Date: Thu, 28 Mar 2002 15:26:44 -0700 From: "Kenneth D. Merry" <ken@kdm.org> To: Lukas Ertl <l.ertl@univie.ac.at> Cc: freebsd-scsi@FreeBSD.ORG Subject: Re: ciss driver and tagged queuing Message-ID: <20020328152644.A23074@panzer.kdm.org> In-Reply-To: <20020328214454.B214-100000@korben.in.tern>; from l.ertl@univie.ac.at on Thu, Mar 28, 2002 at 09:54:50PM %2B0100 References: <20020328134207.A22421@panzer.kdm.org> <20020328214454.B214-100000@korben.in.tern>
next in thread | previous in thread | raw e-mail | index | archive | help
On Thu, Mar 28, 2002 at 21:54:50 +0100, Lukas Ertl wrote: > On Thu, 28 Mar 2002, Kenneth D. Merry wrote: > > > > I have a brand new Compaq Proliant box here, which has a Smart Array 5302 > > > controller (ciss driver). I did some performance testing on it (with > > > bonnie++), but I wasn't quite satisfied, the throughput on the Smart Array > > > is rather low. So I thought I have a look with camcontrol on it: > > > > What were your performance numbers, anyway? > > If you are interested, you can find some figures at > <http://mailbox.univie.ac.at/~le/bonnie.html>. I tested two Compaq > Proliant boxes, one with a Smart Array Controller and a Smart Array, one > with a Dual Channel Host Adapter (with Adaptec Chip) and a SCSI/IDE Raid > Subsystem (speaks SCSI on the outside, has IDE disks inside). The SCSI/IDE > Raid figures are often significantly higher. If you want to know more > details, just mail me. What RAID level are you using? There are significant differences (on the smart array controller at least) between the block read and write results. With the SCSI->IDE controller, oddly, the write results are faster than the read results. > > I would suggest using the 'tags' camcontrol command to see how many tag > > openings there are for the device, that might be more informative. > > Ok: > > [root@raidtest /home/le]# camcontrol tags da1 -v > (pass1:ciss1:0:0:0): dev_openings 1 > (pass1:ciss1:0:0:0): dev_active 0 > (pass1:ciss1:0:0:0): devq_openings 1 > (pass1:ciss1:0:0:0): devq_queued 0 > (pass1:ciss1:0:0:0): held 0 > (pass1:ciss1:0:0:0): mintags 2 > (pass1:ciss1:0:0:0): maxtags 255 Yeah, you can only have one tag outstanding to the array. So I would expect that performance might not be as good as if you could have multiple commands outstanding. What does dmesg say about the card? From looking at the driver, it looks like the "max outstanding commands" value that is printed out on boot will tell us how many outstanding transactions the card claims to support. > > > Apart from that: the ciss driver is fairly new AFAIK, I didn't find a lot > > > of references to it on the web. Would you recommend it for production use? > > > Is there any testing going for userland tools (status control a.s.o.)? > > > > Mike Smith <msmith@FreeBSD.org> would be the one to ask. > > Is he on this list? I think so. Ken -- Kenneth Merry ken@kdm.org To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-scsi" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20020328152644.A23074>