> This test creates memory pressure, but shouldn't OOM:
2.5.62-mm1 ran AIM7 compute and other tests without OOM.
> I'll try 2.5.61-mm1 with elevator=cfq.
I ran 2.5.61-mm1 with both default and cfq I/O elevators.
These results could be skewed by the mm_struct leak in 61-mm1,
but FWIW:
Default 2.5.61-mm1 I/O scheduler did better in most cases:
AIM7 dbase test:
load jobs/minute (higher is better)
2.5.61-mm1 8 76.6
2.5.61-mm1-cfq 8 68.7
2.5.61-mm1-cfq 16 100.0
2.5.61-mm1 16 132.4
2.5.61-mm1-cfq 32 136.5
2.5.61-mm1 32 198.7
AIM7 fileserver test
load jobs/minute (higher is better)
2.5.61-mm1-cfq 10 149.8
2.5.61-mm1 10 175.3
Tiobench ext sequential reads
threads MB/sec CPU usage max latency
2.5.61-mm1-cfq 16 9.34 14.06% 497.8 (seconds)
2.5.61-mm1 16 15.52 53.86% 5.2
More quad xeon benchmarks at:
http://home.earthlink.net/~rwhron/kernel/bigbox.html
-- Randy Hron http://home.earthlink.net/~rwhron/kernel/bigbox.html- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Sun Feb 23 2003 - 22:00:29 EST