pop CO/PERIO
1c Pathway Performance ” Fas
Pathway Live Syste Report Dann “Fo
29/03/00
Document Title: Live System Performance Report for February 2000
Document Type: Monthly Report
Abstract: This document reports on the performance of the
live system. This version covers the period January
and February 2000.
Status: Version Complete
Mike Coombs
Terry Austin
Hodaka Imanishi
Steve Muchow
Tony Oppenheim
Martin Riddell
Chris Wannell
Gill Jackson
Peter Wiles
Library
Allan Hodgkinson
James Stinchcombe
John Hunt
Mark Jarosz
Stephan Robson
Dick Long
Eugene Dempsey
David Tanner
Glenn Stephens
Harvey Potts
Allan Hodgkinson and James Stinchcombe
© 2000 ICL Pathway Lit COMPANY IN CONFIDENCE
Filename: SR Jan Feb 2000{32)dev Printed by AIH at 30/06/00 33:43
Nothing contained herein shal be deemed or eanshried a fecting exiting tontrachucbligations o ereating neareontrachial
cbliations between ICL Patharay and POCL
Poge 4 of 56
FUJ00079350
FUJ00079350
Pathwoy
Re CS/PER/O
Pathway Performance en,
Live Syste Report Distés oe aay
O DOCUMENT CONTROL
0.1 DOCUMENT HISTORY
Version
Date I Reason
1.0
27/8/94 I First version — start of Live trail until mich August:
Issued for comment uv TDA only. Not distributed
fully,
°
2.0 I 51/10/49 I Until end of October. Added data for reference
q date. componenty and the ISDN network.
3.0 I 24/01/0 I Report covering the month of December 19949.
0 — I Additional performance monitoring way
40 2.9/03/0 I Report for January and February 2000
0.2 ASSOUATED/ DOCUMENTS
Ref. Ver I Date ITitle Source
Model. 5 I 30/10/ I See Pathway Performance — I Pathway
49 Team
Perpscalstr I 2 I 25/5/4 I Performance and Scalability I Pathway
at 4 Strategy PA/PER/O1S
PerfMan I 0.2 I 7/10/4 I Performance Management I Pathway
q Service Outline Design
Busvoly 0.5 I 30/6/4 I Pathway Performance — Pathway
4 Business Volumes
Brief 5.3 I nla I Pathway- Workload Brief I POCL
0.3 ABBREVIATIONS
[Aes [Automated Payment Service
© 2000 ICL Pathway Lit COMPANY IN CONFIDENCE Poge 2 of 56
Filname: ISR Jan Fel 2000(32).dev Print by AJH a 10/06/00 33:43
Nothing contained here shal be deemed or canst a fecting exiting contract obligations r reat nesreontrachial,
cbliations between ICL Patharay and POCL
FUJ00079350
FUJ00079350
CS/PER/O
ICL Pathway Performance red 43
Pathway Live Systew Report pan Te
24/03/00
Foreigns I OBCS Foreign Transactions ues OBCS paymenty not made at
the normal or ‘home’ Post Office
Host Sequent NUMA-Q system
LsR Live System Report
Luc Cluster Lookup Service
NINO National Insurance Number
NUMA-Q I Sequent implementation of thre NUMA architecture,
PRI Primary Rate Interface of the ISDN Router
RDDS Reference Data Distribution Service
RDMC Reference Data Management Centre
SLA Service Level Agreement
SMC Systemy Management Centre
0.4 CHANGES FORECAST
Reissued on a monthly basis. The content of the report will be
continually monitored and revised.
Addition of section on resources used at the physical layers ie.
correspondence server, agents and host.
0.5 TABLE OF CONTENTS
1 INTRODUCTION.
4
LL SCOPE... 4
1.2 OVERVEW 5
1.3 MANAGEMENT SUMMARY 6
1.4 PERFORMANCE CHANGES... ia : casi " an 6
7
8
8
1.5 Issues.
2 OBS.
2.1 OBCS SUMMARY.
2.2 OBCS HARVESTING... 7 serrata a 2
2.3 OBCS ‘FOREIGN’ TRANSACTIONS. ci
2.4 STOPS PROCESSING wssnssinssnnstnininnsnnninnninnsnnne een
3 APS. 20
3.1 APS SUMMARY...
3.2. APS HARVESTING.
3.3. APS Host PRocessiNG
© 2000 ICL Pathway Lit COMPANY IN CONFIDENCE Poge 3 of 56
Filename: SR Jan Feb 2000{32)dev Printed by AIH at 30/06/00 33:43
Nothing contained here shal be deemed or canst a fecting exiting contract obligations r reat nesreontrachial,
cbliations between ICL Patharay and POCL
FUJ00079350
FUJ00079350
Re CS/PER/O
ICL Pathwaey Performance ‘ fas
Pathway Live Syste Report Dates ee
A TPS crsene
4.1 TPS Summary.
4.2 TPS HARVESTING
4.3 TPS Host PROCESSING ..
on
REFERENCE DATA sss
3.1 REFERENCE DATA SUMMARY...
5.2 REFERENCE DATA PROCESSING
5.3. NON CoRE REFERENCE DATA LOADING .
5.4 CORE REFERENCE DATA LOADING.
°
RIPOSTE. a — ne 32
6.1 NUMBER OF RIPOSTE MESSAGES 32
6.2. SIZE OF ROSTE DATA.
6.3 ARCHIVING...
ISDN NETWORK.
71 CALs
DURATION...
PEAK USAGE...
a
o
ISSUES...
8.1 UNDER INVESTIGATION...
8.2 SOLUTIONS IDENTIFIED
8.3. CLOSED.
© 2000 ICL Pathway Lit COMPANY IN CONFIDENCE Poge + of 56
Filename: SR Jan Feb 2000{32)dev Printed by AIH at 30/06/00 33:43
Nothing contained here shal be deemed or canst a fecting exiting contract obligations r reat nesreontrachial,
cbliations between ICL Patharay and POCL
FUJ00079350
FUJ00079350
FUJ00079350
FUJ00079350
pop CO/PERIO
1c Pathway Performance ” Fas
ersion
Pathway Live Syste Report Dann “Fo
29/03/00
1 INTRODUCTION
1.14 Scope
The Pathway Performance & Scalability Strategy [PerfScalStrat] describes
the process that ICL Pathway has implemented to manage and predict
the performance of the system.
One of the key components of the strategy is the measurement and
monitoring of the live system and comparing the performance of the live
system (workload processed and resources used) with the performance
predicted using:
+ Data derived from Technical Testing and
«Performance models.
The performance predictions are input into the scalability plans for each
component of the system that will:
+ Document the capability and scalability of each layer of the
architecture
+ Relax the stringent targets set out in the default scalability plan,
where applicable
+ Identify any shortfall in performance that will impact on the
scalability of the system.
The edition of the Live System Report documents, in summarised form,
the monitoring of the live system for January and February 2000.
The Live System Report has four main purposes:
+ To document the workload processed (no of messages, transactions,+—{ Formatted: Bullets and Numbering )
etc) by the key components of the live system.
+ To document the resources used (e.g. processor, disk, time) by each
of the key components of the system.
«To compare the workload and resource usage with that predicted by
the performance models [Model].
@+To highlight any discrepancies, and any action taken or
recommended.
© 2000 ICL Pathway Lit COMPANY IN CONFIDENCE Poge S of 56
Filename: SR Jan Feb 2000{32)dev Printed by AIH at 30/06/00 33:43
Nothing contsined hardin shal be detmed or camsired ffcting enlsing contrac sbligatony or resting newrconrachval
cbliations between ICL Patharay and POCL
CS/PER/O
Pathway Performance Re as
Pathway Live Syste Report Dann “Fo
29/03/00
1.2
Overview
The live system report is divided into sections covering the following
services:
FUJ00079350
FUJ00079350
+ OBCS (see §2) + —[Formatted: Bullets and Numbering ]
2+ APS (see §3)
+ TPS (see §4)
Reference Data (see §5)
Riposte (see §6)
ISDN Network (see §7)
+ Issues (see §8)
In each section the key metrics are presented in graphical and/or tabular
form.
In each of the graphs the monitoring data is normally presented for
each individual day in the sample period. The predicted data is the
maximum value expected during a particular week. Where appropriate
the actual volume of transactions or resources used is compared against
the predicted value and any differences are identified and reviewed.
The current prediction models have been calibrated using data gathered
during Technical Testing for Pathway Release CSR (NR2). As the
workload volumes on the live system increase, the models will be re-
calibrated using data from the live system. The point in the rollout at
which the models will be re-calibrated will depend on the particular
component and the usage of that component.
© 2000 ICL Pathway Lit COMPANY IN CONFIDENCE Poge 6 of 56
Filename: SR Jan Feb 2000{32)dev Printed by AIH at 30/06/00 33:43
Nothing contsined hardin shal be detmed or camsired ffcting enlsing contrac sbligatony or resting newrconrachval
cbliations between ICL Patharay and POCL
Re CS/PER/O
ICL Pathway Performance en,
Pathway Live Syste Report Dates ; a ps00
1.3 Management Summary
This report covers the months of January and February 2000. Rollout of
Post Office Outlets recommenced on 24"" January and since then the
number of outlets has increased from 1860 to 3300.
The load on the hardware platforms has increased steadily and has
reached a level when the resources used on key platforms like the
Correspondence Servers can be monitored and the measurements used
to extrapolate future resource usage.
This issue of the Live System Report looks at both the workload being
processed and the resources being used.
The following summarises the current performance:
+ TPS — The EPOSS workload is growing in-line with predictions.
+ OBCE — The OBCS workload is growing more slowly than predicted
and the current projection is that the number of payments by book in
a full year with all outlets rolled out would be approx. 630M ie.
approx. 20% lower than the contractual volumes. The OBCS volumes
will be monitored closely and if, over the next two sampling periods
this trend continues, the capacity models for the platforms and the
network will be updated to reflect these changes.
The order book bar-code scanning ie has not been resolved so the
number of OBCS ‘foreign’ transactions being process is much lower
than predicted. Discussions are continuing between Pathway and
POCL about re-instating the barcode scanning of order books.
+ Reference Data — Procedural and application changes have been
made to the processing of reference data. The POCL generated peaks
experienced earlier in rollout have been significantly reduced. The
focus is now on eliminating the Pathway generated peaks that are
caused by e.g. reference data to support new releases.
+ Network - The issues identified with the ISDN network has
resulted in a significant improvement in the frequency and length of
calls. The focus has now changed to identifying the source of calls
and trying to reduce the number of ISDN calls.
© 2000 ICL Pathway Lit COMPANY IN CONFIDENCE Poge 7 of 56
Filename: SR Jan Feb 2000{32)dev Printed by AIH at 30/06/00 33:43
Nothing contsined hardin shal be detmed or camsired ffcting enlsing contrac sbligatony or resting newrconrachval
cbliations between ICL Patharay and POCL
FUJ00079350
FUJ00079350
Re CS/PER/O
ICL Pathway Performance en,
Pathway Live Syste Report Dates ; a ps00
14 Performance Changey
* During January the number of PRIs in the routers was reduced to
protect the routers from peak on peak overload. This PRIs will be re-
installed during March in the new ISDN Routers that are planned.
The new router configuration will support full roll out volumes.
LS Iyuey
« TPS - The time to harvest TPS is under review. The TPS bulk
harvester agents are processing data input from the Correspondence
Server at, or above, the expected level, but because the number of
messages input is much greater than expected [*] the overall time to
process is longer than expected. This is not an issue with the current
volumes.
+ Performance Data from the Live System - The production
of this report relies on data feeds from the live system. The report
relies on manual data feeds and problems obtaining the feeds in time
caused the January LSR to be delayed and merged with the February
LSR.
Data feeds from:
+ NT& Riposte meters
+ Maestro Logs
+ Energis phone bills
+ Data Warehouse extracts
have been used in the production of this report.
‘A CP and briefing paper have been produced for the purchase of
Athene from Metron Technology a long-standing ICL partner.
Athene provides the tools to collect manage and report on
performance data collected from the live estate and will be used by
OSD, SSC, CS and the performance team.
+ The TPS Bulk Harvester ready all messages but does only proves EPOSS
menagex The large number of additional messages generated by cash
accounting trial balances are read but not processed.
© 2000 ICL Pathway Lit COMPANY IN CONFIDENCE Poge 8 of 56
Filename: SR Jan Feb 2000{32)dev Printed by AIH at 30/06/00 33:43
Nothing contsined hardin shal be detmed or camsired ffcting enlsing contrac sbligatony or resting newrconrachval
cbliations between ICL Patharay and POCL
FUJ00079350
FUJ00079350
FUJ00079350
FUJ00079350
Re CS/PER/O
ICL Pathwaey Performance ‘ “a
Pathway Live Syste Report Dates ; a ps00
* Network — Two periods very long calls have been experienced on
the ISDN network. Mitigating actions have been put in place whilst
the Riposte bug is resolved.
© 2000 ICL Pathway Lit COMPANY IN CONFIDENCE Poge 9 of 56
Filname: ISR Jan Fel 2000(32).dev Print by AJH a 10/06/00 33:43
Nothing contained here shal be deemed or canst a fecting exiting contract obligations r reat nesreontrachial,
cbliations between ICL Patharay and POCL
Re CS/PER/O
ICL Pathway Performance en,
Pathway Live Syste Report Dates ; a ps00
24
OBCS
OBCS Summary
The 3 key components OBCS workload are:
@s Harvesting - the harvesting of the OBCS transactions that have be+—
done at the Post Offices (e.g. payment of benefit). These are then
processed by the host system (see §2.1).
+ Foreigny - The processing of OBCS foreign by the data centre (see
§2.2).
© Stops — The processing of OBCS stops from the benefit agency and
loading them into Riposte (see §2.3).
The volume of OBCS order book transactions (see Figure 2-2 - Number
of OBCS Messages Processed per Day) is lower than expected at this
point in the roll out. Extrapolating the current volumes to full rollout
would result in approx. 630M order book transactions being processed in
a full year. This is approx. 20% lower than the contractual volumes.
Currently only the original 300 live trial Post Offices are bar-code
scanning order book encashments. However most of the remaining Post
Offices are now registering new order books before they are issued. The
Pathway objective is to get all Post Offices registering all new order
books to minimise the learning effect when the bar-code scanning of
order books is re-instated. However, 319 of the remaining Post Offices
did not registered any books during February and a list of these Post
Offices has been provided to POCL who are investigating.
The average OBCS ‘foreign’ response time (see Figure 2-10 - OBCS
Foreign Response Times) is artificially low. The ‘foreign’ response time is
computed from the time to:
- Processing ‘foreign’ order book encashments ana
- Register new book receipts
New book receipts typically have a lower response time and as they
currently make up a higher proportion of all OBCS foreigns than would
be normal, this results in a lower than typical mean ‘foreign’ response
time. A new chart has been added to this report (see Figure 2-10) which
© 2000 ICL Pathway Lit COMPANY IN CONFIDENCE Page 40 of 56
Filename: SR Jan Feb 2000{32)dev Printed by AIH at 30/06/00 33:43
Nothing contsined hardin shal be detmed or camsired ffcting enlsing contrac sbligatony or resting newrconrachval
cbliations between ICL Patharay and POCL
FUJ00079350
FUJ00079350
‘{Formatted: Bullets and Numbering )
FUJ00079350
FUJ00079350
Re CS/PER/O
1c Pathway Performance , an
Pathway Live Syste Report Dann “Fo
24/03/00
separates out the small number of OBCS ‘foreign’ encashments from the
bulk of ‘foreigns’ that are created by the registration of new order books.
The percentage of OBCS ‘foreigns’ that fail is artificially low since new-——( Formatted: Bullets and Numbering )
book receipts must contact the data centre. The percentage of
transactions that fail (see Figure 2-9) has continued to decline as the
network problems are resolved.
2.2 OBCS Harvesting
This section looks at the harvesting and processing of OBCS messages.
OBCS messages include:
+ The payment of benefit by book (local & foreign)
+ The registering of new books upon receipt at a Post Office and
* The issue of new books at a Post Office.
Figure 2-2 shows the number of OBCS messages harvested per day. The
difference between the actual and predicted numbers should be because
of Northern Ireland (NI) Post Offices as they do not support OBCS (the
OBCS workload model assumes that all Post Offices process OBCS).
Even taking the NI effect into consideration the number of OBCS
messages is significantly lower than would be expected from the
contractual volumes:
Year Order Book
Tey
1999-2000 776M
2000-2001 I 750M
Figure 2-1 - Contractual Volume of OBCS Order Book Tramsactiony
Projecting the current volumes forward would result in the expected
number of OBCS order book transactions to be approx. 630M. The
growth in OBCS order book transactions will be carefully monitored as
the rollout progresses.
© 2000 ICL Pathway Lit COMPANY IN CONFIDENCE Poge 42 of 56
Filename: SR Jan Feb 2000{32)dev Printed by AIH at 30/06/00 33:43
Nothing contsined hardin shal be detmed or camsired ffcting enlsing contrac sbligatony or resting newrconrachval
cbliations between ICL Patharay and POCL
FUJ00079350
FUJ00079350
CS/PER/O
ICL Pathwaey Performance i Re as
ersions
Pathway Live Syste Report Dita “FO
24/03/00
ens OCS tg
vm
me
sme
I a=
om
vem
“2 2 @ @ @ 8 8 8 8 8 8 8 8 B
PRESS Re ees ses es es
as bane 6 8 28 88 8 Fas
apt ht cere ei
Figure 2-2 — Number of OBCS Messages Processed per Day
© 2000 ICL Pathwoy Lit COMPANY IN CONFIDENCE Page 12 of Se
Filename: SR Jan Feb 2000{32)dev Printed by AIH at 30/06/00 33:43
Nothing contained here shal be deemed or canst a fecting exiting contract obligations r reat nesreontrachial,
cbliations between ICL Patharay and POCL
ICL Pathway Performance
Pathway
Live Systew Report
Re:
Date:
CS/PER/O
43
40
29/03/00
The breakdown of OBCS transactions by day of the week given in Figure
23.
Week Starting I Mow I Tue I Wed I The I Fri I Sat I Suw
Mow 27-Dee- I 2% I 5% I 41%] 47% I 2% I 2% I 2%
aq
Mow 03-Jan- I 1% I 42% I 17% I 24% I 9% I 5% I 1%
00
Mow 10-Jan- I 31% I 24% I 14% I 18% I 9% I 4% I 1%
00
Mow 17-Jan- I 32% I 21% I 13% I 202 I 8% I 4% I 1%
00
Mow 24-Jan- I 31% I 21% I 14% I 202 I 8% I 4% I 1%
00
Mon 31-Jan- I 38% I 4% I 14% I 26% I 11% I On I 1%
00
Mow 07-Feb- I 30% I 21% I 14% I 202 I 9% I 5% I 1%
00
Mow 14-Feb~ I 30% I 21% I 15% I 19% I 8% I 5% I 1%
00
Mow 21-Feb- I 31% I 20% I 14% I 20% I 9% I Sh I 1%
00
Predicted 32% I 23% I 14% I 21% I 7% I 3% I oO”
Figure 2-3 - Breakdown of OBCS Tramsactiony by Day
The distribution of OBCS transactions across the week given in Figure
2
is similar to the pattern documented in the Performance Business
Volumes [BusVols]. In particular the %age of OBCS transactions
processed on the busiest day is a very close fit to the prediction.
Figure 2-4 shows the time to harvesting OBCS from the Correspondence
Servers and the time to process the messages on the host.
© 2000 ICL Pathway Lit COMPANY IN CONFIDENCE
Filename: SR Jan Feb 2000{32)dev Printed by AIH at 30/06/00 33:43
Poge 43 of 56
Nothing contsined hardin shal be detmed or camsired ffcting enlsing contrac sbligatony or resting newrconrachval
cbliations between ICL Patharay and POCL
FUJ00079350
FUJ00079350
CS/PER/O
Pathway Performance Re as
Pathway Live Syste Report Dita “FO
29/03/00
(08CS Haverting lapsed Tame
® a 8 & & @
a8 ob 8 an
Boy ROR OB
wae
poe
nae
ane
ome
woe
wisi
Fe
oe ne teste tc te cengrtr iene pe tr
Figure 2-4 - Elapsed Time to- Harvest OBCS Tramsactiony,
Figure 2-4 is based on the workload being processed at all Post Offices
and includes both order books scanned and order books processed using
the non bar-coded order book function, The lapsed time is gathered
from metrics in the OBCS Bulk Harvester agent.
The workload peaks were caused by:
+ 1/2"4 February - The introduction into the system of TPS
reconciliation
+ 29" February - The APS bulk harvester agents at Wigan were not
available so all harvesting used the 8 bulk harvester agents on the
Bootle site.
The overheads of setting up the harvester jobs, and the short execution
times, mean that if the elapsed times are used to project future run-
times, the time will tend to be a over estimate. However, as the number
of messages processed grows the impact of the overheads will be reduced
and more accurate run-time estimates can be extrapolated from
measured run-times,
© 2000 ICL Pathwoy Lit COMPANY IN CONFIDENCE Page 14 of Se
Filename: SR Jan Feb 2000{32)dev Printed by AIH at 30/06/00 33:43
Nothing contained here shal be deemed or canst a fecting exiting contract obligations r reat nesreontrachial,
cbliations between ICL Patharay and POCL
FUJ00079350
FUJ00079350
CS/PER/O
ICL Pathway Performance m Re as
jersions
Pothwoy Live Syste Report Date *°
24/03/00
Mort Prcentin lage enor 08S eget
g g gs ¢ &§ 8 6
2 é a f 8
ep ht oe natorenptcbtemcte pti
Figure 2-5 - Elapsed Time for the Host to- Procesy OBCS Messager
The Host processing times for OBCS messages are still very short and
within expected limits,
© 2000 ICL Pathway Lid COMPANY IN CONFIDENCE Page 15 of Se
Filename: SR Jan Feb 2000{32)dev Printed by AIH at 30/06/00 33:43
Noting contained herein shal be deemed or consid ffcting eising contrac obligations or creating new conirachval,
cbliations between ICL Patharay and POCL
FUJ00079350
FUJ00079350
Rep C/PERIO
ICL Pathwaey Performance ‘ “a
ersion
Pathway Live System Report pie “TA
29/03/00
2.3 OBCS ‘Foreign Transactiony
There are two key performance measures for OBCS ‘foreign’ transactions:
«The volume of ‘foreigns’ (which effects the number of phone calls)
and
+ The average response time (which effects the SLAs).
The number of OBCS ‘foreigns’ is shown in Figure 2-6 below. The graph
shows all ‘foreigns’ initiated at the counters and splits them into:
+ ‘Foreigns’ those that failed ['] ie. did not receive a response within
the 20 second time-out time because:
— The phone call to the data centre failed or
— The response was longer than expected because e.g. the OBCS
Host system is unavailable
+ and those that succeeded.
Number of BCS Foreign Transactions
Figure 2-6 - Number of OBCS Foreign Tramsactiony per Day
* “Foreign! transactions that fail have a zero response time. These transactions
are excluded frow the average.
© 2000 ICL Pathway Lik COMPANY IN CONFIDENCE Page 16 of 56
Filename: SR Jan Feb 2000{32)dev Printed by AIH at 30/06/00 33:43
Nothing contained here shal be deemed or canst a fecting exiting contract obligations r reat nesreontrachial,
cbliations between ICL Patharay and POCL
FUJ00079350
FUJ00079350
Re CS/PER/O
ICL Pathwaey Performance ‘ “a
Pathway Live Syste Report Dates ; a ps00
The number of ‘foreigns’ being processed is far lower than expected
because the bar-code scanning of order book encashments are only
being processed the first 300 Post Offices rolled out.
‘The percentage of OBCS ‘foreigns’ is shown in Error! Reference source not
found. below:
Week Ending Actual I Predicte
a
Sun 02-Jan-00 3.5% 24%
Sun 09-Jan-00 32% 24%
Sun 16-Jan-00 3.6% 24%
Suw 23-Jan-00 3.7% 24%
Sun 30-Jan-00 3.9% 8.9%
Sun 06 -Feb-00 4.6% 11.3%
Sun 13 -Feb-00 S.S% 13.4%
Sum 20-Feb-00 I 5.8% 14.3%
Sum 27-Feb-00 I 6.4% 13.1%
Sum OS-May-00 I 6.4% 12.1%
Sun 12-May-00 I 6.7% 11.3%
Sum 14-Mar-00 I 6.9% 10.6%
Sum 26-Mar-00 I 5.9% 10.0%
Figure 2-7 - tage of OBCS Order Book Tramsactiony that generate
OBCS ‘Foreign Tramsactiony
Figure 2-7 records the total number of OBCS ‘foreign’ transactions ie.
encashments and book receipts. An increasing proportion of the rolled
out Post Offices are now registering new books but only 300 Post Offices
are scanning order books when they are presented for encashment. The
actual number of ‘foreigns’ will rise significantly when the bar-code
scanning of order-book encashments is re-introduced.
© 2000 ICL Pathway Lit COMPANY IN CONFIDENCE Poge 47 of 56
Filename: SR Jan Feb 2000{32)dev Printed by AIH at 30/06/00 33:43
Nothing contsined hardin shal be detmed or camsired ffcting enlsing contrac sbligatony or resting newrconrachval
cbliations between ICL Patharay and POCL
FUJ00079350
FUJ00079350
Re CS/PER/O
ICL Pathwaey Performance ‘ “a
ersions
Pathway Live Syste Report Dita “FO
24/03/00
centage of OBC8 Tenstctons
Figure 2-8 - OBCS Foreign Encashmenty ay a mage of OBCS
mn ,
The number of OBCS ‘foreign’ encashments are averaging approx. 1% of
the number of OBCS Transactions currently. This will rise when the bar-
code scanning of order books is re-introduced.
The percentage of OBCS foreigns that fail is shown below. The table
shows the average failure over the month and the worst day.
Month Average I Worst Day
Failure
Rate
June 17.6% 100.0%
July 22.2% 100.0%
August 16.7% 100.0%
September TBR 100.0%
October 4.5% 17.1%
November 1LSh 12.4%
December O4% 1.5%
© 2000 ICL Petnwoy Lid COMPANY IN CONFIDENCE Page 18 of 56
Filename: SR Jan Feb 2000{32)dev Printed by AIH at 30/06/00 33:43
Nothing contained here shal be deemed or canst a fecting exiting contract obligations r reat nesreontrachial,
cbliations between ICL Patharay and POCL
FUJ00079350
FUJ00079350
CS/PER/O
1c Pathwaey Performance ‘ ae
erscons
Pathway Live Syste Report Dita “FO
24/03/00
January 12% 80.8%
February Osh 4.8%
Figure 2-9- OBCS Foreign Failurey
© 2000 ICL Pothuray Lid COMPANY IN CONFIDENCE Poge 39 of 56
Filename: SR Jan Feb 2000{32)dev Printed by AIH at 30/06/00 33:43
Nothing contained here shal be deemed or canst a fecting exiting contract obligations r reat nesreontrachial,
cbliations between ICL Patharay and POCL
FUJ00079350
FUJ00079350
FUJ00079350
FUJ00079350
Re CS/PER/O
ICL Pathway Performance en,
Pathway Live Syste Report Dates ; a ps00
The percentage of OBCS ‘foreign’ failures has steadily reduced as a result
of actions put in place to rectify ISDN network problems.
However, the percentage of OBCS ‘foreigns’ that fail is currently
artificially low as the figure is made up from:
+ The number of OBCS ‘foreign’ encashment failures, plus
+ The number of OBCS ‘foreign’ book receipt failures
and the vast majority of ‘foreigns’ in the sample period are OBCS
‘foreign’ book receipts which must contact the data centre.
Note: The failure of an OBCS ‘foreign’ encashment does not mean that
the transaction is lost. All OBCS ‘foreign’ transactions are
written to the counter message store but, for a variety of
reasons, some messages fail to contact the datacentre and in
such circumstances the counter clerk will operate the fall back
procedure which may include telephoning the Help Desk.
The average response time for OBCS ‘foreigns’ is shown in Figure 2-10
below.
(08S Foreign Response Times
Figure 2-10 — OBCS Foreign Response Timey
The mean OBCS ‘foreigns’ response time for all ‘foreigns’ is obtained
from an analysis of the response times for:
= OBCS ‘foreign’ encashment, plus
* OBCS ‘foreign’ book receipt
© 2000 ICL Pathway Lit COMPANY IN CONFIDENCE Page 20 of 56
Filename: SR Jan Feb 2000{32)dev Printed by AIH at 30/06/00 33:43
Nothing contsined hardin shal be detmed or camsired ffcting enlsing contrac sbligatony or resting newrconrachval
cbliations between ICL Patharay and POCL
CS/PER/O
Pathway Performance Re as
Pathway Live Syste Report Dita “FO
29/03/00
24
that are harvested and stored in the Data Warehouse. This time is
currently artificially low because of the order book scanning problem.
Separating out the OBCS ‘foreign’ encashments from book receipts gives
a better indication of the current response times.
The response time that is currently being measured on the live system is
better than the response time measured in the OBCS ‘Foreign’
acceptance test benchmark i.e. 5.4 seconds,
Stopy Processing
Each night the Benefits Agency sends Pathway details of books that
should not be encashed. This list covers all OBCS beneficiaries
irrespective of whether they normally claim benefit from a Pathway
automated Post Office or not, This means that the number of changes to
the Stops List received by Pathway was at a maximum from day 1 and is
not expected to grow over time.
Within the OBCS Stops database on the Host system are details of all
Pathway automated Post Offices where the order book has been used.
When a stop is received by Pathway details of the stop are sent to
Riposte and replicated to each automated Post Office were that book has
been used.
The variation in the volumes of changes from the benefit agency is hard
to predict. Average volumes of approximately 40,000 stops per night are
expected with a contractual maximum in one night of 100,000. The dip
in volumes during December is within the normal limits expected.
"Number of OBCS Stops Processed by Host
i
© 2000
Filename: LSR Jan Feb 2000{32)dev Printed by AIM at 30/06/00 33:48
Nothing contained here shal be deemed or canst a fecting exiting contract obligations r reat nesreontrachial,
cbliations between ICL Patharay and POCL
FUJ00079350
FUJ00079350
FUJ00079350
FUJ00079350
Re CS/PER/O
ICL Pathwaey Performance ‘ “a
Pathway Live Syste Report Dates ; a ps00
Figure 2-14 OBCS Stopy received from the Benefity Agency
The number of OBCS Stops processed remains within the expected
limits.
© 2000 ICL Pathway Lit COMPANY IN CONFIDENCE
Filname: ISR Jan Fel 2000(32).dev Print by AJH a 10/06/00 33:43
Nothing contained here shal be deemed or canst a fecting exiting contract obligations r reat nesreontrachial,
cbliations between ICL Patharay and POCL
Poge 22 of 56
CS/PER/O
Pathway Performance Re as
Pathway Live Syste Report Dita “FO
29/03/00
(088 Stops Loaded into Riposte
See See pen OE se ome mange tantpset
Figure 2-12 - Number of OBCS Stopy Loaded into Riposte per day
The OBCS Stops model assumed that the proportion of Stops loaded into
Riposte would be proportional to the number of Post Offices rolled out.
The number of OBCS Stops being loaded into Riposte is not rising as
expected because order books are not being resistered at many Post
Offices and Stops can only be distributed to offices that know about a
particular order book.
As current order books expire, the new order books should be registered
when they are received and the number of OBCS Stops loaded into
Riposte should start to increase even before the order book scanning
service is re-started.
The time to process OBCS Stops by the host and the time to load them
into Riposte are shown in Figure 2-13 and Figure 2-14 below.
© 2000 ICL Pathwoy Lit COMPANY IN CONFIDENCE Page 23 of Se
Filename: SR Jan Feb 2000{32)dev Printed by AIH at 30/06/00 33:43
Noting contained herein shal be deemed or consid ffcting eising contrac obligations or creating new conirachval,
cbliations between ICL Patharay and POCL
FUJ00079350
FUJ00079350
CS/PER/O
ICL Pathwaey Performance i Re as
Pathway Live Syste Report Dates ee
oot Processing Flpted Time for ORCS Stops
2 @ ® @ @ & @ & 8 a 8 8
BS & &§ § € ss es a
Be &@ &R BS ERE Soe 8 8
“oe rea nena ii oe cae teeter
Figure 2-13 - Elapsed Time for the Host to Provesy the Daily Stopy
Fue
The time to process the Daily Stops file (see Figure 2-13) is less than three
minutes and does not impose a major load on the host system.
(OCS Stop nding Eby Teme
“““p 2 @ @ ®@ #8 8 &@ 8 &@ 8 B&B B B B
H SS € 8 & & & 8 8 8 8 B
& 8 &@ a & B 8 & RF BF Rao
Figure 2-14 - Elapsed Time to Load OBCS Stopy into- Riposte ow the:
Correspondence Server
© 2000 ICL Pathway Lid COMPANY IN CONFIDENCE Page 24 of Se
Filename: SR Jan Feb 2000{32)dev Printed by AIH at 30/06/00 33:43
Noting contained herein shal be deemed or consid ffcting eising contrac obligations or creating new conirachval,
cbliations between ICL Patharay and POCL
FUJ00079350
FUJ00079350
FUJ00079350
FUJ00079350
Re CS/PER/O
ICL Pathway Performance en,
Pathway Live Syste Report Dates ; a ps00
The elapsed time in Figure 2-14 is gathered from metrics in the OBCS
Stops Bulk Loader agent. The overheads of setting up the loader jobs and
the short execution times mean that the elapsed times do not give a true
picture of the elapsed time .v. number of OBCS Stops being processed.
The reduction in elapsed times in February resulted from the (partial)
resolution of a problem with LUC that had caused the time to connect to
LUC to be much longer than expected (up to 15 mins), thus increasing
the elapsed time for the job.
© 2000 ICL Pathway Lit COMPANY IN CONFIDENCE Page 25 of 56
Filename: SR Jan Feb 2000{32)dev Printed by AIH at 30/06/00 33:43
Nothing contsined hardin shal be detmed or camsired ffcting enlsing contrac sbligatony or resting newrconrachval
cbliations between ICL Patharay and POCL
Re CS/PER/O
ICL Pathway Performance m fas
ersion
Pothwoy Live System Report oa TS
24/03/00
3 APS
3.1 APS Summary
The graphs below show the number of Automated Payment transactions
processed by the system and a breakdown of transactions by day of the
week,
AS Tomson
same
zane
me
jon
~Tooo
sa00
@ @ @ @ @ @ @ @ &@ &@ & & & @
# ¢ # #@ 5&6 5 5 5 5 #8 8 8 8 @
8 2? @ &® @ 2 F *& & B ea 8
See hee apa tin me ec toemeptntnme te printittomm
Figure 3-1 - Number of APS transactiony processed per day
The number of APS transactions continues to grow in-line with previous
projections i.e, approx. 270M across 18,500 Post Offices.
Week Starting I Mow I Tue I Wed I Thu I Fri: I Sat I Sun
Mow 03 -Jan- On 38% I 19% I 20% I 15% 8% I OR
00
Mow 10-Jan- I 27% I 23% I 13% I 16% I 13% 7% I O%
00
Mow 17-Jan- I 27% I 23% I 13% I 17% I 13% 7% I O%
0°
Mow 24-Jan- I 26% I 22% I 13% I 15% I 15% 8% I O%
00
© 2000 ICL Patwwoy Lid COMPANY IN CONFIDENCE Page 26 of 56
Filename: SR Jan Feb 2000{32)dev Printed by AIH at 30/06/00 33:43
Nothing contained here shal be deemed or canst a fecting exiting contract obligations r reat nesreontrachial,
cbliations between ICL Patharay and POCL
FUJ00079350
FUJ00079350
pop CO/PERIO
ICL Pathway Performance vent 4s
ersion
Pathway Live Syste Report Dann “Fo
29/03/00
Mow 31-Jan- I 26% I 23% I 13% I 47% I 13% I 9% I on
00
Mow 07-Feb~ I 25% I 22% I 13% I 47% I 14% I 9% I On
00
Mow 14-Feb~ I 25% I 22% I 13% I 17% I 14% I 8% I oO”
00
Mow 21-Feb~ I 25% I 22% I 13% I 17% I 15% I 9% I On
00
Mow 28-Feb- I 26% I 25% I 13% I 16% I 14% I 8% I O%
00
Predicted 27% I 23% I 13% I 17% I 14% I 7% I on
Figure 3-2 - Proportion of APS Transactiony processed per day
Figure 3-2 above tabulates the distribution of APS transactions processed
per day. The distribution of automated payments was not included in
the Workload Brief so it was assumed that it followed a similar profile to
TPS. The data collected from the live system suggests that the
assumptions used were correct.
3.2 APS Harvesting
The elapsed time to harvest APS messages from the Correspondence
Servers and write them into the host APS database is shown in Figure 3-3
below.
© 2000 ICL Pathway Lit COMPANY IN CONFIDENCE
Filename: SR Jan Feb 2000{32)dev Printed by AIH at 30/06/00 33:43
Page 27 of 56
Nothing contsined hardin shal be detmed or camsired ffcting enlsing contrac sbligatony or resting newrconrachval
cbliations between ICL Patharay and POCL
FUJ00079350
FUJ00079350
CS/PER/O
Pathway Performance Re as
Pathway Live Syste Report Dita “FO
29/03/00
(APS Harventing Flap Time
Figure 3-3 ~ Time to- Harvest APS Messages from the Correspondence
Server
The elapsed time in Figure 3-3 is gathered from metrics the APS Bulk
Harvester agent.
The reduction in elapsed times in February resulted from the (partial)
resolution of a problem with LUC that had caused the time to connect to
LUC to be much longer than expected (up to 15 mins), thus increasing,
the elapsed time for the job.
On 29" February the APS bulk harvester agents at Wigan were not
available so all harvesting used the 8 bulk harvester agents on the Bootle
site.
© 2000 ICL Pathwoy Lit COMPANY IN CONFIDENCE Page 28 of Se
Filename: SR Jan Feb 2000{32)dev Printed by AIH at 30/06/00 33:43
Noting contained herein shal be deemed or consid ffcting eising contrac obligations or creating new conirachval,
cbliations between ICL Patharay and POCL
FUJ00079350
FUJ00079350
FUJ00079350
FUJ00079350
CS/PER/O
ICL Pathwaey Performance i Re as
Pathway Live Syste Report Dates ee
3.3 APS Host Processing
The elapsed time to process the harvested APS messages stored in the
Host APS database is shown in the Figure 3-4 below.
(APS Host Processing Fld Time
eae & 8 8 B 8 B 8 8 8B
Fa BS 8 8 8 8 8 8 B
it a8 8 & 8 5 2 Re a B
‘muerte en ae acme be te mnt ego
Figure 3-4 - Elapsed Time to Procesy Harvested APS Messages
The elapsed times for APS Host processing are currently so short (less
than 5 minutes) that trends cannot be established until thje volumes
grow to the point were the time to processess APS transactions cannot
be distorted by job overheads, inc. Maestro job start-up times.
© 2000 ICL Pathwoy Lit COMPANY IN CONFIDENCE Page 29 of Se
Filename: SR Jan Feb 2000{32)dev Printed by AIH at 30/06/00 33:43
Nothing contained here shal be deemed or canst a fecting exiting contract obligations r reat nesreontrachial,
cbliations between ICL Patharay and POCL
FUJ00079350
FUJ00079350
Re CS/PER/O
ICL Pathwaey Performance ‘ fas
Pathway Live Syste Report Dates ee
* TPS
4.1 TPS Summary
This section looks at the transactions processed by TPS.
The number of EPOSS transactions is growing in-line with expectations.
Figure 4-1 below shows the number of Riposte messages passed to TPS
for processing.
—r
vse
me
—
=
—
2 reenooe
ina
_
a
“2 @ @ @ @ &@ & & & &8 B B8 B BB
aes = eras
Figure 4-1 - Number of Riposte Messagey Processed by TPS
Notes: The figures in Figure 4-1 above do not include messages
generated by cash accounts and stock declarations.
The number of TPS messages predicted by the TPS model (see
Figure 4-1) is for the busiest day in the week. Figure 4-1 shows
that there is a good correlation between the model and the live
system.
© 2000 ICL Pathwoy Lit COMPANY IN CONFIDENCE Page 30 of Se
Filename: SR Jan Feb 2000{32)dev Printed by AIH at 30/06/00 33:43
Nothing contained here shal be deemed or canst a fecting exiting contract obligations r reat nesreontrachial,
cbliations between ICL Patharay and POCL
Re CS/PER/O
ICL Pathway Performance vent 4s
erscons
Pathway Live System Report paw TO
24/03/00
The breakdown by day of the week is given below.
Week Starting I Mow I Tue I Wed I Thu I Fri I Sat I Suw
Mow 03 -Jan- On 36% I 20% I 23% I 14% I 7% I O%
[ole
Mow 10-Jan- I 27% I 21% I 14% I 19% I 12% I 6% I OR
elo)
Mow 17-Jan- I 27% I 21% I 14% I 19% I 12% I 6% I OR
(ole)
Mow 24-Jan- I 26% I 21% I 14% 19% I 13% Th On
oo
Mow 31-Jan- I 27% I 21% I 14% 19% I 12% on On
oo
Mow O7-Feb- I 25% I 21% I 15% I 19% I 13% I 7% I O”
oo
Mow 14 -Feb— 25% I 21% I 14% I 20% I 13% Th O%
00
Mow 21 -Feb— 26% I 21% I 14% I 19% I 13% Th On
[ole
Mon 28-Feb- I 26% I 21% I 15% I 18% I 13% I 7% I O”
[ole
Predicted 27% I 21% I 14% I 19% I 12% I 6% I O%
Figure 4-2 - Proportion of EPOSS Tramsactiony proceed per day
The distribution of TPS transactions across the week is very similar to
the pattern documented in Performance Business Volumes [BusVols]. In
particular the %age of EPOSS transactions processed on the busiest day
is a very close fit to the prediction.
4.2 TPS Harvesting
TPS harvesting is driven by the Maestro schedule. The harvesters are
started at 18:00 and terminated at approximately 20:30. During this
period the harvesters continuously scan for end-of-day markers from
© 2000 ICL Pathway Lit COMPANY IN CONFIDENCE
Filename: SR Jan Feb 2000{32)dev Printed by AIH at 30/06/00 33:43
Nothing contsined hardin shal be detmed or camsired ffcting enlsing contrac sbligatony or resting newrconrachval
cbliations between ICL Patharay and POCL
Poge 34 of 56
FUJ00079350
FUJ00079350
FUJ00079350
FUJ00079350
Re CS/PER/O
ICL Pathway Performance en,
Pathway Live Syste Report Dates ; a ps00
Post Offices and harvest the EPOSS messages when the end-of-day
marker is received.
The time that the harvesters run is not a function of the amount of data
harvested nor the number of automated Post Offices rolled out. There is
no meaningful data available at present about the performance of the
TPS Bulk Harvesters.
However, metrics from the TPS Bulk Harvester Agents indicate that the
rate at which EPOSS messages are being processed is lower than
expected from Cl2 Technical Testing.
© 2000 ICL Pathway Lit COMPANY IN CONFIDENCE Poge 32 of 56
Filename: SR Jan Feb 2000{32)dev Printed by AIH at 30/06/00 33:43
Nothing contsined hardin shal be detmed or camsired ffcting enlsing contrac sbligatony or resting newrconrachval
cbliations between ICL Patharay and POCL
CS/PER/O
Pathway Performance Re as
Pathway Live Syste Report Dita “FO
29/03/00
Analysis of metrics from the TPS Bulk Harvesters indicate that the
problem may be associated with the number of EPOSS messages
scanned but not harvested, which is much higher than expected. The
analysis so far indicates that this is due to a much higher number of:
« Riposte system messages and
« Cash accounting messages.
See §6.1 - Number of Riposte Messages for further details.
Because most cash accounting is run after end-of-day on a Wednesday
this adds a significant load to the TPS Bulk Harvesters on a Thursday
evening. This is also being monitored but changes such as the reduction
in the number of ‘user lock requests’ (CP2253) will reduce the impact on
the TPS Bulk Harvester.
Currently there is no issue with the time taken by the TPS Bulk
Harvesters but the performance is being monitored carefully and
investigations are taking place with B&TC.
TPS Host Processing
Figure 4-2 shows the elapsed time to process TPS on the host.
{PS Heat Processing Fhe Time
pent Tt
ae atc teeueg teint pie
ove
woe
aoe
aw
worm
orm
were
norm
Fem
Figure 4-3 - Elapsed time to process TPS messagey ow the Host
© 2000 ICL Pathwoy Lit COMPANY IN CONFIDENCE Page 33 of Se
Filename: SR Jan Feb 2000{32)dev Printed by AIH at 30/06/00 33:43
Nothing contained here shal be deemed or canst a fecting exiting contract obligations r reat nesreontrachial,
cbliations between ICL Patharay and POCL
FUJ00079350
FUJ00079350
FUJ00079350
FUJ00079350
Re CS/PER/O
ICL Pathwaey Performance ‘ “a
Pathway Live Syste Report Dates ; a ps00
The time to complete TPS host processing is growing in-line with the
predicted time.
© 2000 ICL Pathway Lit COMPANY IN CONFIDENCE Poge 34 of 56
Filename: SR Jan Feb 2000{32)dev Printed by AIH at 30/06/00 33:43
Nothing contained here shal be deemed or canst a fecting exiting contract obligations r reat nesreontrachial,
cbliations between ICL Patharay and POCL
Re CS/PER/O
ICL Pathway Performance en,
Pathway Live Syste Report Dates ; a ps00
St
5.2
REFERENCE DATA
Reference Data Summary
Reference data processing is split into three main components:
+ The processing of reference data changes received from POCL. This
takes place on the Host system,
+ The loading of non-core reference data changes into Riposte and
+ The loading of core reference data changes into Riposte.
The issues discussed in the December LSR concerning the high volume
of messages being processed on a small number of nights is in the
process of being resolved by changes to operational practices combined
with updates to the reference data applications.
The stress points in the reference data system are increasingly due to
operational requirements within Pathway to download new reference
data before a change is made to the system. Smoothing this operation
over a number of nights is not normally possible as new POCL generated
changes cannot be applied until the download of Pathway reference data
is complete.
During January and February the number of POCL generated changes
remained below the level predicted ['].
Reference Data Processing
Before the bulk loader agents can load reference data into Riposte the
data has to be extracted from the RDDS.
A number of changes have been applied to the extraction process (inc.
P2298) to improve performance. The situation will continue to be
monitored.
There is no available information on the amount of data being extracted.
Metrics have been added to the version of the application that will be
delivered at C14.
+ The prediction of the rate of reference data change was probably the most
difficult to- predict ay there way no information against which to- text the
prediction.
© 2000 ICL Pathway Lit COMPANY IN CONFIDENCE Page 35 of 56
Filename: SR Jan Feb 2000{32)dev Printed by AIH at 30/06/00 33:43
Nothing contsined hardin shal be detmed or camsired ffcting enlsing contrac sbligatony or resting newrconrachval
cbliations between ICL Patharay and POCL
FUJ00079350
FUJ00079350
FUJ00079350
FUJ00079350
Re CS/PER/O
ICL Pathwaey Performance ‘ “a
Pathway Live Syste Report Dates ; a ps00
Figure 5-1 shows the elapsed to extract the reference data. The spike on
3" March was due to multiple changes (undo the previous change &
apply the new change) to the price of postage stamps being applied
following the decision to alter the date of the price increase.
© 2000 ICL Pathway Lit COMPANY IN CONFIDENCE Poge 36 of 56
Filename: SR Jan Feb 2000{32)dev Printed by AIH at 30/06/00 33:43
Nothing contained here shal be deemed or canst a fecting exiting contract obligations r reat nesreontrachial,
cbliations between ICL Patharay and POCL
ICL
CS/PER/O
Pathway Performance Re as
Pathway Live Syste Report Dita “FO
29/03/00
S35
The elapsed time for the reference data extraction process is shown in
Figure 5-1 below. The elapsed time data is extracted from the Maestro
schedule log.
lapsed Time to Extract Changes from RODS
Figure 5-1 - Elapsed time to Extract Reference Data Changes
Based on the data that is available, the actions that have been put in
place by Pathway to manage the variability in the amount of reference
data processed appears to be having the desired effect. This will continue
to be monitored.
The RDDS extract load on 3"! March was due to multiple changes (undo
the previous change & apply the new change) to the price of postage
stamps being applied following the decision to alter the date of the price
increase.
Now Core Reference Data Loading
Reference data loading is done in one of two ways depending on whether
the reference data is core or non-core,
For non-core reference data changes, the reference data is loaded
directly from the host system into the Riposte message store for a
particular Post Office.
© 2000 ICL Pathwoy Lit COMPANY IN CONFIDENCE Page 37 of Se
Filename: SR Jan Feb 2000{32)dev Printed by AIH at 30/06/00 33:43
Nothing contained here shal be deemed or canst a fecting exiting contract obligations r reat nesreontrachial,
cbliations between ICL Patharay and POCL
FUJ00079350
FUJ00079350
pop CO/PERIO
ICL Pathwaey Performance ‘ “a
ersion
Pathway Live Syste Report Dita “FO
29/03/00
Mom Con Rleemce Dat Add tpn Per Day
vase
vase
fan
I ae
I ae
5
] m0
awe
@ 2@ 2 @ @ @ 8 @ @ @ @ @
Bs 8 2 8 8 B 6 8 8 8 8 8
a 2 a BB RN 8 2 8 a 8
etter rom ane
Figure 5-2 - Number of Riposte Persistent Objecty Added to Riposte
per Day
The metrics shown in Figure 5-2 were introduced into the system at
Cl2.2R2 so the number of persistent objects that are created in Riposte
can be accurately tracked.
Following the introduction of VPN late in February problems were
experienced with the Tivoli feed into the reference data system which
resulted in no reference data for new outlets being added to Riposte for a
number of days.
The problem has since been resolved.
© 2000 ICL Pathway Lik COMPANY IN CONFIDENCE Page 38 of 56
Filename: SR Jan Feb 2000{32)dev Printed by AIH at 30/06/00 33:43
Nothing contained here shal be deemed or canst a fecting exiting contract obligations r reat nesreontrachial,
cbliations between ICL Patharay and POCL
FUJ00079350
FUJ00079350
FUJ00079350
FUJ00079350
CS/PER/O
1c Pathway Performance m ae
ersions
Pathway Live Syste Report Dita “FO
24/03/00
Faget Ti ol Non Cae iene hn pte
in
“Sp es f & & 8 8 8 8 8 8&8 8 8 8
BE 8 8 8 8228 88 8 &
ae a kB BR be 2 ba
Figure S-3 - Elapsed Time to load non-core reference data into-
Riposte
In Figure 5-3 the long loading time on the 31 January corresponds to the
failure of an agent that required manual intervention by operations to
resolve.
Actions have been put in place to smooth the number of reference data
changes per night. CS has implemented a number of changes in
association with POCL that appear to be having the desired effect.
5.4 Core Reference Data Loading
For core reference data changes, the reference data is first loaded into
the message store for a ‘dummy’ Post Office on the Correspondence
Server. Once this is done the changes are then replicated from the
‘dummy’ Post Office to all the automated Post Offices on the
Correspondence Server.
In addition, when a new Post Office is automated the complete reference
data required for that Post Office is also replicated.
The number of core reference data changes loaded into the ‘dummy’
Post Office and the elapsed time to execute this process are shown
Figure 5-4.
© 2000 ICL Pathway Lit COMPANY IN CONFIDENCE Poge 39 of 56
Filename: SR Jan Feb 2000{32)dev Printed by AIH at 30/06/00 33:43
Nothing contained herein shal be deemed or eanshried a fecting exiting tontrachucbligations o ereating neareontrachial
cbliations between ICL Patharay and POCL
Re CS/PER/O
ICL Pathwaey Performance ‘ “a
Pathway Live Syste Report Dates ; a ps00
(Chemges to Cove Rederence Dat Px Pst O8ice
awe
pia
worm
wore
sii
orem
wom
rem
won
wove
Figure 5-4 - Number of changes to- Core Reference Data per Post
Office per Day
Note R_LD_ALL is the bulk loader agent that loads POCL core
reference data into the dummy outlet on the Correspondence
Server.
R_LD_COLL is the bulk loader agent that loads the Class D
reference data (ie. Pathway reference data) into the dummy
outlet on the Correspondence Server.
© 2000 ICL Pathway Lit COMPANY IN CONFIDENCE Page 40 of 56
Filname: ISR Jan Fel 2000(32).dev Print by AJH a 10/06/00 33:43
Nothing contained herein shal be deemed or eanshried a fecting exiting tontrachucbligations o ereating neareontrachial
cbliations between ICL Patharay and POCL
FUJ00079350
FUJ00079350
pop CO/PERIO
ICL Pathwaey Performance ‘ fas
ersion
Pathway Live Syste Report Dita “FO
29/03/00
Eg Timetot oad Come erence Dana Darn Once
i oom
“Wp 2 2 @ # @ @ &# &@ 8 @ 8 8 8 B
see Fe eee 8 88 8 ee
ab ak 8 8 Fk SF ee kee
Figure 5-5 - Elapsed time to load Core Reference Data into- the
‘dumany! Post Office
The number of changes to core reference data were small but the elapsed
time of the loading process is fairly consistent as this is the default time
to run the process even when no changes are required. At Cl4 changes
will be made to the reference data system to stop the jobs running if no
changes are to be made.
Figure 5-6 shows the number of Riposte messages replicated from the
dummy Post Office and the elapsed time to execute the replication
process.
© 2000 ICL Pathway Lik COMPANY IN CONFIDENCE Page 44. of 56
Filename: SR Jan Feb 2000{32)dev Printed by AIH at 30/06/00 33:43
Nothing contained here shal be deemed or canst a fecting exiting contract obligations r reat nesreontrachial,
cbliations between ICL Patharay and POCL
FUJ00079350
FUJ00079350
CS/PER/O
Pathway Performance Re as
Pathway Live Syste Report Dita “FO
29/03/00
(Come Reteronce Date Ade to Ripon Pex Day
_TEPEET LL
wove
vei
a
wou
ee
ce
Figure 5-6 - Number of Core Reference Data changes per day
"apoed Te to Papi Cone Felon Deon
See tate pert ee pt ete opens en et
Figure 5-7 - Elapsed, time to- Replicate Core Reference Data per day
The replication time is dominated by the process that replicates
reference data to outlets just prior to them being rolled out and this
process remains well within the allocated slot in the Maestro schedule.
© 2000 ICL Pathway Lit COMPANY IN CONFIDENCE Poge 42 of 56
Filename: SR Jan Feb 2000{32)dev Printed by AIH at 30/06/00 33:43
Nothing contained here shal be deemed or canst a fecting exiting contract obligations r reat nesreontrachial,
cbliations between ICL Patharay and POCL
FUJ00079350
FUJ00079350
FUJ00079350
FUJ00079350
Re CS/PER/O
ICL Pathwaey Performance ‘ “a
Pathway Live Syste Report Dates ; a ps00
The major loads seen in Figure 5-6 and Figure 5-7 result from reference
data being downloaded by Pathway in support of a system change.
Methods for smoothing this load are under investigation.
© 2000 ICL Pathway Lit COMPANY IN CONFIDENCE Poge 43 of 56
Filename: SR Jan Feb 2000{32)dev Printed by AIH at 30/06/00 33:43
Nothing contained here shal be deemed or canst a fecting exiting contract obligations r reat nesreontrachial,
cbliations between ICL Patharay and POCL
FUJ00079350
FUJ00079350
Re CS/PER/O
ICL Pathwaey Performance ‘ fas
Pathway Live Syste Report Dates ee
6 RIPOSTE
This section looks at Riposte related performance including:
« The number of Riposte messages generated
+ The size of the messages
+ The time taken to run Riposte Archiving.
6.1 Number of Riposte Messager
—e ne
emo
=I
ae
a
—
ae
“2 2 @ @ ® 8 B 8 BB B
PEER EEE a8 8 & e
eGR EEE epee e 3
arin ncn
Figure 6-1 - Number of Riposte messages generated per day
The growth in the number of messages generated per day is following
the prediction but the model is currently under predicting the maximum
number of Riposte messages generated per day. The biggest
discrepancies identified are in:
+ The large volumes of reference data changes (see Sections)
+ The number of messages generated by cash accounting
+ The number of Riposte system messages.
The prediction in Figure 6-1 represents the number of Riposte messages
generated on the busiest day in each week, If the number of Riposte
© 2000 ICL Pathwoy Lit COMPANY IN CONFIDENCE Page 44 of Se
Filename: SR Jan Feb 2000{32)dev Printed by AIH at 30/06/00 33:43
Nothing contained here shal be deemed or canst a fecting exiting contract obligations r reat nesreontrachial,
cbliations between ICL Patharay and POCL
FUJ00079350
FUJ00079350
Re CS/PER/O
ICL Pathwaey Performance ‘ “a
Pathway Live Syste Report Dates ; a ps00
messages generated for the month of March follows a similar pattern
then the models will be updated to reflect the actual measurements from
the live system,
6.1.1 Cash Accounting
The cash accounting process was one of the most difficult to estimate as
the volume of data generated is based on:
* The number of stock units in a Post Office and
* The number of trial balances generated by the postmaster.
Based on CSR(NR2) Live Trial data, the expected number of cash
accounting messages was approximately 1300 per Post Office per week.
An analysis of the Riposte message store data indicates the number is
currently nearer 1900 per Post Office.
As well as creating more messages per Post Office, postmasters are
taking much longer than was expected to complete cash accounting on a
Wednesday evening. The messages generated by cash accounting after
approx. 18:00 on a Wednesday evening are not processed ['] by the TPS
Bulk Harvester Agent until Thursday evening adding significantly to the
time to harvest TPS messages on a Thursday.
“Total Persistent Objects
7
1 The menager-generated by the cas accounting provesave replicated. from:
the Post Office counter to Correspondence Server message Hore but most of
these messager are not harvested by the TPS Harvester. The additional load
ow the TPS Harvester is caused by the harvester having to sean all of tee
menages (ince cash accounting messages) generated that-day to find tre
mesages that + has to harvest:
© 2000 ICL Petnwoy Lid COMPANY IN CONFIDENCE Page 45 of 56
Filename: SR Jan Feb 2000{32)dev Printed by AIH at 30/06/00 33:43
Nothing contained here shal be deemed or canst a fecting exiting contract obligations r reat nesreontrachial,
cbliations between ICL Patharay and POCL
Re CS/PER/O
ICL Pathway Performance en,
Pathway Live Syste Report Dates ; a ps00
Figure 6-2 - Load over the Day due to- Cash Accounting
The red line (Wed 26/o1/o0) in Figure 6-2 shows the number of
Persistent Object created in Riposte. The difference between the number
of Persistent Objects created on a Wednesday afternoon/evening
provides a good indication of the load created by cash accounting and
when when cash accounting/stock roll-over is performed on the Post
Offices i.e.
* 10:00 ~ 16:00 for outlets that close Wednesday lunchtime
* 1730 21:30 for outlets that close at 17:30 on Wednesday
* Cash accounting also adds to the load on Thursdays up to lunchtime.
6.1.2 Riposte System Messages
The number of messages generated by Riposte functions e,
+ Log on/log off
+ End of day reports
+ Session transfers
+ ete.
is significantly greater than the prediction which was based on the
CSR(NR2) Live Trial system. The prediction was that 200 messages per
counter per day would be generated. Data from the live system indicates
that the number currently exceeds 500 per counter per day.
These messages are a function of the way the counter terminals are used
and not a function of the number of OBCS, EPOSS, APS, etc transactions
processed by the counter.
6.1.3 User Lock Requesty
CPs to remove unnecessary messages are being raised starting with
CP2253 which significantly reduces the number of User_Lock_Requests
generated by the counter. This will both reduce the number of messgaes
in the message store and significantly reduce the load on the Persistant
Object Index in the Correspondence Server during the peak hour.
CP2253 will be introduced at Cl3.2R2 i.e. before Cl4.
© 2000 ICL Pathway Lit COMPANY IN CONFIDENCE Poge 46 of 56
Filename: SR Jan Feb 2000{32)dev Printed by AIH at 30/06/00 33:43
Nothing contsined hardin shal be detmed or camsired ffcting enlsing contrac sbligatony or resting newrconrachval
cbliations between ICL Patharay and POCL
FUJ00079350
FUJ00079350
Re CS/PER/O
ICL Pathwaey Performance ‘ fas
Pathway Live Syste Report Dates ee
6.2 Size of Riposte Data
Until Riposte archiving is enabled [']the amount of data being generated
in the message store of the Correspondence Servers has been carefully
monitored.
The initial sizing of the message store was very conservative, as it was
difficult to predict the variability between the amount of data generated
by outlets of the same size (number of counters). A significant amount of
data about the content of the message store has been collected which
has allowed the target date for the switch on of Riposte Archiving to be
relaxed to March 2000 when the Riposte Integrity Checker will become
available.
“Toh hn ot Ripont Date Gmrsnt Py
BE 88 8 ¢ 8 8
eh ea & ER EB
Figure 6-3 - Size (Mbytes) of Riposte Data Generated per Day
The peak on the 30" January was due to the introduction of TPS
Reconciliation into the system.
CP2253 to remove unnecessary messages have been approved starting
with which significantly reduces the number of User_Lock_Requests
generated by the counter. This will both reduce the number of messgaes
in the message store and significantly reduce the load on the Persistant
Object Index in the Correspondence Server during the peak hour.
+ Riposte Archiving deletes mesages that have expired from the message stores
© 2000 ICL Pathway Lit COMPANY IN CONFIDENCE Poge 47 of 56
Filename: SR Jan Feb 2000{32)dev Printed by AIH at 30/06/00 33:43
Nothing contained here shal be deemed or canst a fecting exiting contract obligations r reat nesreontrachial,
cbliations between ICL Patharay and POCL
FUJ00079350
FUJ00079350
FUJ00079350
FUJ00079350
Re CS/PER/O
ICL Pathway Performance en,
Pathway Live Syste Report Dates ; a ps00
6.3 Archiving
This section looks at Riposte archiving.
Riposte archiving is currently switched off at the data centres. This
section is therefore not reported on.
Riposte Archiving was originally switched off because of operational
problems that resulted in the archiving process interacting with other
processes running on the Correspondence Servers resulting in Riposte
failures.
These problems have been resolved but Archiving will remain switched
off until the SMC Support Server is introduced into service and the
Riposte Integrity Checker has been run on all Correspondence Servers to
ensure that there are no inconsistencies in the message store.
© 2000 ICL Pathway Lit COMPANY IN CONFIDENCE Poge 48 of 56
Filename: SR Jan Feb 2000{32)dev Printed by AIH at 30/06/00 33:43
Nothing contsined hardin shal be detmed or camsired ffcting enlsing contrac sbligatony or resting newrconrachval
cbliations between ICL Patharay and POCL
FUJ00079350
FUJ00079350
CS/PER/O
ICL Pathwaey Performance ‘ Re as
Pathway Live Syste Report Dita “FO
24/03/00
7 ISDN NETWORK
This section looks at the performance of the ISDN network.
This is split into three sections:
+ The number of calls per day. + (Formatted: Bullets and Numbering )
+ The call duration per day.
+ The peak usage of the routers (calls per second and ISDN lines in
use).
The phone calls are divided into 4 types:
@ Core - Made during the core part of the day (7am to 8pm). + Formatted: Bullets and Numbering )
2+ Non Core - Made outside the core part of the day.
+ In — Calls into the data centre from the Post Offices.
= _Out - Calls out to the Post offices from the data centre.
{SDN Calls Per Day Total forall Outlets
£ sviovc}
Figure 7-1 — Total ISDN Cally per Day for all Outtety
© 2000 ICL Pathwoy Lit COMPANY IN CONFIDENCE Page 49 of Se
Filename: SR Jan Feb 2000{32)dev Printed by AIH at 30/06/00 33:43
Nothing contained here shal be deemed or canst a fecting exiting contract obligations r reat nesreontrachial,
cbliations between ICL Patharay and POCL
FUJ00079350
FUJ00079350
Re CS/PER/O
ICL Pathwaey Performance ‘ “a
Pathway Live Syste Report Dates ; a ps00
Note: No data for outbound calls (calls from the datacentre) is currently
available for the period 19" January to the end of February.
© 2000 ICL Pathway Lit COMPANY IN CONFIDENCE Poge SO of 56
Filename: SR Jan Feb 2000{32)dev Printed by AIH at 30/06/00 33:43
Nothing contained here shal be deemed or canst a fecting exiting contract obligations r reat nesreontrachial,
cbliations between ICL Patharay and POCL
FUJ00079350
FUJ00079350
Re CS/PER/O
ICL Pathwaey Performance ‘ fas
Pathway Live Syste Report Dates ee
ISON calls
Figure 7-2 - Average Number of ISDN Cally per Day per Outlet
Note: No data for outbound calls (calls from the datacentre) is currently available for
the period 19" January to the end of February.
The number of ISDN calls per outlet per day has been reduced by the
elimination of most of the network problems that resulted in multiple
calls being made.
The average number of calls per day is currently higher than expected
even though OBCS ‘foreigns’ are not being processed at most offices.
During the normal working day - 09:00-17:30 - 35 calls would be made
by Riposte synchronising every 15 minutes. Figure 7-2 confirms that this
is approx. the number of calls being made (Core-In).
The default broadcast interval was increased from 15 to 20 minutes at the
end of February and this has made caused a further small reduction in
the number of calls.
The challenge is to reduce the number of calls made for other reasons
e.g, there should be a negligible number of calls out during the core day.
© 2000 ICL Pathway Lid COMPANY IN CONFIDENCE Page St of Se
Fulenames LSR Jan Feb2000{32)dee Print by ALM 10/06/00 34:43
Nothing conduined erin shal be deemed tans a fcing esting tontrachualsbligans or etating new contachval
cbistions betwee Pathway and POCL
Re CS/PER/O
ICL Pathwaey Performance ‘ fas
Pathway Live Syste Report Dates ee
'SDN Total Call Duration per Day - Average per Outlet
Figure 7-3 - Total Time of Telephone Cally per oudlet per day
Note: No data for outbound calls (calls from the datacentre) is currently available for
the period 19" January to the end of February.
The two periods of very long calls have been traced to a bug in Riposte.
The problem occurs when the NT clock wraps after approx. 49 days. Ifan
ISDN call is open at the time the clock wraps the call is not torn down as
expected and the line can remain open for many hours.
The cost to Pathway of ISDN calls is a function of the number and length
of calls. Prior to Christmas, a number of issues were identified:
+ CP 2135 - ISDN Timeout value
« Eicon network driver in the gateway counter PC
which were causing the ISDN connection to be regularly kept open
Jonger that was necessary to transfer data, Many of these problems have
now been resolved and the time the telephone line is in use per outlet
per day has dropped significantly.
The usage of phone lines is under investigation and further changes are
anticipated to reduce the duration of calls further.
© 2000 ICL Pathway Lit COMPANY IN CONFIDENCE Poge 52 of 56
Filename: SR Jan Feb 2000{32)dev Printed by AIH at 30/06/00 33:43
Nothing contained here shal be deemed or canst a fecting exiting contract obligations r reat nesreontrachial,
cbliations between ICL Patharay and POCL
FUJ00079350
FUJ00079350
FUJ00079350
FUJ00079350
Re CS/PER/O
1c Pathwaey Performance ‘ “a
Pathway Live Syste Report Dita “FO
24/03/00
© 2000 ICL Pathway Ltd COMPANY IN CONFIDENCE Page S3 of Se
Filename: SR Jan Feb 2000{32)dev Printed by AIH at 30/06/00 33:43
Nothing contained here shal be deemed or canst a fecting exiting contract obligations r reat nesreontrachial,
cbliations between ICL Patharay and POCL
Re CS/PER/O
ICL Pathwaey Performance ‘ fas
Pathway Live Syste Report Dates ee
Peak Callsis during Core Part of Day (7am te 8pm)
Figure 7-4 - Phone Cally per Second acroyy the ISDN Network
Note: No data for outbound calls (calls from the datacentre) is currently available for
the period 19" January to the end of February.
‘The peak load is the number of calls during the peak second during the core
day. On several days (e.g. 12/13 January) there was more traffic out in the peak
second which is not the normal model for a weekday.
The number of calls in during the core day is currently lower than
expected because calls associated with OBCS ‘foreigns’ are not taking
place at present.
© 2000 ICL Pathwoy Lit COMPANY IN CONFIDENCE Page 54 of Se
Filename: SR Jan Feb 2000{32)dev Printed by AIH at 30/06/00 33:43
Nothing contained here shal be deemed or canst a fecting exiting contract obligations r reat nesreontrachial,
cbliations between ICL Patharay and POCL
FUJ00079350
FUJ00079350
Re CS/PER/O
ICL Pathwaey Performance ‘ fas
erscons
Pathway Live Syste Report Dita “FO
24/03/00
PeakI9DN Lines nur ing or Part of ay am to pm)
i =
i
Figure 7-5 - Peak Number of ISDN Liney in Use during the Core Day
(07:00-20:00)
Note: No data for outbound calls (calls from the datacentre) is currently available for
the period 19" January to the end of February.
The peak number of ISDN lines in use has reduced considerably as
actions have been put in place to resolve problems. The number of lines
is still higher than predicted by the network model but is well within the
capability of the system. Currently 720 lines are available and when the
ISDN Routers are upgraded in March this will rise to 1440.
However, the usage should be lower and the usage of the ISDN network
including:
+ The source of calls
+ Eliminating unnecessary calls
+ Reducing the elapsed time of calls
+ Specific scheduling of overnight calls
are under review to determine the best way of reducing the total cost of
calls,
© 2000 ICL Patwoy Lid COMPANY IN CONFIDENCE Page SS of Sé
Filename: SR Jan Feb 2000{32)dev Printed by AIH at 30/06/00 33:43
Nothing contained here shal be deemed or canst a fecting exiting contract obligations r reat nesreontrachial,
cbliations between ICL Patharay and POCL
FUJ00079350
FUJ00079350
FUJ00079350
FUJ00079350
Re CS/PER/O
1c Pathway Performance ” Fas
ersions
Pathway Live Syste Report Dann “Fo
24/03/00
8 ISSUES
This section briefly describes a number of areas of concern that are being
investigated.
Note: An entry on this list does not necessarily mean it is a performance
problem that will require changes to fix it. Changes will only be
required if there is insufficient spare capacity in the system to
absorb the problem,
8.1 Under Investigation
The following are under investigation:
+ The number and total size of Riposte messages are much higher than+——{ Formatted: Bullets and Numbering )
expected. On a typically day 30% of messages generated each day are
not processed by TPS. This can rise to 60% on Wednesday and
Thursday. The majority of this data is generated by counter
applications e.g. trial cash account balances, log on/off, etc. This is
impacting the elapsed time of the TPS Bulk Harvester.
* Reference data changes in support of system changes. These can
create a very heavy load on one night. Methods to spread the load are
under investigation.
8.2 Solutions Identified
A number of problems have been identified and are being resolved.
These will be fully reported on in the March LSR.
* The cause of the long ISDN network connect times has been
identified. The change to Riposte is in production.
8.3 Closed
+ A number of changes in the area of the ISDN network have been
applied and these have significantly reduced the incidences of long
calls and repeated short calls.
© 2000 ICL Pathway Lit COMPANY IN CONFIDENCE Poge 56 of 56
Filename: SR Jan Feb 2000{32)dev Printed by AIH at 30/06/00 33:43
Nothing contsined hardin shal be detmed or camsired ffcting enlsing contrac sbligatony or resting newrconrachval
cbliations between ICL Patharay and POCL