> > (we can kill atime, mke2fs with fancy options or use reiserfs, add the
> > Beowulf hack for multiple Ethernet cards, hack the scheduler, etc.)
>
> changing web serving software, run reverse squid proxies and similar
> should also be on the table for consideration.
>
> -- craig
I did some bechmarks between a straight Debian slink machine
(glibc-2.0) running 2.2.5 and a glibc-2.1 machine running
2.2.7+NVSv3+Channel Bonding and the vanilla Debian machine provided 10x
better performance according to specweb.
This is with both the document root and the logfiles on NFS mounts from a
NetApp box and running Zeus web server.
There seemed to be some network bottleneck with the juiced up machine, CPU
load was <1 and the performance is best described as "bursty".
I have a test machine available that will be running when I can get power
to it that should allow me to find out exactly what the problem was.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/