138. 解决teamcenter性能差的建议----值得学习!!!!

admin

http://www-cad.fnal.gov/PLMWorld2008/Teamcenter%20Unified/The%2520Network%2520is%2520Slow!.pdf
( ~, p: ?0 n) n0 b# E6 W  N; T' Hhttp://www.plmworld.org/p/bl/et/blogid=43&blogaid=218
https://support.industrysoftware.automation.siemens.com/docs/teamcenter/Network_Performance_Tuning_V6.6.pdf

4 {$ z. P- C" a. BTeamcenter is Slow! Teamcenter Performance Bottlenecks & Mitigation

- A9 `2 G3 l( b6 Q. ?# fLarry Carpenter
/ Y% e* T: w% x9 m, a# N4 fAt the Chicago-Wisconsin RUG Fall Conference in October 2012, I gave a presentation on common Teamcenter performance issues and solutions along with listing, in one place, many other past Teamcenter performance presentations and white papers. It was very well received by a standing room crowd, so I thought I'd share it with the rest of the PLM World community. Here is a link to the presentation (you must be a TCUA SIG member to view it):  http://www.plmworld.org/p/do/sd/sid=3758&type=0
$ B! X$ ]. ^3 pFor those unwilling to join the TcUA SIG to view it, I wrote an abbreviated version here:
Why Performance Matters
• Productivity – doing more work with less
, o. M9 f! S! g: K" _o Improve end user productivity
6 A6 s- }6 p" A2 L; n+ h' Uo Improves administrator productivity
 fewer help desk calls responding to or fixing preventable problems
  c/ i6 K0 L% ]/ B. o. }( f6 S• Reliability
8 b9 p! U2 N: k$ Uo A well tuned Teamcenter environment also improves system stability and uptime.
o Less opportunity for data corruption
o Fewer operational errors/problems
2 h1 T: Y" _' a, K+ O* v7 ^3 Q• Cost Savings
. w) N! @* Y% N& l5 E  p  \o Less waiting means less time wasted.
5 ]' ~; o9 w5 _3 n. m! qQuick Case Study: Company ‘S’
Performance was so bad that something which used to take 10 minutes was taking 2 hours to do.Did pretty much everything wrong, performance-wise, at first but ultimately fixed every major problem. What took 2 hours now only takes 2 minutes.
: @7 O! J2 ]  U' iSee my March 2012 article on PLM World’s website for details:
& h9 X, l; S0 i2 j5 j1 JTeamcenter Performance – Hard-Earned Lessons:  http://www.plmworld.org/p/bl/ar/blogaid=152
Common Performance Bottleneck Causes
Using OOTB settings — OOTB settings are intended for development environments; not production environments.
; V5 N$ F' X6 e- ?# H5 m) k9 \. jOverloaded servers — Putting too much on a single server.
! a9 C4 G% N( A2 v2 ^) l: @Operating system settings — Network settings most common (e.g. TCP parameters).
' J: ]  x4 \  B' E( x* u2 V2 SLack of performance monitoring & tuning — Performance doesn’t improve all by itself. It requires human intervention to determine root cause and address it.
5 D' O0 k: f/ K. YDatabases – The Most Likely Culprit
• Database performance is highly sensitive to hardware, software, and DB configuration.
2 m. _- W# s! F3 G/ Z& ^: I• DBAs at companies generally don’t do performance monitoring and tuning of your Teamcenter DB.
. C9 c( W- N% L1 O5 v' H! J• It’s typically the last place that Teamcenter administrators look for performance issues rather than the first.
# X% f& |& q: P) d. O8 \) I) `; GMitigating Database Server Bottlenecks
• Must use dedicated DB server
o Do not use your DB server for anything other than your Teamcenter production database. This includes not serving additional databases from the same server.
• Must have fast dedicated storage for DB files
' ]+ z3 s: O3 io Avoid using a NAS device; especially a shared NAS. Use DAS (Direct Attached Storage) instead.
9 S, ]+ B! ~' C4 C6 _# {0 C: Uo Use multiple fast disk spindles partitioned with proper RAID levels according to the DB vendor & SPLM recommendations.
o Split DB data, temp, and log files across those RAID partitions according to DB vendor & SPLM recommendations.
o Use disk controllers with a battery/flash backed cache.
0 D, G9 G5 x' ?& M" L+ m# ?o Use multiple disk controllers if possible.
• Cram the RAM
o Ideally should be greater than the database size or in-memory footprint. Otherwise disk swapping/paging will occur.
• Use 64-bit OS & DB software
9 Q/ |" I8 y- |! M/ i: e) {o 32-bit software has severe limitations regardless of whether you use “/3G”, AWE, or PAE settings to access memory beyond 3-4G. It’s still a bottleneck.
& g0 P1 k" @. I  }6 Z3 i; L: d• Use a good quality network adapter(s)
o Often overlooked as potential source of bottleneck (e.g. packets/sec limitations). Consider multiple NICs “trunked” for better throughput.
  R6 `7 [5 u8 X  m$ a& \• DB maintenance tasks
8 F3 Q6 @# g8 @0 U( Ao Update statistics and rebuild indexes regularly.
Common Teamcenter Server Bottleneck Causes
& U& d$ m# H; o• Overloaded Tc Servers
7 s9 A) o8 m$ f" d. j1 ?' N• Poor Web Tier Configuration
1 d7 S8 V3 _, \/ G' a: Q• Poor FMS Configuration
" a) X3 |3 I: W- H* n1 L+ F• Debugging Turned ON
• Rich Client using OOTB settings
‘Unload’ Overloaded Teamcenter Servers
2 n8 `$ G6 P: h) a• Put Dispatcher modules on separate computers away from other Tc Servers.
1 N# p! ^: F' z- L: c4 n1 v9 k6 Z• Separate the Web & Enterprise tiers from the Tc Corporate and Volume servers.
) f2 T( W- n8 r0 }+ {7 F# d6 x• Use load balancing for Middle Tier & FMS:
o Use multiple Web/Enterprise Tier servers to open up potential hardware resource bottlenecks (e.g. CPU, RAM, network adapter I/O, disk I/O, MB bus, etc.).
o Set up multiple FSC cache servers to take load off busy volume servers/storage.
Web Tier Configuration
• Do not use port 80 or 8080
# t* o1 v! U: ao HTTP traffic on those ports is considered web browsing traffic and is therefore given lowest priority on any network. Can also cause randomly dropped connections.
6 e. d) h# Y8 m$ _, I& H; Y4 Y; L• Enable http compression if not using WAN acceleration
o Must be done on both web server and on clients to take effect.
% W- b7 j8 n* D3 B• Change/tune OOTB settings for Web Server/App
o E.g. timeout values, max # of threads, Java memory, etc.
o Read tuning guides specific to your chosen web server/app
• Scale it up or sideways
& B4 H* M: A7 b5 T! s/ ?0 V  eo Add another web tier server, or increase # of work processes (e.g. Web Garden mode for IIS).
! c% y* h- N/ i* Y* a7 T/ bCommon FMS Bottleneck Causes
• Data improperly routed
o E.g. Forcing data to go through a remote FSC server over a WAN and back again over the same WAN instead of simply pulling it directly from a nearby Volume/cache server.
• Using OOTB settings
o For development purposes only, remember?
• Missing client IP address subnets
) f2 u( d" W* j! G& R6 a" |2 |• No load balancing
  R0 q& }5 [* q. L; Z8 l• Not using remote cache servers for WAN users
* B7 i2 k7 U7 u3 L; o+ t- \• Not using remote volume servers for WAN users
FMS Configuration
• Ensure routing is correct
o Between multiple FSC groups via entry/exit/link parameters
$ w) f: x/ S, b* S: T. s* Co Between clients and their assigned FSC servers by using complete and accurate ‘clientmap’ parameters for ALL clients.
/ b9 F- {7 U7 d$ y" jo Use correct transport algorithms for LAN and WAN users.
+ r6 f6 {) W9 q; T2 |- c7 C• Tune FSC/FCC cache settings
o Simply picking a total read/write cache size is not enough.
o Read “Sizing the FMS fast cache” in the Teamcenter help documentation. There is also a FMS cache sizing tool available from the GTAC web site. Link to FMS Calculator
* F) i+ k3 N# P) o! N, U9 F! [+ P• Ensure correct client maps
o Determine ALL potential client IP address ranges and where they are located. Don’t forget to account for VPN IP addresses.
o Place client maps in the closest FSC group to their location.
• Use multiple FSCs for load balancing
o Use “front-end” FSC cache servers to reduce load on Volume servers. (Requires disabling of FSC direct routing)
o Use redundant FSC cache servers to load balance each other. Also provides fail-over.
( {$ P4 u5 x% n! {9 L/ Q• Place FSC cache servers close to users
o Greatly improves read performance (load time) over a WAN for groups consuming shared data.
• Place Volume servers close to users
0 c) |; {3 |. e! S1 oo Use Store & Forward or at least place remote volumes near remote users. (Backups of remote volumes are critical)
o Greatly improve write performance (save time) over a WAN especially for CAD data.
• Prepopulate FSC caches
o Run a nightly script to prepopulate shared data across FSC caches.
Misc Teamcenter Changes
0 p8 O6 {# x) v+ C: |6 N0 r6 ]• Disable logging and other debugging tools.  Turn them on only when needed. Examples,
o TC_SECURITY_LOGGING=OFF
o TC_APPLICATION_LOGGING=OFF
  J8 y. u3 m0 `3 _% @# `6 \+ Do TC_SLOW_SQL=-1
• Rich Client ‘Teamcenter.ini’ configuration:
o Modify Teamcenter.ini file to increase Java Memory and other settings. Will improve performance/stability when perform large operations (CAD, PSE expansion, large Workflows).
# ^( S  U/ }0 f* |) p- c• Enable FCC File Warming
4 V5 }, D+ K4 y6 v: M& e2 A. Qo Improve startup time by having FCC and Tc files pre-loaded at OS login.
Network Performance
5 _  a) W8 i( X• Use a hardware-based WAN accelerator if using Teamcenter between multiple facilities over a WAN.
o If none then make best use of software-based Web server and FMS WAN acceleration settings. (i.e. http compression & FMS compression)
• Optimize OS network settings
* R1 `. k+ D( @' V$ l4 yo OOTB settings are insufficient
+ K! C. d4 s1 h2 k2 Wo Applies to both servers and clients
: S9 S, e5 T) o: z. J( U! ~. xo Examples, increasing # of tcp ports, disabling NAGLE algorithm, increasing buffer sizes, trunking multiple NICs, enabling Compound TCP, Large Windows Scaling
Recommended Server Changes (Windows)
# K5 b( J, ?* G+ g4 D2 ?8 t6 R3 [See Presentation.
Recommended Client Changes (Windows)
/ o3 A$ f4 N+ k$ ^See Presentation.
Performance Monitoring Tools
• Some useful performance monitoring tools:
% {) m5 ?  y/ h7 q5 {! o3 C6 w• Performance Monitor (comes with Windows)
7 a) Y, a) J% \$ {# M+ f* y1 J- m• Used to monitor OS, Teamcenter, and SQL use of resources
# K" ~. _. A1 L" m• Wireshark (freeware network packet sniffer)
• Used to help identify network TCP performance between computers.
! R4 w1 Q! O: Q/ ?  H1 x- e3 K• Available at http://www.wireshark.org/
: H/ V' e5 b1 f: y8 l4 N• Sysinternals Suite (free from Microsoft web site)
* v9 B7 ]" V: t6 S, E  `• http://technet.microsoft.com/en-us/sysinternals/bb842062
  S. u. Z& m& v! t4 }5 j( @9 nReference Materials
+ F: @7 @9 C' ^+ Y% c% L) ~• Available on GTAC Web Site @ http://support.industrysoftware.automation.siemens.com/docs/teamcenter/
' [) B. A7 k) {+ D/ Z* o2 g0 f% @# H, B# Mo Teamcenter Deployment Guide
o Teamcenter Network Performance Tuning
o JBoss 4.2.2 Installation & Tuning Guide
o Optimizing Teamcenter Client Performance
o Teamcenter System Performance Analysis
• Oracle documentation & web sites
" F- D5 Y) |1 t1 Z, \6 j' B• MS SQL Server documentation in addition to:
3 S7 l) l6 I# q9 U) vo Best Practices for Running Siemens Teamcenter on SQL Server
: v% c8 |# D1 Y. ]0 R; p http://download.microsoft.com/download/7/3/6/7365D2BB-BB34-4D28-A128-F2C8FBA6E995/Siemens-Teamcenter-and-SQL-Server-Best-Practices.pdf
o Siemens Blog on Technet
 http://blogs.technet.com/b/sql_server_isv/archive/tags/siemens/
o Siemens-Teamcenter-SQL-Resource-Page
 http://blogs.technet.com/b/sql_server_isv/p/siemens-teamcenter-sql-server-resource-page.aspx
, S* H3 C& Y9 `- G9 a+ V$ l# A• Past PLM World Conference Presentations available at www.plmworld.org,
o Teamcenter 4-Tier Deployment – Best Practices
" ?+ `+ l. t5 M! e5 v. ^% Eo Teamcenter – Database Performance
  J: \: c0 d& o$ X$ g+ d) d$ no JBoss Optimization for Teamcenter
& L0 l& o* U3 Oo Optimizing Teamcenter Client Performance
o Teamcenter Database Server Configuration & Tuning
+ S& J% f- `: }4 q% IContact Information
Larry Carpenter P.E.,  PLM World Teamcenter SIG Chair,  tcua@plmworld.org
Teamcenter UA SIG:  http://www.plmworld.org/TC_UA
LinkedIn:  http://www.linkedin.com/pub/larry-carpenter-pe/44/5b8/aaa
. E6 _# T) i; X: E8 c5 QAlternate Contact Info: ideas2nx@plmworld.org, larry.carpenter@siemens.com
7 ]2 f* _9 }, k

xiefeixiang

调优厉害啊

a150159

总结全面，我们也是类似情况，找了西门子的团队过来优化的