Teamcenter Manufacturing Performance Studies
Last update February 2016
Preliminary Notes
If not stated otherwise, all the 10.1 based tests were executed on a 64 bit installation.
Latest updates are based on 10.1.x released versions, however, in some cases where no significant changes were introduced the reported results are for the latest available version (10.1/9.1.x).
We strive to align all studies on the same released version however, the study results are being updated based on specific areas with available results , therefore, some results will be focused on the latest version service pack released where others will still be on previous service pack, none the less the base major version for all areas will be aligned.
The following presentation is a work in progress. Further updates will be provided as the study advances through the Siemens PLM releases and additional improvements are introduced.
In line with the above, in general, this study covers the basic areas of the Teamcenter Manufacturing use cases in respect to performance.
Over time, it is expected to extend the coverage and address the main areas of interest to the manufacturing use cases.
The data presented in the following slides is reflective of the test environment only, projection to other environments can be suggested.
However, at best, it will only reflect the software behavior and not the correlation to any other environment settings.
Recent Updates Highlights
Main manufacturing use cases p.6
Updating key use cases
Teamcenter manufacturing baseline studies p.17-42
Updating summary numbers on some use cases with TC 10.1.5
BOM Management section p.22-24
Aligning numbers with TC 10.1.5
BOP Authoring section p.25-32
Aligning numbers with TC 10.1.5
BOP Validation section p.33-37
Aligning numbers with TC 10.1.5
Documentation section p.38-41
Aligning numbers with TC10.1.5
Introduction
Status Highlights
Baselines Studies
Summary
Table of Contents
Introduction
Main Manufacturing Use Cases of Focus
Manufacturing Process Planning
(MPP)
BOM Management
• EBOM-MBOM ACC
• Configuration
• Structure search
• Occurrence Groups
BOP Authoring
• MBOM-BOP ACC
• Structure Search
• Assignments
• Assembly Viewer
• Pert, Gantt, Time, IPA
• Line balancing
• Enterprise BOP
Validation
• Process Simulate
• Assembly Viewer Documentation
• Product Views
• 3DPDF
• EWI
• Standard Text
• MFG reports Downstream
• PLMXML
• TCXML
• MES Integration
• EWI
Develop Key Performance Indicators (KPI) focus
The “heavy hitters” notion
More integration performance focus (ERP,
MES …) More integration performance focus (ERP,
MES …)
General key performance behavior focus
• TC-DB Trips
• SQL executions & distribution
• TC-DB Trips
• SQL executions & distribution RAC behavior
efficiency RAC behavior
efficiency
• Client-Server Trips
• API calls count & distribution
• Client-Server Trips
• API calls count & distribution
Backbone behavior efficiency Backbone behavior
efficiency
Main Performance Categories of Focus in our landscape
Load / Scalability Stressing the system Sensitivity
Impact of configurations
Baseline
Use case driven performance profiling
Usability studies Application usability derived from use-cases and workflows Feature Performance Develop for
performance acceptance
HW/SW sizing (configuration)
Main focus on baselines
Baseline methodology principals
Heavy Hitters – focus on what matters to our customers essentially
Sound performance foundation with no regressions to our customers
Methodical scientific approach – use cases driven, clear, measurable, reflective data
KPI baseline building blocks
Standard environment
Prepare analysis tools
From day in a life to Heavy Hitters
Performance script development
Baseline execution cycles
Baseline analysis and report Baseline
established
Update to new version and then run comparison again
Baseline Execution Cycles Flow
1. Calibration Cycles monitoring off 1. Calibration Cycles
monitoring off Stable wait time?
Stable wait time?
No
2. Response & Calls profile cycles
2. Response & Calls profile cycles
Enough results, at least 2?
Enough results, at least 2?
Yes
No
3. Footprint cycle DB load / Resource
performance 3. Footprint cycle DB load / Resource
performance
Yes
4. Detailed Logs cycle / Details parameters on 4. Detailed Logs cycle / Details
parameters on
All Results Saved in the Analysis Folder
All Results Saved in the Analysis Folder
Start
Start Environment Tuning
Cycles Environment Tuning
Cycles
Environment Tuned? Environment
Tuned?
Yes No
Sample execution efforts for phase 2
• Average cycle per workflow – takes ~1-2 hours
• 3 cycles for per workflow
• Per tested version – Total of 3 (cycles) x 2(hrs) = 6 hrs
• Additional 5-10 hrs spent on workflow test variations
`
Web Server Application Servers
DB Server Oracle 11G /
SQLServer File server:
Tecnomatix Sysroot or TC volume
` ` ` ` `
ESX 4i 8x2.39GHz Xeon
144GB RAM Backbone virtual server
Win2008R2 enterprise 64bit 2x2.39GHz 8GB RAM
Customer / customer like data: backups
Ethernet 100/1000 Mbit
WAN Simulator Windows 7 64bit
2x2.6 GHz Xeon 4GB-32GB RAM
DM Lab with standard engineering workstations
R&D client For analysis
Standard Study Environment
Use Case # lines: Product # lines: Process #JTs / Size (GB)
BOM Management 15,000 - -
BOP Authoring 30,000 16,000 5,000 / 3.5
Validation 28,000 26,000 5700 / 3
Documentation 15,000 10,000 5,000 / 3
Study Data Main Highlights
Status Highlights
Teamcenter 10.1.x Areas of Improvements
Stability
• Continuous reduction of crashes on exit/unload scenarios
• Continuous reduction of invalid tag issues
• Improved status bar synchronization on MPP bar
Usability
• Line Balancing support
• More streamlined work instructions leveraging Active Workspace client
• More templates to simplify user’s work and reduce “clicks”
Performance
• End Item Light Weight BOM and Dynamic IPA for more efficiency
• Continuous improvement on the basics with reduction of database trips
• Refinement of new technologies such as Enterprise BOP
Use Cases Overall Performance Improvements
TC 9.1 => 101.5
Notes:
Downstream execution systems was only looked at from PLMXML export perspective as the output of the BOP Authoring
Wherever relevant the data is based on cached FCC information
Using LAN except when mentioned otherwise
Main focus: a single user software performance profiling & baselines when comparing between versions 47
49 31
82
0 10 20 30 40 50 60 70 80 90
BOM Management BOP Authoring Validation Documentation
9.1 => 10.1.5 % Performance Improvement
Specific Heavy Hitters Performance Improvements
TC 9.1 => 10.1.5 Comparison
Notes:
Heavy hitters are specific actions heavily used in the context of a use case and may present performance challenges 29
19
29
72 59
0 10 20 30 40 50 60 70 80
Load to embedded viewer Load to Process Simulate Structure Expand BOM-BOP Accountability Configured Process Export using PLMXML
9.1 => 10.1.5 % Performance Improvement
Continuous Improvement Trend
BOP Authoring Use Case – Response Time Example
29.09
49.25
8.3.2 9.1 10.1.5
%
Teamcenter Version
BOP Authoring %
Performance Improvement
0 200 400 600 800 1000 1200
8.3.2 9.1 10.1.5
Seconds
Teamcenter Version
Heavy Hitters Performance Progress
13-LoadFullProd 16-LoadSubProc 27-AccBOM_BOP
Continuous Improvement Trend
Validation Use Case – Response Time Example
32.22 34
8.3.3 9.1 10.1.5
%
Teamcenter Version
Validation % Performance Improvement
0 100 200 300 400 500 600 700 800 900 1000
8.3.3 9.1 10.1.5
Seconds
Heavy Hitters Performance Progress
MPP Un-config. station display IPA MPP config. station display FIPA PS config. station load IPA
PS config. station load FIPA & PreGen BOM
Baseline Studies Highlights
The baseline studies are continuously being updated. The material in the next slides refers to the information available up to February 2016.
BOM Management
Comparison Results Overview
Heavy Hitter item # BOMLines 9.1
(sec)
10.1.5 (sec)
% Improvement 9.1 => 10.1.5
Send CC to MPP 7.2 6 17
MBOM Expand Below 5.7K 31.8 28.8 9
MBOM Structure Search 10K 115.2 25.2 78
EBOM MBOM Assign 2.6K 95.4 28.8 70
MBOM Remove Assignments 2.6K 105 73.8 30
EBOM-MBOM Accountability Check 5.6K 54 52.8 2
Total 408 216 47%
Improvements on Assign, Search and Accountability
Load CC to MPP regression fixed
Average improvement is 47%
Wait Time vs. Calls Count
BOM management TC 9.1 vs. TC 10.1.1
In general, improvements on every measured task mostly attributed to server
side improvements
Clients server trips regression in MBOM expand, Acc and send CC
currently under investigation
BOP Authoring
8.3.x uses closure rules … so, pre-expand not needed 8.3.x uses closure rules … so, pre-expand not needed
Use Case Heavy Hitters
Send CC Send CC Login
Login Product root expand
one level Product root expand
one level
Product root Load to Viewer
Product root Load to Viewer
Sub Process expand one level
Sub Process expand one level
Sub Process Load to Viewer
Sub Process Load to Viewer Undo Incremental
Change Undo Incremental
Change
Accountability Check BOP to MBOM Accountability Check
BOP to MBOM
Users Activities
Admin work
PLMXML Export
Nightly batch program
Heavy Hitters
Expand Below Product
Expand Below Product Unpack Below ProductUnpack Below Product Expand Below ProcessExpand Below Process
Unpack Below Process Unpack Below Process
Comparison Results Overview
Heavy Hitter item # BOMLines 8.3.2
(sec)
9.1 (sec)
10.1.5 (sec)
% Improvement 9.1 => 10.1.5
SendCCtoMPP 7.72 7.44 5.11 31.32
ExpandProdOneLevel 129 4.98 4.99 2.27 54.60
ViewerLoadFullProduct 30000 687.20 437.58 271.63 37.92
ExpandProcessOneLevel 52 8.25 11.68 8.18 29.96
ViewerLoadSubProcess 1100 127.04 35.11 31.27 10.93
ExpandFullProduct 15000 345.83 172.08 99.21 42.35
UnpackFullProduct 30000 388.10 244.39 67.05 72.56
ExpandFullProcess 11000 188.25 187.77 227.64 -21.23
UnpackFullProcess 16000 161.86 75.27 88.28 -17.29
AccCheckBOM_BOP 30000/16000 959.59 864.92 235.33 72.79
Total 2878.82 2041.24 1036 49.25
Significant improvement in response time, specifically with accountability check Load to viewer
Incremental reduction in client server trips except for Accountability , this is also specific to some properties with follow up analysis underway
Some regression in BOP expand continuous investigation is underway
Database trips were only collected for 10.1.5 version but within the boundaries of under 100000 per action
Keep striving to reduce client server calls to 100 per action
Wait Time vs. Calls Count
TC 9.1 vs TC 10.1.5
DB trips only measured for 10.1.5
Latency Client Server Calls Sensitivity Projection Process Plan Authoring TC10.1.5 cached JTs
Factor calculated by: [(time in <1ms) + (# API calls)*(Latency in seconds)]/(Original time in sec measured in LAN)
If we consider also JTs download impact: [(time in <1ms) + (# API calls+#JTs)*(Latency in seconds)]/(Original time in sec measured in LAN)
Excluding files download time impact only calls impact , so in case of Load Graphics JTs download also play a role
By average 1.9Xslower in latency of 300ms
Most sensitive areas: Load CC , Accountability and unpack
Least sensitive areas: Structure expand
0 1 2 3 4 5 6 7
X MULTIPLE
Latency Factor
100ms 200ms 300ms
• If we factor the amount of JTs (~5000) downloaded (not cached) multiple would be:~4.5X
• If we factor the amount of JTs (~5000) downloaded (not cached) multiple would be:~4.5X
Peak Memory Footprint
New session
TcServer similar footprint between 9.1 and 10.1 base , 64 bit consumes up to 45% more memory
RAC 10.1 base slightly better footprint compared with 9.1 with improvements on unpack
Most intense memory consumption attributed to large assembly displaying, accountability between root structures and large process expand
Memory footprint did not exceed 1.5 GB for the entire sessions both on client and server
New session
New session New session
Database Load Profile
BOP Authoring – end to end
Load Profile Per Second
Category 8.3.3 9.1 10.1.5
Redo size 3,116.67 10483 20,050.91
Logical reads 12114 41924 12,798.64
Block changes 27.46 133.11 212.02
Physical reads 0.2 0.02 0.02
Physical writes 0.65 1.17 2.46
User calls 189.78 292.01 135.7
Parses 189.68 285.27 125.73
Hard parses 0.29 0.97 0.03
Sorts 1.58 21.62 6.21
Logons 0.02 0.03 0.03
Executes 190.5 287.3 126.59
Transactions 0.67 6.66 2.16
Top 5 Timed Events Time(s)
Event 8.3.3 9.1 10.1.5
db file sequential read 181.00 0.00 0
CPU time 328.00 1105 146
SQL*Net more data to client 4 4.00 4
SQL*Net more data from client 8 9.00 0
control file parallel write 33.00 40.00 12
More data is being processed per second in 10.1
10.1 is more efficient on the database as db trips have been
reduced
Validation
Use Case Heavy Hitters
Send CC to MSE/MPP Send CC to MSE/MPP Login
Login Configure / un-configure
Product
Configure / un-configure Product
Select Occurrence type Filter
Select Occurrence type Filter
Load to Embedded Viewer/Process Simulate
Load to Embedded Viewer/Process Simulate
Logout Logout
Perform study with Process Simulate / Embedded Viewer Perform study with
Process Simulate / Embedded Viewer
Navigate to desired station/line Navigate to desired
station/line
Admin work
Bounding box creation
Bomwriter export
FIPA Create/Update
Heavy Hitters
Users Activities
Comparison Results Overview
Heavy Hitter item # BOMLines 8.3.3
(sec)
9.1 (sec)
10.1.5 (sec)
% Improvement 9.1 => 10.1.5
1-Create FIPA On a Station 28k 12k 243.00 58.20 39 33
2-Flat FIPA Update 28k 12k 127.80 55.00 37 34
3-MPP un-configured station Live Display 28000/ 5654JTs
total 3GB 918.00 798.00 599 25
3.1 MPP un-configured station DIPA Display 5645 JTs total 3GB NA NA 83 NA on 9.1
4-MPP configured station display with FIPA 5100/ 1704JTs total 663MB 96.00 129.00 75 42
5-PS configured station Live load using IPA 14700/2654 JTs total 1.1GB 636.00 330.00 189 43
6-PS configured station load using FIPA & PreGen Product 5100/1700 JTs total 670MB 70.20 46.20 44 5
Total 2091.00 1417.20 941 31%
• All FIPA based on flat mode. Additional study comparing to nested is followed
• Case 3.1 added to accommodate Dynamic IPA , it is comparable to heavy hitter item # 3 as far as the data is concerned where displaying a station at the end of a line is tested
Overall positive trend in particular as we look into the larger structure display in Process Simulate and EmbeddedViewer
Client server trips (RAC trips) stayed stable overall with still room for improvement – focused work underway
Database trips were only collected for 10.1.5 version
Keep striving to reduce client server calls to 100 or less and server database trips to 10000 or less
Wait Time vs. Calls Count
TC 9.1 vs TC 10.1.5
DB trips only measured for 10.1.5
Leveraging DIPA and LWB
Embedded visualization graphics load boost
# lines # JTs Total JTs (GB)
10.1.4 (IPA no LWB) 29000 5600 3
10.1.5 + DIPA-LWB 1034 5600 3
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
1-Create FIPA on a Station
2-Flat FIPA Update
3-MPP Un- configured station
live display with IPA
4-MPP configured station display
with FIPA*
5-PS configured station live load
using IPA
6-PS configured station load using
FIPA* & PreGen Product
Total
X Multiple
Latency Factor
100ms 200ms 300ms
Latency Client Server Calls Sensitivity Projection Process Plan Validation TC10.1.5 cached graphics
Factor calculated by: (time in <1ms) + (# API calls)*(Latency in seconds)
If we consider also JTs download impact:[(time in <1ms) + (# API calls+#JTs)*(Latency in seconds)]/(Original time in sec measured in LAN)
If JTs are not cached in case of Load Graphics JTs download also plays a role
By average 1.5Xslower in latency of 300ms over cached graphics
The impact of JTs download is relatively higher on small structures with less API calls
0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40 1.60 1.80 2.00
1-Create FIPA on a Station
2-Flat FIPA Update 3-MPP Un-configured station live display
with IPA
4-MPP configured station display with
FIPA*
5-PS configured station live load using
IPA
6-PS configured station load using
FIPA* & PreGen Product
Total
X Multiple
Latency Factor
100ms 200ms 300ms
If we factor the amount of JTs (~5650) downloaded multiple would be:~2.5X If we factor the amount of
JTs (~5650) downloaded multiple would be:~2.5X
If we factor the amount of JTs (~2654) downloaded multiple
would be:~18X If we factor the amount of JTs
(~2654) downloaded multiple would be:~18X
Documentation
Use Cases Heavy Hitters
Send CC to MSE/MPP Send CC to MSE/MPP Login
Login Configure
structures Configure
structures Navigate to desired itemNavigate to desired item Load to embedded viewer Load to embedded
viewer
Create snapshot
Create snapshot Apply to
snapshot Apply to snapshot Update snapshot
Update snapshot Find in product views *
Find in product views *
Logout
Logout
Heavy Hitters
Users Activities
* Not yet tested
Comparison Results Overview
Heavy Hitter item # BOMLines 9.1
(sec)
10.1.5 (sec)
% Improvement 9.1 => 10.1.5
Apply Snapshot 1 90/38 16 16 0%
Apply Snapshot 2 90/38 189 8 96%
Create Snapshot (Use Current) 90/38 17 6 65%
Update Snapshot 90/38 11 10 9%
Create Snapshot (Use configuration from Product
View) 90/38 6 4 33%
Total 239 44 82%
Resolving some issues produced much better results in 10.1.5
Average improvement is 82%
Overall better performance and similar # of calls , apply 1 time a snapshot has a bit more calls
currently analysis underway
Wait Time vs. Calls
Summary
Performance Improvement
Continuous performance improvement of 31% in the validation use case, 47% in BOM management use case and 49% in the BOP authoring use case when moving from 9.1 to 10.1.5 base
Positive trend in database trips reduction e.g. for accountability check of up to 74%
10.1.1+ Introduction of new mechanisms to enable nimbler faster expansions and graphic loads through introduction of LWB via end-item capabilities
Continuous focus on reduction of client server trips, in some areas regressions are being handled for 10.1.x versions
Further method improvements such as the Dynamic IPA additional capabilities where only desired data is being
loaded “on the fly” in a flat manner
Expanding Coverage of Performance Studies …
Additional opportunities for improvement, especially as new mechanisms are introduced helping us break through some problematic bottlenecks
Few areas under investigation specifically where there is a substantial number of server-database trips as well as some regression aspects in client server trips
Continue to expand performance Baselines & Scalability coverage mostly around the key heavy hitters derived from the main use cases
Continuous refinement of new functionality enabling new workflows for Manufacturing is being analyzed from
performance and scalability perspectives
On going development of performance holistic approach
Engaging at System Integration Testing:
• Establish good understanding of new developments
• Provide earlier cleanup of performance baselines
• Augment baselines with new functionality
Basic scalability analysis:
• Leveraging Siemens PLM performance and scalability lab services
• Data scalability studies leveraging customers like data
Basic criteria for unit clearance:
• Unit check points for development before delivering main code units
• Absolute performance criteria
Performance dashboards:
• Continuous refinement of dashboards
• Dashboards publications
Planning sessions engagements:
• Providing performance feedback for major performance projects
• Driving new performance projects for improving productivity
Continuous Improvements Approach
Continuous Improvements
• Reduce waste
• Optimize heavy hitters
• Optimize resources footprint
New Projects
• Further performance improvements
• Better methods for achieving goals
Continues Studies
• Continuous customers productivity assessments
• Measurable baselines
• Develop clear targets
Improve
Productivity
Appendix
Heavy Hitters Breakdown Studies
Load to Embedded Viewer
Comparison Results Overview
Embedded Viewer Load Time Comparison
Case Description #
BOM lines
# JTs
Sum of JTs size (GB)
8.3.3 (min)
9.1 (min)
10.1.5 (min)
9.1 10.1.5
% Improvement
1 Load large product 43000 8000 4.65 9.75 7.3 4.5 38%
2 Load small process 1877 2000 1.20 0.8 0.59 0.51 11%
3 Load configured IPA station over flat FIPA 5100 1600 0.66 1.57 2.15 1.25 42%
4 Load un-configured IPA station 28000 5500 3.00 15.3 13.3 10 25%
Average Improvement 29%
29% performance improvement by average
The larger the structure we see more improvement capacity
Comparison Results Overview
Embedded Viewer Load Client Server Calls Count Comparison
Case Description #
BOM lines
# JTs
Sum of JTs size (GB)
8.3.3 Calls
9.1 Calls
10.1.5 Calls
9.1 10.1.5
% Improvement
1 Load large product 43000 8000 4.65 1772 243 237 2.47
2 Load small process 1877 2000 1.20 80 50 61 -22.00
3 Load configured IPA station over flat FIPA 5100 1600 0.66 337 193 75 61.14
4 Load un-configured IPA station 28000 5500 3.00 2391 1648 600 63.59
Average Improvement 4580 2134 973 54.40
10.1.5 shows overall improvement of 54% in number of calls
Slight increase in the calls count for case 2 however this is still well below our sensitivity threshold hence 100 calls
Load Un-Configured IPA Station (2/2)
Client Server Calls Breakdown
DB Caching improves structure traverse by ~47%.
JTs download attributes ~20% of the time on cached DB and
~12% of the time on non cached DB.
Collecting properties for BOM lines has less impact by DB
cache.
Latency Sensitivity and client server trips correlation
Load Configured IPA Station (based on TC 8.3.2)
Linear pattern
The impact of latency increases with the number of calls
Load to Process Simulate
Comparison Results Overview
Process Simulate Time Comparison
Case Description #
BOM lines
# JTs
Sum of JTs size (GB)
8.3.3 (min)
9.1 (min)
10.1.5 (min)
9.1 10.1.5
% Improvement
1 Config. IPA Station over flat FIPA & PreGen. Prod. 5100 1600 0.66 1.17 0.77 0.73 5.19
2 Configured IPA station over flat FIPA 5100 1600 0.66 Not Tested 1.57 1.43 8.92
3 Configured IPA station with PreGen. Prod. Not Tested 2.80 2.7 3.57
4 Configured IPA station 14700 2508 1.08 10.6 5.53 3.15 43.04
5 Un-configured IPA station 28000 5500 3.00 Not Tested 10.87 7 35.60
Average Improvement 46% 19%
Average of 19%
Biggest improvements on cases 4,5
Case Description # BOM lines
# JTs
Sum of JTs size (GB)
8.3.3 Calls
9.1 Calls
10.1.5 Calls
9.1 10.1.5
% Improvement
1 Config. IPA Station over flat FIPA & PreGen. Prod. 20 10 -
2 Configured IPA station over flat FIPA 5100 1600 0.66 20 10 -
3 Configured IPA station with PreGen. Prod. 20 10 -
4 Configured IPA station 14700 2508 1.08 69 21 10 -
5 Un-configured IPA station 28000 5500 3.00 21 10 -
Average Improvement 71%
Comparison Results Overview
Process Simulate Load Client Server Calls Count Comparison
Structure size has less impact on the client server trips
Load to Process Simulate is positioned well for 4 tier
Process Simulate Load Flat FIPA vs Nested FIPA
68 – 75% spent on Teamcenter export activities mostly attributed to PLMXML export
Nested FIPA contains additional structure levels thus more time is spent on extracting the structures from Teamcenter
Preference controlled: MEFipaStructure
Structure Expand
Tree Expand Use Case Context
Send CC to MSE/MPP
Send CC to MSE/MPP Login
Login Configure
structures Configure
structures Navigate to desired itemNavigate to desired item
Full /x levels expand Full /x levels expand Unpack
Unpack Configure expanded
structure*
Configure expanded structure*
Logout Logout
* Not yet tested
Comparison Results Overview
Structure Expand Response Time Comparison
Case # Description #
BOM lines
8.3.3 (min)
9.1 (min)
10.1 (min)
9.1 10.1
% improvement
1 Large product expand below 15000* 4.49 2.87 1.65 42.50
2 Large process expand below 10000* 3.10 3.08 3.7 -20.12
3 Large product unpack all 30000 4.88 4.1 1.11 72.92
4 Large process unpack all 16000 1.28 1.25 1.46 -16.8
5 Product expand one level 130 0.08 0.08 0.04 50
6 Process expand one level 52 0.13 0.13 0.13 0
7 Cumulative time 13.96 11.51 8.09 29%
Significant improvement on large product expand
10.1 shows additional improvement of ~46% on average
Large process expand and unpack regression introduced in 10.1.5 (was not in 10.1.4) work underway to reduce the impact
* With packed lines
Comparison Results Overview
Structure Expand Client Server Calls Count Comparison
Case # Description #
BOM lines
8.3.3 Calls
9.1 Calls
10.1 Calls
9.1 10.1
% Improvement
1 Large product expand below 15000* 45 26 26 0.00
2 Large process expand below 10000* 36 20 20 0.00
3 Large product unpack all 30000 681 909 232 74.48
4 Large process unpack all 16000 379 458 198 56.77
5 Product expand one level 130 7 6 6 0.00
6 Process expand one level 52 11 2 2 0.00
7 Cumulative time 1159 1421 484 65.94
Significant improvement for 10.1, specifically in unpack
Unpack area is the most sensitive and shows regression in 9.1
Overall positive trend
* With packed lines
MPP Expand Large Product
Client Server Calls Breakdown
Significant improvement in ExpandOneLevel No change on client server calls
MPP Expand Large Process
Client Server Calls Breakdown
Significant improvement in ExpandOneLevel
Similar pattern on client server calls
Large Product Unpack All
Client Server Calls Breakdown
Significant improvement in response time as well as reduction
in calls count
Accountability Check
Accountability Use Case Context
Send CC Send CC Login
Login Open 1ststructure in
upper panel Open 1ststructure in
upper panel
Open 2ndstructure in lower panel Open 2ndstructure in
lower panel
Configure structures as needed
Configure structures as needed
Logout
Logout Closure Rule Based Accountability Check: LinkedAssembly or leaves / ProcessConsumption
Closure Rule Based Accountability Check: LinkedAssembly or leaves / ProcessConsumption
8.3+. Closure Rule based
No need for expand/unpack
More accurate report
Accountability Check Main Option Used
Classic
Closure
OOTB closure rule for the source structure (product)
Can create custom closure rule to address specific data
Closure Rules used:
Source: AccountabilityLinkedAssymOrLeaves
Target: AccountabilityProcessConsumption Display Options:
Missing target
Missing source
Multiple match
Comparison Results Overview
BOM/BOP Accountability Check # lines under selected
roots # lines returned 8.3.3 9.1 10.1 9.1 10.1.5
% Change
Response Time Compare in minutes 30k / 16k 13k / 11k 16.29 14.19 3.9 72
Client-server trip count compare 30k / 16k 13k / 11k 3906 121 550 -350
Server-database trip count compare 30k / 16k 13k / 11k 207127 194846 50740 74
Significant improvement in response time of 72%
Client server trips regression is currently under investigation. Significant reduction has been introduced on 10.1.3 compared with 10.1.1 however still ways to go to get back to 9.1 levels
Significant database trips reduction of 74%
Reduction of database trips helps in database and server scalability aspects
9.1 accountability in this case is best positioned to work well over 4 tier WAN
Not advisable to run very large accountability checks over WAN.
BOM/BOP Accountability Check
Client Server Calls Breakdown
Significant improvement in response time attributed to the
accountability server main call
Regression in the client server trips mostly attributed to getproperties type of calls. Those are in effect when accountability results are displayed in the UI.
This regression is currently under investigation and will be fixed in 10.1.x
Currently for large accountability studies either run on LAN or run only with report to excel
Accountability Check BOM-BOP Footprint Capture
Capturing Accountability Check footprint procedure
1. Set an environment variable MFG_TRAV_TIMER_ON=1 either on the session from which Teamcenter server starts or as user environment variable on the Teamcenter server machine
2. Run your accountability session, for example: TC login Load CC object set accountability check menus execute accountability check Log out collect the associated session associated syslog file
Analyzing the footprint
1. Open the associated syslog in a text editor
2. The 4 lines shown below can be found in the syslog as ordered:
1. 214529286 26957 1m 12.23s cpu, 1m 49.59s real MFGBVRClosureRulesServices::traverseByClosureRules Num = 10878
2. 22931483 1334 14.09s cpu, 20.99s real MFGBVRClosureRulesServices::traverseByClosureRules Num = 13552
3. 0 0 0.14s cpu, 0.14s real StructureVerificationImpl::fillOutSOAResults srcLines=4028 ,targetLines=1354, srcparents=524, targetparents=252
4. 251104654 45799 1m 50.79s cpu, 2m 48.50s real StructureVerificationImpl::accountabilityCheck End of MessageTimer
# of BOM lines traversed
# of BOP lines traversed
# of BOP lines returned
# of BOM lines returned
Total Server time
Server time to traverse the BOM
Server time to travers the BOP
PLMXML Export
PLMXML Export Use Case Context
Configure Process to reflect work order Configure Process to
reflect work order
Save configuration
Save
configuration Export Process Using PLMXML Export Process Using PLMXML Reformat PLMXML through style sheet
Reformat PLMXML through style sheet
PLMXML Main Count
Consumed parts 12454
Operations 1894
Workcenters 1981
Tools 210
Total PLMXML elements 57602
Category Product Process Plant
# of lines 30000 16000 16000
Structure Hierarchy 12 levels 15 levels 16 levels
# CAD files 5000 2000 4500
Overall CAD data size 500 MB 150 450
Data Export
PLMXML Configured Process Export 8.3.3 9.1 10.1 9.1 10.1
% Improvement
Response Time Compare (min) 30.00 25.00 10.13 59
# of Database trips 958659 762451 310392 59
Comparison Results
Configured Process Data
PLMXML Main Unpacked Packed
Consumed parts 12454 6618
Operations 1894 1894
Workcenters 1981 1981
Tools 210 210
Total PLMXML elements 57602 37371
Category Product Process Plant
# of lines 30000 16000 16000
Structure Hierarchy 12 levels 15 levels 16 levels
# CAD files 5000 2000 4500
Overall CAD data size 500 MB 150 450
Comparison Results
Exporting as Packed vs Unpacked
Data Export
PLMXML Configured Process Export 9.1 Unpacked 10.1 Unpacked 10.1 Packed Unpacked Packed
% Improvement
Response Time Compare (min) 25.00 10.13 7.8 23
# of Database trips 762451 310392 245032 21
Scalability Challenge Notes
Category Value
Required export capacity Assume up to: (10 models) x (10 orders per model) = 100 plans per day Time per order 0.4 hours / 40 min (based on average order per model)
Overall cumulative time 100 x 0.4 = 40 hours
Theoretical (sequential exports) 720 (min or 12 hour IT win) / 100 plans = 7.2 min per export required Practical target (parallel exports) Up to 10 plans in parallel based on experiments 720/10 = 72 min
Parallel PLMXML export per server 1 server will run with 10 plan in parallel would run 12 hours assuming database scalability 5 server running 50 plans in parallel would need ~ 2.4 hours to complete
Limit 10 exports per server to run in parallel for a 16GB RAM Tcserver, assuming database scalability
PLMXML BOP Export Scalability Study Highlights
(Based on TC 8.3.3.8)
# lines # Operations # Consumptions Sequential (min) In parallel (min) X impact
Program1 39340 1589 8050 12.41 35.00 2.8
Program2 42738 1706 8886 13.04 33.00 2.5
Program3 64432 2158 14532 26.64 64.00 2.4
Program4 54530 2214 11404 30.38 73.00 2.4
Program5 49009 1814 10243 19.18 48.00 2.5
Program6 32593 1089 7330 19.84 52.00 2.6
Program7 44187 1706 9109 42.10 93.00 2.2
Program8 39487 1629 8318 72.09 119.00 1.6
Program9 36624 1523 6745 63.32 119.00 1.8
Program10 65462 2166 14857 38.80 87.00 2.2
Average 46840.2 1759.4 9947.4 33.7 72.3 2.3
2.3X Teamcenter time factor on average when exporting 10 mix routings in parallel. The complete picture with T4S and
SAP is going to be impacted as well
Note: Program stands for Bill of Process
Scalability Analysis
Checking up to 30 parallel orders (Based on TC 8.3.3.8)
Be able to scale up and increase shop floor
orders capacity
Import the closure rule PreExpandProcess to the database
1. tcxml_import -u=<user_name> -p=<password> -file=<TC_ROOT>\data\defaultTransfermodes.xml - scope_rules scope_rules_mode=ignore
2. Ensure the above closure rule has been imported
Operation mode
1. Open a TCShell
a. Set PLMXML_Pre_Expand_BOP_Rule_Name=PreExpandProcess (case sensitive) b. Set MFG_TRAV_TIMER_ON=1
c. plmxml_export -u=<user> -p=<pass> -g=<group> -uid=<Process UID / CC UID> -
xml_file=%TC_TMP_DIR%\Name_export.xml -transfermode=<Relevant transfermode>
2. Check the plmxml_exportxxx.syslog a. search for “CPU” keyword
b. There should be 2 entries. The first entry is the relevant entry indicating a pre-expand 3. Define on what level to set the PLMXML_Pre_Expand_BOP_Rule_Name and set it (i.e. as a
preference on a user, as a preference on site or as an environment variable)
Note: from TC 8.3.3.8+ this is defined as a site preference and therefore the default behavior would be pre-expansion of PLMXML. To revert back to the original behavior create a dummy entry (ddd, for instance) which will override the pre-expand behavior.
PLMXML Export Pre Expand Settings
(TC 8.3.3.8 +)
PLMXML_export_packed_bom_<export transfermode>
Values: TRUE/FALSE
Description: Export as packed lines for the specific <export transfermode> when set to TRUE
Recommended value: TRUE on the user’s level to
PLMXML Export Packed Lines Settings
Enterprise Bill of Process (EBOP)
EBOP Use Case Heavy Hitters
Send CC to MPP Send CC to MPP Login
Login
PBOP Resolve Logical Assignment (LA)
PBOP Resolve Logical Assignment (LA)
Logout Logout
Clone PBOP from GBOP Clone PBOP
from GBOP
Launch Assignment Viewer
Launch Assignment Viewer
Auto allocate
PBOP#2 based on PBOP#1to PlantBOP
Auto allocate
PBOP#2 based on PBOP#1to PlantBOP
** Glossary
PBOP Product BOP, Bill of process based on product GBOP Generic BOP, Generic bill of process
PlantBOP Specific plant bill of process
LA Logical Assignment, ready made logical assignments to be resolved Resolve Physically connect to the assigned structure
Allocate Copy assembly into plantBOP
PBOP to PlantBOP Accountability Check
PBOP to PlantBOP Accountability Check
Allocate PBOP to PlantBOP
Allocate PBOP to PlantBOP Propagate changes
Propagate changes
TC 10.1.1 Results Overview
Response Time
# Functional step #
Lines # Operations #
LAs
9.1 Response (min)
10.1.1 Response
(min) % Improvement
1 Clone PBOP from GBOP 151 143 1.78 1.88 -6
2 Resolve 10 LAs 1 1 10 0.12 0.07 42
3 Resolve 100 LAs 1 1 100 0.57 0.45 21
4 Launch Assignment Viewer 1 op, 100 Resolved
LAs 1 100 0.2 0.12 40
5 Create new LA 1 op, 100 Resolved
LAs 1 100 0.28 0.15 46
6 Recursive Resolve 10 Proc, 100 Ops 100 550 3.95 2.35 41
7 Allocate resolved Operations (PBOP Plant BOP) 700 100 550 4.33 3.45 20
8* Accountability Check PBOP Plant BOP 1400 100 550 0.38 NA
9* Propagate LA changes 5 5 21 0.77 NA
10 Auto allocate PBOP#2 based on PBOP#1 110 110 550 4.77 3.57 25
Total 16 12 25
25% Average response time improvements
* At the time of testing, EBOP Accountability and Propagate were not available
TC 10.1.1 Results Overview
Client Server Trips
# Functional step #
Lines # Operations #
LAs
9.1 # Client-Server Trips
10.1.1 #
Client-Server Trips % Improvement
1 Clone PBOP from GBOP 151 143 77 102 -32
2 Resolve 10 LAs 1 1 10 33 16 52
3 Resolve 100 LAs 1 1 100 23 16 30
4 Launch Assignment Viewer 1 op, 100 Resolved
LAs 1 100 6 10 -67
5 Create new LA 1 op, 100 Resolved
LAs 1 100 14 20 -43
6 Recursive Resolve 10 Proc, 100 Ops 100 550 1088 11 99
7 Allocate resolved Operations (PBOP Plant BOP) 700 100 550 12 11 8
8* Accountability Check PBOP Plant BOP 1400 100 550 13 NA NA
9* Propagate LA changes 5 5 21 3043 NA NA
10 Auto allocate PBOP#2 based on PBOP#1 110 110 550 2 2 0
Total 4311 298 93
Average 93% reduction in trips Slight regression in Clone GBOP
* At the time of testing, EBOP Accountability and Propagate were not available.
