Monitoring with PerfSuite database (manual check)*

CMSSW_6_0_0_pre7 to pre4

The following prereleases were checked on July 9th, 2012:

# CMSSW_6_0_0_pre7 >> TTBAR >> GEN,FASTSIM,HLT:GRun (RECOSIM, LowLumiPileUp, MC_53_V6::All) >> slc5_amd64_gcc462
# CMSSW_6_0_0_pre6 >> TTBAR >> GEN,FASTSIM,HLT:GRun (RECOSIM, LowLumiPileUp, MC_53_V6::All) >> slc5_amd64_gcc462
# CMSSW_6_0_0_pre5 >> TTBAR >> GEN,FASTSIM,HLT:GRun (RECOSIM, LowLumiPileUp, MC_53_V4::All) >> slc5_amd64_gcc462
# CMSSW_6_0_0_pre4 >> TTBAR >> GEN,FASTSIM,HLT:GRun (RECOSIM, LowLumiPileUp, MC_53_V4::All) >> slc5_amd64_gcc462

Contact with RelVal (Alan)

*600_pre7_event.png600_pre6_event.png

CPU TIME

* CMSSW_6_0_0_pre7 and pre6 * Both prereleases presented many problems in the plots * request rerun to confirm software malfunction * CMSSW_6_0_0_pre5 * it does not show problem, but it is running in less than 1 sec (problem?!) * famosSimHits is consuming 20% more CPU time than the previous prerelease * CMSSW_6_0_0_pre4 * Modules are consuming CPU time as expected

CPU TIME PER EVENT

600_pre5_cputime_event.png

CMSSW_6_0_0_pre2 and CMSSW_5_2_0_pre5

The following prereleases were checked on May 21st, 2012:

# CMSSW_6_0_0_pre2 >> TTBAR >> GEN,FASTSIM,HLT:GRun (RECOSIM, LowLumiPileUp, MC_52_V09::All) >> slc5_amd64_gcc462
# CMSSW_5_2_0_pre5 >> TTBAR >> GEN,FASTSIM,HLT:GRun (RECOSIM, LowLumiPileUp, MC_52_V01::All) >> slc5_amd64_gcc461

Module RECOSIMoutput taking too much CPU time

* an average CPU time of 0.300, while the module famosSimHits has an average of 0.302. * the average CPU time for this module is 8.72% (3rd in rank), but consistent with previous releases.

Behavior in CPU time per event

600_pre2_cpu-time.png600_pre2_cpu-time.png

VSIZE reaching high memory limit

* 8000th event at 2900MB.

VSIZE chart

* it has a more smooth growing -- the same is observed in releases 5_3_0, 5_2_5, and 5_1_2. * other releases have a step-like growing with virtual memory. * 6_0_0_pre2 and 5_2_0_pre5: more events are using less memory than previous releases, but not the 3-peak behavior as seen in 5_3_0. * there are small oscillations in the virtual memory chart than in previous releases (5_3_0). * Maybe a different method has been implemented in these prereleases, or it happens due to the MC conditions.

600_pre2_VSIZE.png530_VSIZE.png530_VSIZE.png

RSS chart

* 6_0_0_pre2 and 5_2_0_pre5 have large oscillations than previous releases. * however RSS size per event is more spread (more events using less memory)

CMSSW_5_2_0_pre4 to 5_2_0_pre1

The following prereleases were checked on May 21st, 2012:

# CMSSW_5_2_0_pre4 >> TTBAR >> GEN,FASTSIM,HLT:GRun (RECOSIM, LowLumiPileUp, MC_52_V01::All) >> slc5_amd64_gcc461
# CMSSW_5_2_0_pre3 >> TTBAR >> GEN,FASTSIM,HLT:GRun (RECOSIM, LowLumiPileUp, MC_50_V13::All) >> slc5_amd64_gcc461
# CMSSW_5_2_0_pre2 >> TTBAR >> GEN,FASTSIM,HLT:GRun (RECOSIM, LowLumiPileUp, MC_50_V09::All) >> slc5_amd64_gcc461
# CMSSW_5_2_0_pre1 >> TTBAR >> GEN,FASTSIM,HLT:GRun (RECOSIM, LowLumiPileUp, MC_50_V09::All) >> slc5_amd64_gcc461

VSIZE chart

* fewer events are using more memory.

RSS chart

* fewer events are using more memory (peak around 1200MB).



CMSSW_5_X_X series

The following releases were checked on May 21st, 2012:

# CMSSW_5_3_0 >> TTBAR >> GEN,FASTSIM,HLT:GRun (RECOSIM, LowLumiPileUp, MC_53_V4::All) >> slc5_amd64_gcc462
# CMSSW_5_2_5_cand1 >> TTBAR >> GEN,FASTSIM,HLT:GRun (RECOSIM, LowLumiPileUp, MC_52_V9::All) >> slc5_amd64_gcc462
# CMSSW_5_2_5 >> TTBAR >> GEN,FASTSIM,HLT:GRun (RECOSIM, LowLumiPileUp, MC_52_V9::All) >> slc5_amd64_gcc462
# CMSSW_5_1_2 >> TTBAR >> GEN,FASTSIM,HLT:GRun (RECOSIM, LowLumiPileUp, MC_50_V15A::All) >> slc5_amd64_gcc462

All releases

RSS chart

* the increasing of RSS with the number of events shows oscillations like in VSIZE, with a smaller amplitude observed in 5_2_5. * other releases have a step-like growing in RSS.

CMSSW_5_3_0

CPU time per module=

* Module =hltL1SeededElectronGsfTracks is loaded only in this release. * it has an average CPU time of 0.142 -- Is that OK?

CMSSW_5_2_5_cand1

VSIZE chart

* There are smaller oscillations than in previous releases. * virtual memory reaching 2950 MB at the 8000th event. * If compared to other releases, fewer events are using more memory.



CMSSW_4_X_X series

The following releases were checked on May 21st, 2012:

# CMSSW_4_4_3 >> TTBAR >> GEN,FASTSIM,HLT:GRun (RECOSIM, LowLumiPileUp, MC_44_V7::All) >> slc5_amd64_gcc434
# CMSSW_4_4_2_patch10 >> TTBAR >> GEN,FASTSIM,HLT:GRun (RECOSIM, LowLumiPileUp, MC_44_V7::All) >> slc5_amd64_gcc434
# CMSSW_4_4_2_p10JEmalloc >> TTBAR >> GEN,FASTSIM,HLT:GRun (RECOSIM, LowLumiPileUp, MC_44_V7::All) >> slc5_amd64_gcc434
# CMSSW_4_4_2 >> TTBAR >> GEN,FASTSIM,HLT:GRun (RECOSIM, LowLumiPileUp, MC_44_V7::All) >> slc5_amd64_gcc434
# CMSSW_4_4_1 >> TTBAR >> GEN,FASTSIM,HLT:GRun (RECOSIM, LowLumiPileUp, MC_44_V6::All) >> slc5_amd64_gcc434
# CMSSW_4_2_8_patch7 >> TTBAR >> GEN,FASTSIM,HLT:GRun (RECOSIM, LowLumiPileUp, MC_42_V12::All) >> slc5_amd64_gcc434
# CMSSW_4_2_3__STRAIGHT >> TTBAR >> GEN,FASTSIM,HLT:GRun (RECOSIM, LowLumiPileUp, MC_42_V12::All) >> slc5_amd64_gcc434
# CMSSW_4_2_3 >> TTBAR >> GEN,FASTSIM,HLT:GRun (RECOSIM, LowLumiPileUp, MC_42_V12::All) >> slc5_amd64_gcc434

Many releases

RSS chart

* No uniform behavior in the RSS. * CMSSW_4_4_2_p10JEmalloc: oscillations observed like in 5_X_X series, but the behavior of number events vs RSS size was not seen in other releases. * CMSSW_4_X_X: step-like growing in RSS, but no oscillations (new method employed?). * 4_2_3 and 4_2_3__STRAIGHT: step-like growing in RSS with small oscillations around the 1000th event.

CMSSW_4_4_3

* The module RECOSIMoutput has an average CPU time higher than all other modules. * the same behavior is seen in all releases in this series since 4_4_1.

CMSSW_4_2_X

* Module allConversions has a high average CPU time (14.5%). * 4_2_8_patch7, 4_2_3__STRAIGHT and 4_2_3: average CPU time is around 0.490, while for RECOSIMoutput it is around 0.360, and in famosSimHits is 0.315.

Bumps

The following releases presented a bump in CPU time per event.

CMSSW_4_4_2 CMSSW_4_4_1 CMSSW_4_2_3__STRAIGHT, and CMSSW_4_2_3
bump at the 7500th event bump at the 4500th event not at the same range as seen in preceding releases bump at the 7300th event large peak around the 550th event with CPU time of 40s
442_bump.png 441_bump.png 423_STR_bump.png



Edit | Attach | Watch | Print version | History: r1 | Backlinks | Raw View | WYSIWYG | More topic actions
Topic revision: r1 - 2013-01-20 - GustavoGilDaSilveira
 
    • Cern Search Icon Cern Search
    • TWiki Search Icon TWiki Search
    • Google Search Icon Google Search

    CMSPublic All webs login

This site is powered by the TWiki collaboration platform Powered by PerlCopyright & 2008-2019 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback