[MMTests] dbench4 async on ext3

From: Mel Gorman
Date: Mon Jul 23 2012 - 17:22:26 EST

Configuration: global-dhp__io-dbench4-async-ext3
Result: http://www.csn.ul.ie/~mel/postings/mmtests-20120424/global-dhp__io-dbench4-async-ext3
Benchmarks: dbench4


In general there was a massive drop in throughput after 3.0. Very broadly
speaking it looks like the Read operation got faster but at the cost of
a big regression in the Flush operation.

Benchmark notes

mkfs was run on system startup. No attempt was made to age it. No
special mkfs or mount options were used.

dbench 4 was used. Tests ran for 180 seconds once warmed up. A varying
number of clients were used up to 64*NR_CPU. osync, sync-directory and
fsync were all off.

Machine: arnold
Result: http://www.csn.ul.ie/~mel/postings/mmtests-20120424/global-dhp__io-dbench4-async-ext3/arnold/comparison.html
Arch: x86
CPUs: 1 socket, 2 threads
Model: Pentium 4
Disk: Single Rotary Disk


Generally worse with a big drop in throughput after 3.0 for small number
of clients. In some cases there is an improvement in latency for 3.0
and later kernels but not always.

Machine: hydra
Result: http://www.csn.ul.ie/~mel/postings/mmtests-20120424/global-dhp__io-dbench4-async-ext3/hydra/comparison.html
Arch: x86-64
CPUs: 1 socket, 4 threads
Model: AMD Phenom II X4 940
Disk: Single Rotary Disk
Status: Ok

Similar to arnold, big drop in throughput after 3.0 for small numbers
of clients. Unlike arnold, this is matched by an improvement in latency
so it may be the case that IO is more fair even if dbench complains
about the latency. Very very broadly speaking, it looks like the read
operation got a lot faster but flush got a lot slower.

Machine: sandy
Result: http://www.csn.ul.ie/~mel/postings/mmtests-20120424/global-dhp__io-dbench4-async-ext3/sandy/comparison.html
Arch: x86-64
CPUs: 1 socket, 8 threads
Model: Intel Core i7-2600
Disk: Single Rotary Disk

Same story, big drop in throughput after 3.0 with flush again looking very
expensive for 3.1 and later kernels. Latency figures are a mixed bag.

Mel Gorman
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/