LETTER TO MR. W. M. SLINGLAND FROM(Sanitized)
Document Type:
Collection:
Document Number (FOIA) /ESDN (CREST):
CIA-RDP80-01794R000100230010-5
Release Decision:
RIPPUB
Original Classification:
K
Document Page Count:
21
Document Creation Date:
December 16, 2016
Document Release Date:
October 16, 2004
Sequence Number:
10
Case Number:
Publication Date:
December 12, 1975
Content Type:
LETTER
File:
Attachment | Size |
---|---|
CIA-RDP80-01794R000100230010-5.pdf | 999.85 KB |
Body:
Approved For Releaae 2004/10/28: CIA-RDP80-01794R000, p230010-5
25X1A
Main Post Office
Washington, DC 20013
12 December 1975
R E G I S T E R E D
Ampex Corporation
Advanced Technology Division
Terabit Memory Systems
1020 Kifer Road
Sunnyvale, California 94086
Attention: Mr. W.M. Slingland
Reference: Cable to AMPEX Corporation from Contracting
Officer dated 4 December 1975.
Gentlemen:
The error rate stated in section la of the referenced cable
is wrong. The specification states an unrecoverable error rate
of only 1 in 62.5 billion bytes. The specification also states
there is a raw error rate, without rereads, of only 1 error in
each .375 billion bytes.
The translation of error rates into file loss at 70 files
each day stands.
Very truly yours,
STAT
Contracting Officer
Approved For Release 2004/10/28 : CIA-RDP80-01794R000100230010-5
Approved F%rRelease 2004/10/28 : OA-RDP80-017 000100230010-5
AMPEX CORPORATION
January 9, 1976
Serial No. TIS-2/071
STAT Mr.
Contracting
Officer
STAT
Post Office
Box
Main Post Of
Washington,
fice
D.C.
2001
3
STAT
Subject
Contracts
TMS-2
Mass Storage
System
Reference:
(a) Agency Contracting officer's
V.E. Ragosine dated 4 Decembe
Cable/Letter t
r 1975
o
(b) Agency Contracting officer's
dated 12 December 1975
(c) Ampex Letter TMS-2/070 dated
Letter to W.M.
23 December 19
Slingland
75
Consistent with our Reference (c) letter, this will acknowledge re-
ceipt and understanding of the Agency's concern over Ampex progress
in the performance of the subject contracts as expressed in the
Reference (a) correspondence and at the 18 December 1975 Management
Review Meeting at the Agency. It is Ampex position that the "items
of concern" as stated by the Agency need to be discussed, understood,
and resolved; however, they are not of sufficient magnitude to in-
dicate that successful'completion of the contracts might be impaired.
Ampex is committed to the timely completion of the TMS-2 Mass Storage
System.
This letter is Ampex formal response to the Reference (a) and (b)
Agency correspondence. Our response to the issues raised in the same
sequence as Reference (a) is as follows:
a. We believe that there caeremisunderstandings between us as
to the testing approach and philosophy during the November
1975 Redwood City test. These must be resolved prior to
scheduling a formal Redwood City Preshipment Acceptance
Approved For Release 2004/10/28 : CIA-RDP80-01794R000100230010-5
STAT.
STAT
Mr.
Approved A%Welease 2004/10/28: CIA-RDP80-017000100230010-5
- 2 - January 9, 1976,
Test. Ampex entered the November 1975 test believing that
we had agreed at the 30 October 1975 Management Review
Meeting to proceed with the test on a "mutual discovery"
basis to assess overall development progress and status,
thus, removing the contractual "pass/fail" connotation
indicated in Reference (a).
Consequentl Am ex did not insist on its prerogative under
Contract Exhibit II, for prior approval of Agency
prepared fes-c p edures. Neither did Ampex complete a
full "wring out" of the individual tests provided by the
Agency. It was our understanding that it would be possible
to work out the details as testing proceeded. In the
formal context in which the November 1975 tests were run
and without prior agreement on test procedures, definition
of test sessions, and accept/reject criteria, it was im-
possible for Ampex to pass the Redwood City tests.
Based on the above and prior to scheduling the Redwood City
Preshipment Acceptance Test, a detailed agreement relating
to conduct of the test should be worked out between us,
considering the following elements:
2. Definition of test procedures; sequence of tests, and
planned action in case of'difficulty or failure
during testing.
3. Definition of discrete test sessions and the accept/
reject criteria for such testing.
From Ampex viewpoint, the November 1975 tests were success-
ful in exposing problem areas and demonstrating Ampex capa-
bility to resolve them promptly.
During the first four days of testing (test sessions 1
through 8), an inordinate number of failures occurred in
moving some files from disk to TBM tape (descend) and from
TBM tape to disk (ascend). These failures were diagnosed
and an -in-process installation of Read verify software as
well as an adjustment of the data channel hardware dramat-
ically reduced the failure occurrences.
AMP
Approved For. Release 2004/10/28 : CIA-RDP80-01794R000100230010-5
Mr.
Approved Ror.Release 2004/10/28 : CIA-RDP80-0172000100230010-5
- 3 - January 9, 1976
Thereafter, in test sessions 9 through 20, following im-
plementation of the corrective actions by Ampex, 95% of
the attempted descends (112 out of 118) were successful
and 100% of the attempted ascends were successful. On
this basis, no files were lost following incorporation
of Read Verify software and data channel adjustment.
Five of the six unsuccessful descends attempted following
corrective action relate to functions which Ampex had
never planned to support during a T1S-2 Release 1 PSAT.
The remaining item relating to.automatic demarking was
recovered on retry at a different tape location. Ampex
was not permitted to remove a known bad cylinder on the
staging disk from use during testing which resulted in re-
petitive descend failures. Detailed summary data on the
.ascend and descend success ratio by test sessions, indi-
cating reasons for failure and corrective actions taken,
is provided as Attachment I.
There were also eighteen failure occurrences during the
tests relating to software. Ampex acknowledges twelve of
these occurrences as functional deficiencies in Release 1
Software; eleven have already been corrected, installed,
and tested successfully.. Partial capability of the
twelveth item (set segment size less than 10) was install-
ed at that time and we will have full capability for Final
Release Software. The remaining six occurrences are con-
sidered by Ampex to have recovery implications which were
not planned to be tested but will be incorporated for
Final, Release Software.
Attachment II is provided to detail the software failure
occurrences and corrective actions.
The unrecoverable error rate stated in the Reference (b)
correction of Reference (a) is incorrect and should be one
error occurrence in 2.5 billion bytes read in lieu of 62.5
billion bytes as stated by the Agency.
As a result of the discussions during the December Manage-
ment Review Meeting relating to the four design/action
items, Ampex has re-evaluated its position on these issues.
This re-evaluation has confirmed that the specification
applicability trail within Contract D is not clear STAT
and is subject to widely varying interpretation differences.
AM
Approved For Release 2004/10/28 : CIA-RDP80-01794R000100230010-5
Approved F Release 2004/10/28 : CIA-RDP80-017 000100230010-5
STAT_ Mr.
- 4 - January 9, 1976
Further research and analysis of these design issues, in
accordance with our commitment at the December meeting,
has been completed and the following are Ampex revised
positions:
1. Regarding concurrent access to individual transport
drivers from both SCP's, the Agency cited Section
3.3.2.1.3, Paragraph 2, of the MSS Design Spec of
19 March 1975 to support its position. Ampex cited
Section 10, Paragraph 2, and the specific floor.
cabling diagram on Page 172 of the same specification
as justification for a different position.
In recognition of the desirability of separating one
SCP and one or more transport drivers from the online
production TMS-2 equipment, Ampex agrees to provide
dual floor cabling and manual switching capability
to enable the Agency to run one SCP and mane or more
transport drivers offline from the remaining equipment,
which can be run concurrently.
Ampex will provide the above capability as a hardware
function. The Agency should recognize that there are
other procedural requirements which it must be res-
ponsible for in order to ensure system integrity.
These include the proper sequence of operator commands
.to vary offline remaining DSS and CCS hardware from
the primary SCP before reconfiguring the system with
the second SCP. As the requirement for a split/dual
system has never been clearly stated, Ampex requests
that the Agency state clearly the purpose, intended
use, and example configuration of the system with
concurrent SCP access to transport drivers.
2. The specific hardware already ordered by the Agency,
second increment, and o tions are covered under the
Fixed Price Contract o
The Agency cited the STAT
MSS General Specification, Exhibit I, to Contract
XG 3766 as basis for the position that perfect switch-
ing of a maximum hardware configuration is a contract
requirement. It is Ampex view that the preparation
and incorporation of the Mass Storage System Design/
Specification of 19 March 1975 supersedes and cancels
the MSS General Specification under the "Design and
A"
Approved For Release 2004/10/28 : CIA-RDP80-01794R000100230010-5
STAT ' Mr.
STAT
Approved Fd&Release 2004/10/28: CIA-RDP80-017W000100230010-5
-- S January 9, 1976
Analysis" provision of Contract Further, ?ffT
it is Ampex position that delivery of a maximum
system capability was clearly not a requiremer~t
authorized,. planned, or costed under Contract
verted from a standard CPFF to a CPFF with a ceiling,
no costs were included in our proposal of 20 March
In addition, at the time Contract was con- STAT
D
1975 to'cover any expansion or reconfiguration beyond
the hardware ordered or covered by option in Contract
Future expansion capability to the maximum configu-
ration has always been planned for the TBM product
line to be accomplished consistent with marketing/
operational needs. It will be offered to the Agency
when available or when specifically required and con-
tracted for.
The present TMS-2 system design provides for incre-
mental module expansion with perfect switching capa-
bility to a maximum of six transport drivers, sixty-
four tape transports, two data channels, and two
external data channel processors (EDCP) within the
currently structured contract ceiling. A'design
change currently being implemented will enable field
retrofit modification expansion to four external
data channel processors.
In reviewing the contract performance requirements,
the hardware and software capability for the existing
system, and the above stated maxir.um configuration,
it is Ampex view that two data channels and two ex-
ternal data channel processors are sufficient to
handle the workload of a full six transport driver
and sixty-four tape transport system. This is based
upon utilization of one data channel during second-
and third shift operations for dispersal functions
involving preparation of PRODTAPES and SPECTAPES.
Our detailed analysis supporting our position that
two data channels and two EDCP's are sufficient to
handle a maximum system workload is provided as
Attachment III.
A P!X
Approved For Release 2004/10/28 : CIA-RDP80-01794R000100230010-5
Approved Felease 2004/10/28 : CIA-RDP80-017SdR000100230010-5
STAT Mr.
6 - January 9, 1976
3. During the October Management Review Meeting, Ampex
indicated that a dedicated disk controller was re-
quired to access the MSS private files. The Agency
took exception to this, citing Figure 3 of the MSS
Design/Specification of 19 March 1975. Ampex reviewed
its earlier position and indicated during the December
Management Review that TMS-2 can operate with two
shared BSS/MSS controllers as shown in Figure 3. Our
detailed analysis supporting a position that two disk
controllers shared between the staging disks and MSS
private files are sufficient to handle the maximum
workload is provided as Attachment IV. If, after in-
stallation and acceptance of TMS-2 at the Agency, it
becomes apparent that the two shared disk controllers
shown in Figure 3 of the MSS Design/Specification are
inadequate to meet the representative MSS activity
shown in Table 12, then Ampex will reimburse the
Agency for the cost of procurement of an additional
CDC 38302 controller.
4. Ampex software for the TMS-2 Mass Storage System has
been implemented in a single storage control processor
which will meet the performance requirements of the
subject contracts. Other factors such as TBM tape
search time and data transfer rate are more critical
to system performance than CPU cycles available in a
single SCP. Splitting the software functions into
two parts, a master and a slave, may in fact degrade
performance due to the increased communication over-
head. The entire system becomes more complex, more
difficult to manage and operate, and more difficult
to develop, integrate and test. We believe the single
SCP implementation approach is beneficial to the
Agency as it enables a single SCP to meet full system
performance requirements while the second unit is
available for redundancy/backup purposes, similar to
the IBM 3850 Mass Storage System. Further, the off-
line/stand-alone dual system mode, made possible by
the additional cabling and switching being provided
by Ampex in connection with item I above, reinforces
the desirability of TMS-2 system software resident
in a single SCP.
A Approved For Release 2004/10/28 : CIA-RDP80-01794R000100230010-5.
Approved FbwRelease 2004/10/28 : CIA-RDP80-01792000100230010-5
STAT' Nr.
January 9, 1976
The Ampex Cost to Complete Proposal of 20 March 1975,
on which Contract II ceiling price was based, STAT
was predicated on continuation and completion of the
software approach then under development.
Ampex considers that the intent of the Master/Slave
SCP capability is fullfilled by providing for a stand-
by SCP "watch dog" program to insure that the slave
unit is available for system redundancy/backup in the
event of failure of the master.
It is our view that the overall TMS-2 software design
approach has been dramatically changed since the Ampex
Technical Proposal #T1MP73-l dated 24 January 1973 and
the split software was only a possible implementation
approach and not a specification requirement.
c. The TMS-2 Mass Storage System under Contract [::::A has STAT
always been developmental in nature. As a result of the
Agency's concern and Ampex confidence in the overall system
as defined in MSS Design/Specification of 19 March 1975,
Ampex accepted the responsibility to perform the remaining
development effort within a cost ceiling. The type of pro-
gram undertaken by Ampex and the Agency has never before
been accomplished and there are no easy solutions to
scheduling, planning, and organizing the work that remains
to be done. We have, however, accepted the responsibility
to perform the subject contracts and have entered into the
work with diligence and believe that we understood the
basic requirements. In view of the issues raised via
Reference (a) and our own assessment of progress as a re-
sult of the November 1975 tests, a total reassessment of
the completion schedule is in order following discussions
and agreement with the Agency. We are presently planning
a concentrated software system design/specification review
to plan and schedule the remaining software development
for the project. As we indicated during the 18 December
1975 Management Review Meeting, our target schedule for
.hardware shipment is July 1976 and for Final Acceptance
Test at the Agency in November 1976; however, resolution
of the outstanding issues must be accomplished in the very
near term to make these dates possible.
AMPS:
Approved For Release 2004/10/28 CIA-RDP80-01794R000100230010-5
STAT Mr'.
Approved FNYRelease 2004/10/28 : CIA-RDP80-0179iit000100230010-5
- 8 January 9, 1976
Ampex fully intends to complete the TMS-2 Mass Storage System at the
earliest possible time without reauest~any further funding beyond
the present ceiling under Contract assuming satisfactory STAT
resolution to the above issues. We do, however, believe that many
of the issues raised in the refrenced correspondence have developed
as a result of a breakdown in communications and understanding with
people at both Ampex and the Agency. To improve project communica-
tions, Ampex believes it would be helpful to reinstate a resident
Agency representative for the remainder of the project. Since the
completion of the renegotiation of the T,'S-2 project with the Agency
in May 1975, Ampex has doubled its software staff primarily to sup-
port this project. WW1e have restructured and reorganized the software
group and added a full-time THIS-2 software manager. In further re-
cognition of the importance of the project to Ampex Corporation, key
project and engineering personnel participate in regular bi-weekly
management review sessions for Advanced Technology Division Management.
We trust the above provides a full response to the Agency's concern
and look forward to meeting with you at the earliest possible date
to resolve the PSAT ground rules and other open issues identified.
Sincerely yours,
V. E. Ragosine
Vice President-General Manager
Advanced Technology Division
VER:fp
ABAPEX
Approved For Release 2004/10/28 : CIA-RDP80-01794R000100230010-5
ALLQ4..111LtC11L I
Approved F4 elease 2004/10/28: CIA-RDP80-0179iR0001002300 to Ampex Letter
~3~-/071
TMS-2 MASS STORAGE SYSTEM
NOVEMBER 1975 REDWOOD CITY TEST
DESCEND SUMMARY
SESSION
> DESCEND
4 FAILURES
% SUCCESSFUL
COMMENTS - See P.3
2
23
1
96%
(1) Reason I
3
31
1.0%
--
4
3
3
.0%
(3) Reason I
Software Modification
READ VERIFY INSTALLED
6
3
3
0%
(2) Reason II
(1) Reason I
10
3
77%
(3) Reason I
41
1.
80%
(4) Reason I
(6) Reason II
Hardware Modification
DATA CHANNEL ADJUSTED
9
15
88%
(2) Reason I
1O
.0
11
23
100%
12
0
13
20
1
95%
(1) Reason IV
14
21
1
95%
(1) Reason I
15
0
16
18
2
89%
(2) Reason III
17-20
21
100%
--
Total
1-20
229
26
89%
Total
9-20
118
6
9s%
After Corrective Actions Implemented
Approved For Release 2004/10/28 CIA-RDP80-01794R000100230010-5
It t- UCL l:i 11LLC 11 L. I
to Ampex Letter
Approved Fo`PRelease 2004/10/28 CIA-RDP80-0179ii 0001002300TOI,5-2/ 071
P.2 - TMS-2 Mass Storage System - November 1975 Redwood City Test
ASCEND SUM.:,IARY
SESSION
ASCENDS
h
FAILURES
% SUCCESSFUL
COtiLMENTS - See P.3
1-2
p1
3
20
95%
(1) Reason II or V
4
6
67%
(2) Reason II or V
5
0
Software Modification
READ VERIFY INSTALLED
6-7
0
7
77%
(1) Reason VII
(5) Reason V
(1) Reason VI
Hardware Modification
9-20 77
Total
1-20 126
DATA CHANNEL ADJUSTED
1~~ % No file loss after
Read Verify, installed
and Data Channel align-
ment modification made.
92%
Approved For Release 2004/10/28 : CIA-RDP80-01794R000100230010-5
to Ampex Letter
11
Approved NwRelease 2004/10/28 CIA-RDP80-0100010023001-2/071
uo-b
P.3 - TMS-2 Mass Storage System - November 1975 Redwood City Test
(15) Reason I
(11) Reason II
2) Reason III
1) Reason IV
Bad tracks existed at two locations on BSS001.
Handling of these tracks not planned for PSAT.
(Recovery type item.) Specific tracks were:
Cylinder 28, Track 8 & Cylinder 288, Track 5.
Data Channel Read electronics improperly set-up to
handle dro,outs. Found prototype unit requires
separate set-up procedure. Adjusted R11 on Assembly
621154' to eliminate errors.
overflow record which spanned cylinder boundary.
Function not planned for PSAT. (Recovery type item.)
Demarkable block found. Recovered on retry at
different tape location.
( 7) Reason V - Read Verify function required as part of each Descend
operation. Installed for usage after Session #5.
1) Reason VI - Data Channel wire found disconnected after PSAT at
Pin 22, J2, of Assembly 621139. This file was re-
c'overable after reconnecting this wire.
1) Reason VII
Reserve failure. Not supported. .(Recovery type
item.)
Two failures attributed to malfunction of this
device. Replaced first portion after Session #1,
remainder after Session n4.
SA11/TDIF Eight(8) failures attributed to inability of SAIL
Interrupt and TDIF hardware to properly handle simultaneous
Handling interrupts. Missing wire added to SAIL. TDIF
design modified. Solution implemented and fully
tested.
Approved For Release 2004/10/28 : CIA-RDP80-01794R000100230010-5
to Ampex Letter '1'MS?2/071'
Approved For Releas'et'B`A0,V1 W8,$ C CM$5-0A` 0100230010-5
NOVEMBER 1975 REDWOOD CITY TEST
SUMMARY OF SOFTWARE ERRORS
Session #'s
Occurrences Description
Fix Implemented
2
2 - Error messages & completion messages not
Fixed during Test
printed on flush command.
? "Setup complete" message in error should
have been "breakdown complete". Inter-
Fixed during Test
preter cancelled job.
Segment time limit currently at 10
instead
Final release
of 1. File transfer segmentation
limit
(R1 limited)
currently at 17 segments for descend-
disallowing large files.
d SRQ'full - request queue filled with job Partial fix during
stream $180311BL - set up for Biggie 1,2,3, Test
& 4. Messages prioritized during test to
drop low priority information only when
full. Final system will have sufficient
space.
? Flush failure - nothing happened.' Software. Final release
timeout to TDP after lost interrupt. Error
recovery will correct this.
? MSS aborted incorrect MFD entries due to Final release
manual restart procedures. Error recovery
will correct this.
? MSS sent incorrect message to host then Final release
hung looping. Manual restart procedure
and incorrect MFD entries. Error recovery
will correct this.
Key: ? Hard failure - caused system abort or required IPL to continue
Interpretation of operation - no files lost, jobs ran ok
Approved For Release 2004/10/28 : CIA-RDP80-01794R000100230010-5
Incorrect operation - system did not function to specification
AtLaeltl1 i1t i I
to Awpux LeLt(,r
P.2 - TMS-2 Mass StorageAp ediForRefeesi 0 1W7258 RG D@88C41t 94'116&10A230010-5
Summary of Software Errors
Session #'s
Occurrences
Description
Fix Im
le
t
d
p
men
e
13
1
? File conflicts. WM logic erroneously
treated DISP on file required for 3 steps
of a job.
Fixed during Test
13,16
3
? Priority conflicts - LIFO instead of FIFO
within priority.
Fixed after Test
13
1
? Non-recognition of change in BSS config-
uration unless MSG091 received at IPL time,
Fixed after Test
17,18,19
3
? Deallocation did not stop - min/max logic
not functioning correctly.
Fixed after Test
17
1
p Allocation lockout, space de-allocated on
disk without flSS notification, wait loop.
for space. Error recovery will handle this.
Final release
20
1
1 Display file abort - allocation lockout,
Final release
op command had conflicting information on
aborted. Error recovery will handle this.
Kam: ? Hard failure - caused system abort or required IPL to continue (6)
? Incorrect operation - system did not function to specification (9)
? Interpretation of operation - no files lost, jobs ran ok (3)
P.2 of 2
Approved For Release 2004/10/28 : CIA-RDP80-01794R000100230010-5
HcLdcriinent 11
to Ampex Letter TMS-2/071
Approved Fd2elease 2004/10/28 : CIA-RDP80-017'!1000100230010-5
TMS-2 MASS STORAGE SYSTEM
DESIGN/ACTION ITEM 2
Perfect Switching - Expanded System
Position
Up to 6 Transport Drivers may be configured in a TBM system with
perfect switching. Up to 32 Dual Transport Modules may be don-
figured in a TBM system with perfect switching. Up to 2 Data
Channels may be configured in a system with perfect switching.
Adding a 3rd Data Channel will reduce data channel utilization
by less than 5% (from 11% utilization with 2 data channels to
7% utilization with 3 data channels). Up to 4 EDCP's may be con-
figured in a system with perfect switching to 2 Data Channels.
A 5th or 6th EDCP requires a 3rd data channel.
Analysis
Based on the Representative MSS Activity in Chapter 11 of the
19 March 1975 design specification, the following tables indicate
Data Channel and EDCP utilization.
Data Channel utilization (expressed in hours busy and percent of
time available busy) is shown for one Data Channel (this. is the
degraded mode of the current contract configuration), for Two
Data Channels (current contract and present capability), and for
Three Data Channels (the disputed capability desired by the Agency).
The Read half of any data channel is utilized much more than the
Write half because read requests are more frequent than write
requests by a factor of about four and because the read half is
used after every write operation for verification.
During the daytime peak of 10 hours, a single Read Data Channel
can service all file read requests and read verify functions in
2.19 hours or 22% of the available 10 hours. With both Data
Channels operational, the utilization of the read half of each
is only 11%. A third Data Channel would reduce the duty cycle.
of each read data channel by only 4% from 11% to 7%. During the
daytime 10 hours, the present contract and current capability.
provides 8.9 hours on each data channel out of 10 hours avail-
able during which the read data channel is not being used. This
excess capacity of 8.'9 hours can be used to perform tape-to-tape
copy operations or to absorb an increased workload above that
specified for the present contract.
-P. 1 of 4
Approved For Release 2004/10/28 : CIA-RDP80-01794R000100230010-5
ALGdL 111ttC11l. iii
to Ampex Letter TMS-2/071
Approved F' elease 2004/10/28 : CIA-RDP80-017VAR000100230010-5
P.2 TMS-2 Mass Storage System - Design/Action Item 2
The night time activity has been shown compressed into 10 hours
instead of the available 14 hours. This presents a worse case
in terms of percent utilization and more realistically provides
for equipment downtime for maintenance or system testing.
The read half of the data channel is again the most utilized.
The production night workload would use 9% of a single read data
channel at night, 4% of each of the two data channels provided
in the present contract, and 3% of each of the three data channels
if three were installed.
The log tape dispersal activity at night requires more read data
channel utilization than the production workload. The dispersal
function (elapsed'time about 6 hours at night) requires only
1.76 hours of the 10 available hours of read data channel utili-
zation. Even in a degraded mode of the present contract and
current capability, a single read data channel can handle all
night time production workload plus dispersal with an excess
capability of 7.36 hours of 10 available hours. With both data
channels operational (present contract and current capability),
the data channel which is used for dispersal has an excess capa-
city of 7.8 hours and the other data channel has an excess capa-
city of 9.56 hours out of the 10 hours available. This excess
capacity of 9.5 hours can be used to perform other tape-to-tape
copy operations or to absorb an increased workload above that
specified for the present contract. Adding a third data channel
reduces the utilization of each data channel by about 1.5%.
EDCP utilization (expressed in hours busy and percent of time
available busy) is shown for one through six EDCP's. The EDCP's
are utilized most heavily during the day during which a single
EDCP can service the production workload in 2.63 hours or about
26% utilization. The present contract and current capability
with two.EDCP's can support the representative MSS activity with
a 13% utilization of each EDCP.
Approved For Release 2004/10/28 : CIA-RDP80-01794R000100230010-5
One
Data
Channel
Two
Data Channels
IF Three
Data
Channels
(Each)
(Each)
Read
Write
R
ead
Write
Read
Write
Day
.2.19
hrs
.44
hrs
1.1
hrs
.22 hrs
.73 hrs
.15 hrs
(10 Hrs.)
22%
4%
11%
2%
7%
1.5%
Night
.88
hrs
.44
hrs
.44
hrs
.22 hrs
.29 hrs
.15 hrs
(10 Hrs.)
9%
4%
4%
2%
3%
1.5%
Dispersal
1.76
hrs
.88
hrs
1.76
hrs
.88 hrs
1.76 hrs
.88 hrs
(10 Firs.)
18%
9%
18%
9%
18%
9%
(at
i
ht)
n
g
-
-
2nd
DC
2nd DC
2nd & 3rd
DC
2nd & 3rd DC
not
used
for
not used for
not used f
or
not used for
dispersal
dispersal
dispersal
dispersal
Excess
7.81
hrs
9.56
hrs
8.90
hrs
9.78 hrs
9.27 hrs.
9.85 hrs
Capacity
Day
Excess
7.36
hrs
8.68
hrs
7.80
hrs
8.90 hrs
7.95 hrs
8.97 hrs
Capacity
1st
DC
1st DC
1st DC
lst DC
Ni
ht
g
9.56
hrs
9.78 hrs
9.71 hrs
9.85 hrs
2nd
DC
2nd DC
2nd & 3rd
DC
2nd & 3rd DC
Desired by Agency
'Present Contract
'Current Capability
(Degraded Mode Performance
'(One of Two Data Channels down)
DATA CHAR UTILIZATION to Ampex Letter
.1010-2/071
Approved For Release 2004/10/28 : CIA-RDP80-01794R000100230010-5
P.3 of 4
Approved For Release 2004/10/28 : CIA-RDP80-01794R000100230010-5
to Ampex Letter TMS-2/071
Approved For Release 2004/10/28 : CIA-RDP80-01794R000100230010-5
EDCP UTILIZATION
One
Two
IF Three
IF Four
IF Five
IF Six
EDCP
EDCP's
EDCP's
EDCP's
EDCP's
EDCP's
(Each)
(Each)
(Each)
(Each)
(Each)
Day
2.63 hrs
1.32 hrs
.88 hrs
.66 hrs
.53 hrs
.44 hrs
(10 Hra.)
26%
13%
9%
7%
5%
4%
Night
1.32 hrs
.66 hrs
.44 hrs
.33 hrs
.26 hrs
.22 hrs
(10 Hrs.)
13%
7%
4%
3%
3%
-2%
Dispersal
0
0
0
0
0
0
Present Contract
Current Capability,
Current Capability with
No Development Cost
Requires 3rd Data Channel
P.4 of 4
Approved For Release 2004/10/28 CIA-RDP80-01794R000100230010-5
raiI-auIa[Len t IV
to Ampex Letter TMS-2/071
Approved For Release 2004/10/28 : CIA-RDP80-01794R000100230010-5
TMS-2 MASS STORAGE SYSTEM
DESIGN/ACTION ITEM 3
Position
Ampex does not require a dedicated control unit for the MSS private
disks. All functions can be performed using a shared control unit
configuration as illustrated in Figure 3 (Page 9) of the 19 March
1975 design specification. The system performance specified in
Chapter 11 of the 19 March 1975 design specification can be met with
the same shared control unit configuration.
Reservation
The Agency must be aware that there.may be some infrequent times
during which both EDCP's in Figure 3 (Page 9) are actively staging
two large files through both. of the shared control units. The
duration of this busy condition could be as long as 153 seconds
(two 403 cylinder files being staged in parallel). During the 153
seconds maximum busy time, neither SCP can access data on the pri-
vate packs. This means that no new file request messages from the
host computers may be received or processed. This is because the
private packs are unavailable for message spooling or master file
directory search operations.
During the busy time, the system does not fail because of the busy
condition, no messages are lost, and no error conditions arise.
The system simply must finish one of the two staging functions be-
fore it can begin any new work.
Analysis
Some recent measurements made on the system indicate the following
statistics for private pack disk and controller utilization:
o One MVT Job requesting one TBM file causes
100 program page disk accesses, and
120 data base disk accesses, or
220 total private pack disk accesses to service this job
from beginning to end.
Approved For Release 2004/10/28 : CIA-RDP80-01794R00010023001-0-5
to Ampex Letter TMS-2/073
Approved Foelease 2004/10/28 : CIA-RDP80-01794&000100230010-5
P.2.- TMS-2 Mass Storage System - Design/Action Item 3
io A disk access requires 2.5 milliseconds of SCP CPU time
to set up and issue and 25 milliseconds of SA-11, control
unit, and private pack disk busy time to execute.
o The above numbers are worst case.
- memory. mapping and subtask work space swapping (both
items planned for final release) will reduce the pro-
gram page rate by a factor of three or four.
system tuning and some repackaging will reduce both
data base accesses and program paging by a factor of
two or three.
- one file per job is not typical and the overhead per
job is not a linear function of the number of files
per day used below.
the LRU disk management scheme means that many files
may not require staging and therefore will not require
as many private pack disk accesses.
The peak daytime workload (0800-1800) from Figure 12 (Page 158)
of the 19 March 1975 specification shows 2500 file's ascended or
descended. Assuming the worst case of one file per job, this
represents:
2500 files (jobs) in 10 hours
X 220 private pack disk accesses
550,000 disk accesses in 10 hours
X 25 millisec busy time per access
13,750 seconds (3.82 hours) out of 10 hours during which
the SA-11, shared disk control unit, and private
disk packs are busy supporting the SCP.
During this same 10 hour peak period, 7875 seconds (2.19 hours)
are required to stage or destage.20,000 cylinders of data for
2500 files.
Therefore, if there were only one disk controller servicing both
the private disk packs and the backfill storage disk packs, that
control unit would be busy 6.01 hours (3.82 + 2.19) out of the
10 hour peak period or 60% utilization. This is the worst possible
case and yet performance is easily met.
Approved For. Release 2004/10/28 : CIA-RDP80-01794R000100230010-5
t r_LCL ill[1Cl1C IV
Approved Fcelease 2004/10/28: CIA-RDP80-01791 0001M2EP?* Letter TMS-2/071
P.3 - TMS-2 Mass Storage System - Design/Action Item 3
With both control units functioning, one of the control units is
busy 4.92 hours (3.82 + 1/2 X 2.19) and the other is busy 1.09
hours out of the 10 hours available. The representative MSS
activity could double before the shared control unit is saturated.
Interpretation
All of the above is worst case. With memory mapping, system
tuning, two or three files per job, and LRU management of the
staging disks, the representative MSS activity could increase
bya factor of five before the load on the shared private pack
disk control unit became a critical factor. If and when the
load does increase to this level, then the remedy is to add a
third (non-shared) private pack disk control unit.
Approved For Release 2004/10/28 : CIA-RDP80-01794R000100230010-5