Encouraging results from 2.4.10-pre14 + preempt

Encouraging results from 2.4.10-pre14 + preempt

[J Sloan]

I am very encouraged by the latest linus kernel, and
thought I'd pass this along -

Platform info -
----------------
PIII 933/Intel mobo
512 MB RAM
2 30 GB IDE disks
Ensoniq ES1371 sound card
Red Hat 7.1 + bits of rawhide
kernel 2.4.10-pre14 + preempt patch

The testing was done while running X windows (icewm), Netscape 4.78,
iptraf, and misc other services (named, sendmail, apache, iptables
firewalling etc.

During the dbench runs mpg123 was playing mp3s -

Highlights -
------------
 - There was absolutely no mp3 skipping during dbench 16 & 32.
 - There was minor skipping during dbench 40, when load was near 40.
 - Best of all, the desktop remained completely responsive the entire
time.


dbench summaries:
---------------------------
Throughput 49.6162 MB/sec (NB=62.0202 MB/sec  496.162 MBit/sec)  16
procs
Throughput 32.5531 MB/sec (NB=40.6913 MB/sec  325.531 MBit/sec)  32
procs
Throughput 24.8711 MB/sec (NB=31.0889 MB/sec  248.711 MBit/sec)  40
procs


Screenshot at http://mirai.cx/dbench40.png

Re: Encouraging results from 2.4.10-pre14 + preempt

[Robert Love]

On Sat, 2001-09-22 at 20:38, J Sloan wrote:
> I am very encouraged by the latest linus kernel, and
> thought I'd pass this along -

I appreciate this.

> Highlights -
> ------------
>  - There was absolutely no mp3 skipping during dbench 16 & 32.
>  - There was minor skipping during dbench 40, when load was near 40.
>  - Best of all, the desktop remained completely responsive the entire
> time.

This is great.  While I am sure all three of this points wouldn't be
true without preempt patch, I would be interested in seeing your opinion
of the comparison between with and without, if you had the time.  Also
comparing the dbench would be nice, but you must be careful to replicate
the system similarly.

Thanks much for the feedback,

-- 
Robert M. Love
rml at ufl.edu
rml at tech9.net