Date: Sat, 4 Feb 2006 21:31:51 -0400 (AST) From: "Marc G. Fournier" <scrappy@hub.org> To: Ivan Voras <ivoras@fer.hr> Cc: stable@freebsd.org, freebsd-proliant@freebsd.org Subject: Re: Compaq Proliant CISS slow writes Message-ID: <20060204213028.D3207@ganymede.hub.org> In-Reply-To: <43E52A89.4020700@fer.hr> References: <43E52A89.4020700@fer.hr>
next in thread | previous in thread | raw e-mail | index | archive | help
Not sure if this is standard, but on an idle machine (ie. just set it up,
nothing running on it yet), with RAID1+0 across 4 drives:
Writing the 853 Megabyte file, 'iozone.tmp'...15.656250 seconds
Reading the file...3.921875 seconds
IOZONE performance measurements:
57129601 bytes/second for writing the file
228063191 bytes/second for reading the file
The server is a new DL360 G4p, drives are SAS ...
On Sat, 4 Feb 2006, Ivan Voras wrote:
> I need to get a Proliant machine with 2 P3 processors running FreeBSD 6. I
> don't know much about the machine, I think it's ML 380 G2 or close to that,
> but I have physical access. So far, everything is fine (once the inability to
> boot from CD-ROM is circumvented), except one "detail": horrible write
> performance on its CISS 5 RAID5 array. I get ~75MB/s burst (large blocks)
> reads, and only 5MB/s burst writes. I know how RAID5 works, but still, this
> is bad. The machine has been running Linux before this and performance was Ok
> - I didn't benchmark it but the "feeling" when working on it was normal,
> while on FreeBSD it's noticably slow in mixed read/write load.
>
> Is there anything I can try to improve this? In the verbose boot log there's
> a line that says the controller supports "simple, performant and MEMQ" modes,
> and the one that's used is "simple" - does this have any influence? If so,
> how to change it?
>
> Thanks!
>
----
Marc G. Fournier Hub.Org Networking Services (http://www.hub.org)
Email: scrappy@hub.org Yahoo!: yscrappy ICQ: 7615664
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20060204213028.D3207>
