Here are the results of running dbench, kernbench, tbench, and SDET on
2.6.13-rc3-mm3 with and without CKRM.
Most results are within 1 standard deviation of a non-CKRM-enabled
kernel.
I used 10 to 20 iterations of each test to calculate the mean and
standard deviation. Standard deviations are provided in percent of the
baseline to give some idea of the resulting precision. dbench runs used
ramfs thereby factoring out imprecision apparently caused by disk
access.
All tests with CKRM disabled are the baseline at 100%.
All tests with CKRM enabled include the existence of 100 classes
without shares set and without the numtasks controller enabled.
Highest CKRM score: 102% +- 2 (dbench)
Lowest CKRM score: 97% +- 4 (tbench)
Highest standard deviation: +-4 (tbench)
Is anyone interested in seeing comparitive results from other
benchmarks?
4-processor NUMAQ with 3GB of RAM
********************************************************************************
dbench
================================================================================
Config Throughput (%)
--------------------------------------------------------------------------------
CKRM with 100 classes 102.11 +- 1.9
Baseline 100.00 +- 0.8
kernbench
================================================================================
Config User (%) System (%) Elapsed (%) %CPU
--------------------------------------------------------------------------------
CKRM with classes 99.98 +- 0.0 99.79 +- 0.3 99.96 +- 0.1 100.00 +- 0.1
Baseline 100.00 +- 0.0 100.00 +- 0.4 100.00 +- 0.2 100.00 +- 0.2
SDET [1]
================================================================================
128 scripts
Config Throughput (%)
--------------------------------------------------------------------------------
CKRM with 100 classes 99.84 +- 0.3
Baseline 100.00 +- 0.4
64 scripts
--------------------------------------------------------------------------------
CKRM with 100 classes 99.65 +- 0.5
Baseline 100.00 +- 0.5
16 scripts
--------------------------------------------------------------------------------
CKRM with 100 classes 99.62 +- 1.3
Baseline 100.00 +- 1.3
4 scripts
--------------------------------------------------------------------------------
CKRM with 100 classes 100.08 +- 1.3
Baseline 100.00 +- 1.2
1 script
--------------------------------------------------------------------------------
CKRM with 100 classes 99.80 +- 0.2
Baseline 100.00 +- 0.2
tbench
================================================================================
Config Throughput (%)
--------------------------------------------------------------------------------
CKRM with 100 classes 97.07 +- 4.4
Baseline 100.00 +- 3.8
4-processor pSeries Power4 with 3GB RAM
********************************************************************************
dbench
================================================================================
Config Throughput (%)
--------------------------------------------------------------------------------
CKRM with 100 classes 99.74 +- 1.8
Baseline 100.00 +- 2.0
kernbench
================================================================================
Config User (%) System (%) Elapsed (%) %CPU
--------------------------------------------------------------------------------
CKRM with classes 100.24 +- 0.1 99.51 +- 0.4 99.97 +- 1.5 100.19 +- 1.5
Baseline 100.00 +- 0.1 100.00 +- 0.5 100.00 +- 1.2 100.00 +- 1.2
tbench
================================================================================
Config Throughput (%)
--------------------------------------------------------------------------------
CKRM with 100 classes 99.14 +- 3.4
Baseline 100.00 +- 4.2
[1] DISCLAIMER: SPEC(tm) and the benchmark name SDET(tm) are registered
trademarks of the Standard Performance Evaluation Corporation. This
benchmarking was performed for research purposes only, and the run
results are non-compliant and not-comparable with any published results.
-------------------------------------------------------
SF.Net email is Sponsored by the Better Software Conference & EXPO
September 19-22, 2005 * San Francisco, CA * Development Lifecycle Practices
Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA
Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf
_______________________________________________
ckrm-tech mailing list
https://lists.sourceforge.net/lists/listinfo/ckrm-tech
