|
|
Subscribe / Log in / New account

OSDL: AIO-Stress summary for AIO patch set vs linux-2.6.8

From:  Judith Lebzelter <judith@osdl.org>
To:  <suparna@in.ibm.com>
Subject:  OSDL: AIO-Stress summary for AIO patch set vs linux-2.6.8
Date:  Fri, 27 Aug 2004 16:16:43 -0700 (PDT)
Cc:  <linux-aio@kvack.org>, <mason@suse.com>, <linux-osdl@osdl.org>, <linux-kernel@vger.kernel.org>

Hello;

I have done a summary of some aio-stress runs on our 1CPU and 2CPU hosts
to compare the performance of the latest version of Suparna's buffered AIO
patch set http://marc.theaimsgroup.com/?l=linux-aio&m=109285137...
(less the aio-ra) to the linux-2.6.8 kernel.  I found up to a 442%
improvement and was wondering if anyone could explain it.

The buffered results on 1CPU/single disk/512M RAM/1 2G File showed:
     Sequential Reads:    -13.2%
     Random reads:        +32.7%


The buffered results on 2CPU/5 Disk striped Megaraid/1G RAM/1 4G File
showed:
    Sequential reads:      +2.1%
    Random Reads:         +442%


The improvement in buffered random reads from this patch set was +32% on
1CPU and +442% on 2CPU.  This was more than expected on the 2CPU host at
+442%.  It seems as though the 'buffered random reads' performance on the
unpatched linux-2.6.8 kernel is unexpectedly low at only 7.57MB/s on our
5 Disk striped Megaraid partition.

The buffered sequential reads on the 1CPU host were 13% slower, but they
did not change significantly on the 2CPU host.

There were no other serious regressions or changes within error.

The full summary is at:

http://developer.osdl.org/judith/aio/result_compare_2.6.8...


Regards;

Judith Lebzelter
OSDL







-
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/


Copyright © 2004, Eklektix, Inc.
Comments and public postings are copyrighted by their creators.
Linux is a registered trademark of Linus Torvalds