Calculating IOPS and bandwidth for a disk / disk array
in [General Linux]
|
Prev: max international, mlm business opportunity, MaxGXL, network marketing opportunity , Max GXL, MaxWLX, Max N-fuze, Max WLX,
Next: Finding all cron jobs...
From: Rahul on 17 Sep 2009 17:54 I have disk specs on my 15K SAS drives that I am going to put in a RAID5 box with 8 drives. I think I know how to get IOPS but how do I get a effective throughput number? IOPS = 1000 / (average read seek in millisec + latency) [Correct me if I am messing it up] But how do I get Mbits/sec numbers now? Someone said I multiply IOPS by 512 since that how big a block is but the reasoning seems a little shaky to me. I need the Mbits/sec since that allows me to estimate what the size of the links ought to be. -- Rahul
From: Chris Cox on 17 Sep 2009 18:02 On Thu, 2009-09-17 at 21:54 +0000, Rahul wrote: > I have disk specs on my 15K SAS drives that I am going to put in a RAID5 > box with 8 drives. > > I think I know how to get IOPS but how do I get a effective throughput > number? > > IOPS = 1000 / (average read seek in millisec + latency) > > [Correct me if I am messing it up] > > But how do I get Mbits/sec numbers now? Someone said I multiply IOPS by 512 > since that how big a block is but the reasoning seems a little shaky to me. They are wrong. The only way to get this is by doing a block transfer test. You could use bonnie++ or a time transfer of something more than twice as big as your memory. I'd use bonnie++ though. Then you get reads, writes, etc. > > I need the Mbits/sec since that allows me to estimate what the size of the > links ought to be. >
From: Rahul on 17 Sep 2009 18:22 Chris Cox <chrisncoxn(a)endlessnow.com> wrote in news:1253224971.29662.177.camel(a)geeko: > They are wrong. The only way to get this is by doing a block transfer > test. You could use bonnie++ or a time transfer of something more than > twice as big as your memory. I'd use bonnie++ though. Then you get > reads, writes, etc. > Thanks! But how does one go about buying a disk? I looked at some SAS disk spec sheets on the web and they only have seek times and latency. If I have a requirement in mind how do I evaluate what this particular disk can or cannot generate for IOPS and MB/sec? -- Rahul
From: Maxwell Lol on 18 Sep 2009 12:02 Rahul <nospam(a)nospam.invalid> writes: > Chris Cox <chrisncoxn(a)endlessnow.com> wrote in > news:1253224971.29662.177.camel(a)geeko: > >> They are wrong. The only way to get this is by doing a block transfer >> test. You could use bonnie++ or a time transfer of something more than >> twice as big as your memory. I'd use bonnie++ though. Then you get >> reads, writes, etc. >> > > Thanks! But how does one go about buying a disk? I looked at some SAS disk > spec sheets on the web and they only have seek times and latency. > > If I have a requirement in mind how do I evaluate what this particular disk > can or cannot generate for IOPS and MB/sec? Like they said - Run the bonnie benchmark. I haven't used the bonnie++ benchmark, but I'ved used bonnie a lot. Recently I compiled it under cygwin, and ran it on a Windows box using an encrypted disk. In bonnie, You do have to specify a file size that is larger than the largest memorty buffer. Otherwise you are measuring the cache. Here's a sad truth. marketeers lie. Specs lie. You can't trust them. If you are luckly, someone else has done a suitable benchmark for you. If not, do it yourself.
From: Keith Keller on 18 Sep 2009 13:25
On 2009-09-18, Maxwell Lol <nospam(a)com.invalid> wrote: > > Here's a sad truth. marketeers lie. Specs lie. You can't trust them. > If you are luckly, someone else has done a suitable benchmark for you. > If not, do it yourself. In other words, you need to buy the disk and benchmark it before deciding which disk to buy. ;-/ I'm not disagreeing with you--it's the only way to fairly compare disks to each other. But not all organizations have the resources to go out and purchase a bunch of hard disks just for benchmarking purposes. Does anyone know of a centralized location where various benchmarks on hard drives are posted/shared? Obviously these too need to be taken with some grains of salt (not everyone tunes their configuration appropriately, vendors might poison the site, benchmarks might be fudged and/or fabricated. But it would be better than nothing. --keith -- kkeller-usenet(a)wombat.san-francisco.ca.us (try just my userid to email me) AOLSFAQ=http://www.therockgarden.ca/aolsfaq.txt see X- headers for PGP signature information |