COMPUTER SYSTEM PLANNING REPORT MARCH 1967 - MARCH 1968
Document Type:
Collection:
Document Number (FOIA) /ESDN (CREST):
CIA-RDP80-01003A000100020001-2
Release Decision:
RIPPUB
Original Classification:
S
Document Page Count:
58
Document Creation Date:
December 14, 2016
Document Release Date:
September 26, 2002
Sequence Number:
1
Case Number:
Publication Date:
March 17, 1967
Content Type:
REPORT
File:
Attachment | Size |
---|---|
![]() | 1.97 MB |
Body:
Approved For Release 2005/04/(7 F 1 -RDP80-01003A00 0020001-2
Woo
COMPUTER SYSTEMS PLANNING REPORT
March 1967 - March 1968
OFFICE OF COMPUTER SERVICES
DIRECTORATE OF SCIENCE AND TECHNOLOGY
CENTRAL INTELLIGENCE AGENCY
17 March 1967
GROUP I
Excluded from atito?-
Approved For Release 2003/04/17: CIA-RDP80-01003A00010062OOO1l2wngradi.ng
S E C R E T and declassification
Approved For Release 2003IO 111'F:%IA RTDP80-01003A000100020001-2
Page
1.
SCOPE I 1
2.
GENERAL PROBLEMS 3
2.
1
IBM 360 Hardware Experience
3
2.
2
IBM 360 Software Performance
3
2.
3
Conversion Experience
2.
4
User Needs
2.
5
Organization and Policy
3.
4.
OBJECTIVES
CURRENT EQUIPMENT STATUS
4. 1 IBM 360/65
4.2 IBM 360/50
4. 3 RCA 501 and 301
4-14 IBM 7010
5.'
IMMEDIATE PROBLEMS
5. 1 Computing Load
5. 2 Heterogeneous Systems
5. 3 Hardware Configuration
6.
SPECIFIC, IMMEDIATE OBJECTIVES
13
7:
NEAR -TERM COMPUTER PLAN - SUMMARY
14
~.
PLANNING DETAILS AND SCHEDULING
19
R
pP80-01003A0001000202
001-2
Approved For Release 20034/17 n' R : CIq E T
Approved For Release 200~0f/1Cl :-GIA-P80-01003AO00100020001-2
Page
9.
COSTS
23
10.
ALTERNATIVES, CONTINGENCIES, AND
IMPLICATIONS
25
10.1
Alternatives
25
10. 1. 1 Alternatives to this Plan
25
10. 1. 2 Alternatives within this Plan
29
10..2
Contingencies
30
10.3
Effects on On-Going or Planned
Applications
34
10.4
Effects on Long-Range Computer Systems
Planning
36
TAB A - Computer Utilization Graphs
TAB B Computer Load (Jan 1966 vs. Jan 67)
TAB C - Cost Analysis and Justification for RCA Spectra 70/45
TAB D - Equipment Planning Milestones
TAB E - FY-67 Rentals
TAB F - Equipment Rentals and Purchases (FY-68)
TAB G - Summary of Purchase Items
Approved For Release 2003/04/17 : CIA-RDP80-01003A000100020001-2
S E C R E T
Approved For Release 206WC I/ 7 C! -JDP80-01003A000100020001-2
17 March 1967
OFFICE 'OF COMPUTER SERVICES
COMPUTER SYSTEMS PLANNING REPORT
March 1967 - March 1968
1. SCOPE
This report provides a short-term plan for revising and ex-
tending the computer systems in OCS. The plan considers only
general-purpose computer equipment; it does not include a discussion
of related peripheral devices in the Computer Center.
This report should be viewed as an addendum to the "OCS Com-
puter Systems Planning Report," dated 1 June 1965; it provides an
updating of that plan for the next twelve months. It is needed now
because significant changes have occurred in user requirements and
in the assumptions concerning equipment performance and OCS
capabilities .
As we gained experience in the use of IBM 360 systems and
became aware of deficiencies in performance and delays in software
delivery, we made several changes to the original plan. But, until
today, the changes in the plan were concerned with interim means for
providing a bridge between second-generation and third,? generation
Approved For Release 2003/04/17 : CIA-RDP80-01003A000100020001-2
Approved For Release 2003 O 1 :il-FpP80-01003A000100020001-2
equipment--the IBM 360/67 remained the ultimate goal. This paper
represents a major departure from the original plan because of recent
announcements from IBM reflecting their inability to design and imple-
ment an effective software package for the 360/67 system within the
time frame of concern to us.
Even before this announcement, however, we knew that we had
to contend with an increased computational load with sub-optimal 360
software performance; we also recognized the gaps in our understanding
of precisely how the 360/67 would accommodate the wide spectrum of
computing problems the Office must handle.
Briefly, we have concluded that what were interim measures
before must now be considered as semi-permanent. Before we were
focused on the 360/67 as the third-generation equipment goal, with the
assumption that this goal was clear and the nature and sequence of the
interim steps (smaller computers and interim software) was well-
defined. Our view now is that, having made the first few steps, we find
that the interim systems are not performing as anticipated, and the
360/67 (with its time-sharing system) is not a panacea, but that our
investment in 360 systems is beginning to bear fruit. We want this
investment to continue at low risk, however, by making the most., of
systems that work and remaining skeptical of the unproven. This shift
in direction and pace need not be more expensive, will permit continued
Approved For Release 2003/04/17 : CIA-RDP80-01003A00010002b001-2
SECRET
Approved For Release 20V1Q407A 9IAIRDP80-01003A000100020001-2
full-scale service, but must be viewed as a slow evolution of our
overall capabilities.
2. GENERAL PROBLEMS
Many of the developments since the writing of the June 1965
planning report reflect major problems whose solutions require
modifications to the basic plan. The more important of these are
discussed below.
2.1 IBM 360 Hardware Experience
Major delays have been encountered in the delivery and
acceptance of 360 equipment. Our 360/65 was the eleventh such
system delivered by IBM and has required over 300 man/hours
of on-site engineering changes. Although delivered on 18 July
1966, it did not pass the acceptance test until 9 October 1966.
IBM customer engineers lacked experience with this system,
hence OCS operations suffered. Nagging hardware failures have
been a constant problem. When the problems are identified and
corrected, however, these systems meet the hardware perfor-
mance levels on which planning was based. Such is now the case
with our 360/65.
2. 2 IBM 360 Software Performance
The IBM design plan for 360 software is ambitious and com-
plex. The documentation is lengthy (it covers about five linear feet
Approved For Release 2003/04/17 : CIA-RDP80-01003A000100020001-2
Approved For Release 200 OI l1! . CIS-F DP80-01003A000100020001-2
already), but poor? The Operating System* for the IBM 360
has been in use only one year, but already has gone through
nine major revisions (this could be contrasted with the Operating
System for the IBM 7090, which is now in its fourteenth version
since its release about four years ago).. The software concepts
are valid, but their early implementation has produced ineffi-
ciencies and considerable error. The flexibility of the software
is a mixed bles.s_ing. The applications programmer must make
choices in areas where none were available (or needed) before.
This places a heavy burden on the programmer in specifying his
job and responding to diagnostics provided by the system. Con-
sequently, we have encountered a major learning problem and
program test time has risen considerably.
2.3 Conversion Experience
The logistical problems involved in equipment changeover
have occupied a considerable portion of OCS Staff time. The
modularity of the 360 hardware, while an asset to configuration
* An Operating System is a set of control and service programs
which perform such functions as loading application programs in
memory, allocating facilities needed by the application, translating
FORTRAN and COBOL statements into machine instructions, main-
taining program libraries, etc.
Approved For Release 200WV/17 jCJ - DP80-01003A000100020001-2
Approved For Release 2003)64/q '1ATRDP80-01003A000100020001-2
5-
planning, has provided us with a seemingly infinite number of
component combinations whose relative merits are difficult to
analyze. Conversion programming is now progressing at a
satisfactory rate, but, of necessity, has been given low priority
compared with meeting basic user needs. This task has been
reasonably straightforward for COBOL or FORTRAN programs.
.2.4 User Needs
Projections of computer.time needed by users have been
consistently underestimated by OCS and its users. In general,
our customers are becoming more sophisticated in ADP, have
explored new applications as confidence in OCS increased, want
to modify and expand lobs which have been in a production status,
and are anxious to exploit the facilities of third-generation equip-
ment. Further discussion of user needs is not given here because
of the current effort by the O/PPB Information Processing Staff
to generate an Agency Five-Year ADP Plan. We assume that we
are aware of most of the requirements projected over the next
few years. However, Parkinson's Law has applied to date. This
near-term equipment plan represents a means of solving our
immediate and critical needs.
2. 5 Organization and Policy
In the past eighteen months, several changes in Agency
polic and OCS organization have occurred; the former is
Approvey d For Release 2003/04/17 : CIA-RDP80-01003A000100026001-2
;SECRET
Approved For Release 2003/04/17 : CIA-RDP80-01003A000100020001-2
-6 -
25X1A
reflected in the publication of --]which provides mechanisms
for coordinating ADP activities and establishes a philosophy of
partially centralized ADP. In addition, a significant increase in
OCS resources ha,s been approved. Finally, our early experience
with the 360 systems has engendered a more cautious attitude on
the part of OCS management and its users with respect to the
schedule within which the full promise of third-generation systems
cazi'be realized.
3. OBJECTIVES
The basic objectives outlined in the June 1965 report remain valid:
the aim toward a homogeneous set of hardware and software, equipment
with large growth capability, provision for remote consoles, efficient
software from the operator and programmer points -of-view, and around-
the-clock computing capability. To this group of objectives might be
added a need to provide a stable hardware environment from the pro-
grammer/user point-of-view. The need to reconfigure our systems
several times in recent months because of significant new requirements
and manufacturers' failures has been most unsettling.
4. CURRENT EQUIPMENT STATUS
The current OCS equipment picture is as follows:
Approved For Release 2003/04/17 : CIA-RDP80-01003A000100020001-2
S E C R E T
Approved For Release 2003/04/17 : CIA-RDP80-01003A000100020001-2
4.1 IBM 360/65
The 360/65 system includes a 7090 emulator*, one-half
million byte (byte = $ bits) core memory, twelve magnetic
tape drives, four disc storage drives, card reader/punch, and
high speed line printer. With its high speed and flexibility,
this system has considerable computing potential, but its
potential has not been fully realized yet because of software
inefficiencies, notably in control program compilers and the
lack of a flexible multi-programming** capability. The multi-
programming delay is the most significant. However, from a'
cost point-of-view, this system can be justified if only because
of its replacement potential for the 7090. In the emulator
mode, it operates somewhat faster than the 7090 at lower cost.
4.2 IBM 360/50 ,
We now have two of these systems. The first one,
acquired as a replacement for the 360/30, provides our basic
input/output facility (tape-to-print and card-to-tape conversion),
and was planned as the prototype time-sharing system. It has a
An emulator is a combination of hardware and software facilities that
permits efficient simulation of one computer on another.
** Multi-programming permits concurrent operation of two or more
independent tasks on one computer.
Approved For Release 2003/04/17 : CIA-RDP80-01003A00010002b001-2
S E C R E T
Approved For Release 2003/vA7 iAER6P80-01003A000100020001-2
1/8 million byte core memory, five tape drives, two disc storage
drives, a card reader/punch, and two high speed line printers.
It has been operating effectively for one year as an input/output
processor with a'program designed by the OCS Technical Staff
which executes three independent tasks concurrently.
The second 360/50 was acquired in December and passed its
acceptance test in March. Its facilities include an IBM 1410/7010
emulator, quarter-million byte core memory, ten magnetic tape
drives, four disc storage drives, card reader/punch, and high
speed line printer. As with the 360/65, this system can be
justified on a cost basis alone; emulator performance is equivalent
to IBM 7010 performance at lower cost.
4. 3 RCA 501 and 301
These systems continue to perform satisfactorily and ab-
sorb a very heavy computing load- -a load which has required
backup of about 60.hours per month at the CEIR Service Bureau
in 'Washington. The 501 has 48, 000 character (character = 6 bits)
core memory, and the 301 has 10, 000 character core memory.
Other facilities include nine magnetic tape drives, card reader/
punch, paper tape reader, and high speed line printer.
St
4. 4 IBM 7010
Our IBM 7010 system includes 80, 000 character (character
Approved F)or Releasee g /~W/17e~yi Ib d603e01r0b0cfd reader/
=SECRET
Approved For Release 2003/04/17 : CIA-RDP80-01003A000100020001-2
-9-
punch, and high speed line printer. The load on this system has
been heavy in the past, but is dropping quickly because several
applications" are now being converted to the 360 systems, and
the 360/50 emulator will assume the remaining unconverted 7010
programs.
5. IMMEDIATE PROBLEMS
5. 1 Computing Load
The June 1965 report estimated that the OCS computer
load would quadruple over the next five years. Our experience
in the past 18 months indicates that this estimate was low. An
indication of the upward trend is given in the charts in Tab A,
showing for each computer system the hours logged each month
in CY-1966. All current systems are being used at least 500
hours per month. Additional statistics in Tab B give a
comparison of the operational loads for the months of January
1966 and January 1967. For example, the printing load for IBM
systems increased 44 percent, there was an increase in number
of jobs processed on IBM systems of 130 percent, the OSA/OSP
computer load increased by 72 percent, and RCA systems
processed 21 percent more work. An indication of the
current upward trend is the table below which shows the number
of jobs processed by the Computer Center on IBM systems and
the total hours logged and metered for all systems for the past
Approved For Release 2003/04/17 : CIA-RDP80-01003A000100020001-2
S E C R E T
,SECRET
Approved For Release 2003/04/17 : CIA-RDP80-01003A000100020001-2
- 10 -
five months (where a "job" is an individual run submitted to the
Computer Center). Note that the number of jobs has increased
by 50 percent since September 1966. Metered and logged time*
stayed nearly constant, at or near the maximum available in
the month.
Sept
Oct
Nov
Dec
Jan
No. of Jobs
2027
2462
2543
2674
3246
Logged Hours
1775
1709
1853
1621
1709
Metered Hours
1345
1257
1204
1129
1286
With respect to the RCA 501/301, as noted above, the load hap
exceeded the capacity of the Agency systems and has required
significant outside support which is cumbersome from a security
and procedural standpoint and is more costly.
The reasons for the increasing load on all OCS systems
- The conversion load is reaching a peak, which will not
slack off for at least another year.
- Existing applications, particularly on the RCA 501, have
expanded, requiring increased test time and larger
volumes of output.
A large'number of staff programmers are now writing
programs for the 360 and have encountered learning
* ' Metered time is the time (read from meters installed on the equip-
ment) during which the computer is performing work. Logged time is
metered time plus job set-up time.
Approved For Release 200// ~7 RCJ-~DP80-01003A000100010001-2
',SEC RE T
Approved For Release 2003/04/17 CIA-RDP80-01003A000100020001-2
- 11 -
difficulties which require extensive program test
time and many reruns. (This is a major reason
for the increase in the number of jobs noted in
the table above. )
J
- Difficulties in diagnosing basic source of System/360
errors -(hardware, software, applications program)
due to the lack of adequate diagnostic facilities --
which must be quite elaborate because of the many
layers of control interposed between the programmer
and ultimate machine executions in the 360 system.
- Several new scientific applications requiring many runs,
each measured in hours.
5.2 Heterogeneous Systems
We continue to be plagued with the need to handle a variety
of operating systems, programming languages, and application
peculiarities which demand special attention. When taken together,
these variations and resultant special procedures constitute a
major scheduling problem in the Computer Center. The Center,.
by virtue of its mission, must accommodate the needs of a wide
spectrum of applications. Commonality of procedures remains
an unattained goal which can be met only through patient attention
to programming standards and gradual changeover to .a new,
homogeneous software system.
5. 3 Hardware Configuration
The June 1965 report mentions the basic problem of the
incompatibility of the RCA and IBM equipment in the Center.
This problem remains, but is being attacked through (a) piece-meal
Approved For Release 2003/04/17 : CIA-RDP80-01003A00010002YO001-2
S E C R E T
.SECRET
Approved For Release 2003/04/17 : CIA-RDP80-01003A000100020001-2
- 12 -
conversion of, selected 501 programs, (b) the planned acqui-
sition of an RCA Spectra 70 computer which will be partially
compatible with the IBM 360 and will emulate 501/301 pro-
grams, and (c) ultimately, complete replacement of 501
programs with the DD/S SIPS System now under development.
A secondary'- incompatibility problem has developed
within the IBM 360 systems themselves: each of the three
systems we have acquired have different memory sizes and
peripheral devices. Because of these differences, very
few 360 jobs will run on more than one computer. We have
been unable to dedicate a single system to normal 360
activity (note that. two of these systems must be dedicated
partially to emulation and the third entirely dedicated to
input/output tasks). Thus, we now have a heterogeneous 360
facility which generates significant scheduling problems and
provides little backup in the event of machine failure. Finally,
one system--the first 360/50--was acquired in anticipation
of the need for experimentation with time-sharing applications;
it includes a variety of control units and terminal devices
which have yet to be exploited because of the necessity to
sacrifice the pace of development of this system for increased
Approved For Release 200* '1f : bI-RBP80-01003A000100020001-2
S E C R E T
Approved For Release 2003/04/17 : CIA-RDP80-01003A000100020001-2
- 13
attention to immediate Operating System/ 360 problems.
6. SPECIFIC, IMMEDIATE OBJECTIVES
The near -term (FY-67 ,and 68) equipment plan outlined below
should provide the means for meeting the following objectives:
a. A standard, stable, and powerful 360 configuration
provided to programmers (see para. 7. b. below).
b. At least two systems having the standard configura-
tion, which will provide backup and scheduling flexibility for the
360 (see para. 8.b. below).
c. Partial relief of the 501 computing load problem
through the acquisition of an RCA Spectra 70/45 with 501 and 301
emulators (see para. 7. c. below and Tab C).
d. A minimal stand-alone 360 system dedicated to proto-
type time-sharing applications which will be manageable and
represent a low risk to overall production needs (see para. 7. d.
below).
Ability to continue processing 7090, 7010, 501, and 301
programs for as long as necessary until all conversion is complete.
* Full exploitation of Operating System/ 360 is basic to our production
capability and requires full-time attention of several OCS system, pro-
grammers. It might be noted that the complexity and scope of the-soft-
ware is such that several thousand programmers at IBM are engaged in
its development.
Approved For Release 2003ree1 dI-RT)P80-01003A000100020001-2
S E C R E T
Approved For Release 2003/04/17 CIA-RDP80-01003A000100020001-2
- 14 -