From: Alexei_Roudnev (Alexei_Roudnev_at_exigengroup.com)
Date: Wed Apr 11 2007 - 20:03:22 CEST
Message-ID: <19f701c77c63$b1846980$6f31a8c0@sjc.exigengroup.com> From: "Alexei_Roudnev" <Alexei_Roudnev@exigengroup.com> Date: Wed, 11 Apr 2007 11:03:22 -0700 Subject: Re: [suse-sles-e] More benchmarks with SLES10 on HPProLiant DL360 G4p SCSI (was: catastrophic disk-performance on SLES10 with SATA drives)
As I said before, looks as something is wrong with the driver. Like - system
initiates IO but don't got 'IO is completed' response for a unusually long
time.
It don't influence local FS with wriuteback mode and long IO query (all IO
are passing thru in some time), but it can kill any performance on
NFS server (which is syncronous by design) and can explain high LA on the
server.
I'd better compare it all with other versions.
Btw, your LA is not consistant (2 tasks running, io wait == 20%, but LA ==
17 - strange). Can you run vmstat 5 and send output here (when you see LA ==
17), just as iostat -x 5 (send 1 or 2 last outputs).
----- Original Message -----
From: "Markus Roeffen" <markus.roeffen@iws.uni-stuttgart.de>
To: <suse-sles-e@suse.com>; <Alexei_Roudnev@exigengroup.com>;
<matilda@grandel.de>
Sent: Wednesday, April 11, 2007 7:25 AM
Subject: [suse-sles-e] More benchmarks with SLES10 on HPProLiant DL360 G4p
SCSI (was: catastrophic disk-performance on SLES10 with SATA drives)
Hello Andreas and Alexei,
I've tested the new filesystem options on the with two users and a
mounted nfs.
Each user has run this command:
time dd if=/dev/zero of=./bla.dat bs=160k count=10000
This was the output:
10000+0 records in
10000+0 records out
1638400000 bytes (1,6 GB) copied, 314,761 seconds, 5,2 MB/s
real 5m15.824s
user 0m0.013s
sys 0m7.244s
The load average on the server with two (!) users and only two (!)
running processes was quite ... (annoying? frustrating?) ... interesting:
> top - 16:12:51 up 9 days, 5:02, 4 users, load average: 17.57,
> 12.26, 6.76
> Tasks: 134 total, 2 running, 132 sleeping, 0 stopped, 0 zombie
> Cpu(s): 0.7%us, 6.6%sy, 0.0%ni, 67.5%id, 19.9%wa, 0.0%hi,
> 5.3%si, 0.0%st
> Mem: 1028988k total, 1019388k used, 9600k free, 4196k buffers
> Swap: 4192924k total, 140k used, 4192784k free, 528900k cached
I'll continue my research.
Best regards
Markus
-- ------------------------------- Markus Roeffen Institut für Wasserbau IWS Universität Stuttgart Pfaffenwaldring 61 70569 Stuttgart Tel: +49(0)711/685-67010 Fax: +49(0)711/685-67020 Mail: markus.roeffen@iws.uni-stuttgart.de --------------------------------------------------------------------- To unsubscribe, e-mail: suse-sles-e-unsubscribe@suse.com For additional commands, e-mail: suse-sles-e-help@suse.com
This archive was generated by hypermail 2.1.7 : Wed Apr 11 2007 - 22:09:50 CEST