Jobs . .
21
27
26
?
23
16
12
First Shift Large lobs . .
22
24-
30
? 54
' 34
23
DRAFT vs. FINAL
TTFN I'
Percent DRAFT ( only)
69
53
64
. 67
61 '
61.
Percent FINAL 0-rvxr only)
31
40
35
32
37
- 34
Total Monthly DRAFT . .
1688
1557
1922
2060
1326
2172
Total Monthly-FINAL . .
1356
1330
1870
1551
11134
2063
Normal Workday DRAFT .101
.
73
71
91

84
97
Normal Workday FINAL .
59 .
117
. 90
73
X3
, 96
RTC vs. STC (TtExT only)
?
Percent RTC ..... .
73
63
- 64
52
49
47
Percent STC
24
28
32 ?
43
47
49
LATE Option Usage
Total
25
93
li5
151
170
84
W
Normal Workday Percent
1
4
3
? 5
6
? 2
Approved For Release 2001/07/12 : CliZaral4a933R000500120001-5
?