On 4/10/06, Peter Tribble <P.Tribble at herts.ac.uk> wrote:
> System 1: SunBlade 200, twin 1.015GHz processors, single 72G disk.
My test:
$ uname -i; psrinfo -v
SUNW,Sun-Blade-1500
Status of virtual processor 0 as of: 04/10/2006 19:28:41
on-line since 02/28/2006 09:36:56.
The sparcv9 processor operates at 1062 MHz,
and has a sparcv9 floating point processor.
> The limiting factor in current installs appears to be the cpu
> required to bzcat the archives. It took the test machine 2 minutes
> to unpack the staroffice archive - going at 2M/s which is the
> network data rate I observed during install. That's about 1/6 of
> the total data - so 12 minutes of the 22 is bzcat. That's the
> lowest hanging fruit. Even when you've solved that there's still
> 10 minutes left - and only 4 of those appear to be down to the
> contents file. There is of course the actual time it takes to write
> the data to disk - and, being scattered around in small files,
> those writes won't be as efficient as the contents file.
I recreated something that is close to the SUNWcsr.bz file with:
$ grep -w SUNWcsr /var/sadm/install/contents | nawk '{print $1}' \
| cut -d= -f1 | cpio -o > SUNWcsr.cpio
I also bzip2'd it. The resulting files:
$ du -h /tmp/SUNWcsr.cpio*
20M /tmp/SUNWcsr.cpio
8.3M /tmp/SUNWcsr.cpio.bz2
Taking a look at the time to bunzip the file...
$ time bzcat < /tmp/SUNWcsr.cpio.bz2 > /dev/null
real 0m5.938s
user 0m5.060s
sys 0m0.050s
So, 5.11 seconds. This implies that the CPU limts NFS reads to 1.62
MB/s from NFS and 3.9 MB/s writes to disk. Both should be able to do
better.
I have brought this issue up before, but buried with other stuff.
Look for "chunks" at
http://www.opensolaris.org/jive/thread.jspa?messageID=29919瓭
This will only grow as a problem as data continues to grow and
per-core performance stagnates.
Mike
--
Mike Gerdts
http://mgerdts.blogspot.com/
