Messages in this thread | | | Date | Tue, 17 Jun 2003 16:27:06 +1000 | From | Nick Piggin <> | Subject | Re: i/o benchmarks on 2.5.70* kernels |
| |
On Tue, Jun 17, 2003 at 01:32:32AM -0400, rwhron@earthlink.net wrote: > Some benchmarks on 2.5.70* kernels. Hardware is quad P3 xeon > with 3.75 GB ram. For some workloads, -mm has higher numbers > than 2.5.70, for others, 2.5.70 is better. Overall, 2.5.70 > seems to give more balanced results on these benchmarks. >
Thanks.
Recent mm's have had some experimental fair request allocation code in them which should be sorted now (not yet in a released mm), and should improve performance of "queue full" loads such as many writers or readers.
It might not get the throughput of mainline kernels, but I have seen 5x better max latency in some cases. It would be interesting to see how it goes on a big box.
Don't know whats going on with AIM7, but there have been a few bugs in AS being fixed.
tiobench on SMP results are not very good, lots of fragmentation, the random IO throughput drops is probably due to AS strangling TCQ though. You are using SMP and TCQ, right?
- 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/
| |