Date: Wed, 5 Sep 2001 07:15:58 -0700 From: Doug Hardie <bc979@lafn.org> To: "Lawrence Farr" <l.farr@epcdirect.co.uk> Cc: "'Greg Lehey'" <grog@FreeBSD.ORG>, "'Lawrence Farr'" <lawrence@epcdirect.co.uk>, "'David Gilbert'" <dgilbert@velocet.ca>, "'Chris BeHanna'" <behanna@zbzoom.net>, "'FreeBSD-Stable'" <stable@FreeBSD.ORG> Subject: RE: [stable] Re: RAID5 Message-ID: <f04330116b7bbe195e610@[10.0.1.100]> In-Reply-To: <002c01c135e4$69c924d0$c80aa8c0@lfarr> References: <002c01c135e4$69c924d0$c80aa8c0@lfarr>
next in thread | previous in thread | raw e-mail | index | archive | help
At 9:26 +0100 9/5/01, Lawrence Farr wrote: >I did these tests in response to a recent thread on hardware (look for >"3ware stuff not ready for heavy duty useage"). > > >On Tuesday, 4 September 2001 at 13:52:02 +0100, Lawrence Farr wrote: >>> Just to add another benchmark, I got: >>> >>> Pass 23 - 1048576 kb written in 115 seconds, at 9118 kb/Sec >>> Pass 23 - 1048576 kb read in 15 seconds, at 69905 kb/Sec >> >>This shows there's a big difference. Which is which? What is it > >really doing here? Perhaps some experiences I had with several years running large RAID 5 systems may be of help. I used a RAID 5 configuration where each bay had 7 disks for striping and parity data. The 8th disk was a fast-write/recovery disk. If all the drives were working correctly then writes to the system were first written to the fast-write drive and the I/O terminated. The RAID system then distributed the information to the 7 data disks. If the fast-write disk was not full, then this ran a little faster than read speed. However, we frequently did large Oracle tablespace initializations. Generally all 7 disks were used for the tablespace. We were able to monitor disk usage on all the drives plus the line from the computer to the RAID system. During the first few minutes, all the drives (fast-write and data) would be running a 100%. The line from the computer to the RAID system was operating at very close to 100%. However, after just short of 2 GB transfered (the size of the fast-write disk), the line to the RAID dropped to zero transfers. It would stay at zero for about 10 minutes. Then a few blocks would be transferred and back to sleep. It continued like this till the tablespace was initialized. The average line utilization was very low. The disk utilization was at 100%. Writes to the fast-write disk are fast. Distributing the information the RAID 5 way to multiple disks is slower and hence the RAID system bottlenecked and quit accepting data from the host. It had no place to put it until the fast-write disk had been processed. While the information above is not complete enough to say for sure, it looks like thats what occurred in that test. -- -- Doug To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-stable" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?f04330116b7bbe195e610>