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: Maxwell Lol on 18 Sep 2009 20:35 Keith Keller <kkeller-usenet(a)wombat.san-francisco.ca.us> writes: > Does anyone know of a centralized location where various benchmarks on > hard drives are posted/shared? I would google bonnie <DiskModel> and hope for the best.....
From: Rahul on 24 Sep 2009 21:08 Maxwell Lol <nospam(a)com.invalid> wrote in news:87zl8s8cmq.fsf(a)com.invalid: > 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. > I now ran bonnie++ but have trouble figuring out if my perf. stats are up to the mark or not. Any comments are very appreciated. Data posted below. Since this is an NFS store I ran bonnie++ from both the client and the server. Server side bonnie++ http://dl.getdropbox.com/u/118481/io_benchmarks/bonnie_op.html Client side bonnie++ http://dl.getdropbox.com/u/118481/io_benchmarks/bonnie_op_node25.html Caeveat: The cluster was in production so there is a chance of externalities affecting my data. (am trying it hard to explain why some stats seem better on the client run than the server run) Subsidary Goal: This setup had 23 clients for NFS. In a new cluster that I am setting up we want to scale this up about 250 clients. Hence want to estimate what sort of performance I'll be looking for in the Storage. (Other specs: Gigabit ethernet. RAID5 array of 5 total SAS 10k RPM disks. Total storage ~ 1.5 Terabyte; both server and client have 16GB RAM; Dell 6248 switches. Port bonding on client servers) -- Rahul
From: Maxwell Lol on 25 Sep 2009 23:18
Rahul <nospam(a)nospam.invalid> writes: > Maxwell Lol <nospam(a)com.invalid> wrote in news:87zl8s8cmq.fsf(a)com.invalid: > >> 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. >> > > I now ran bonnie++ but have trouble figuring out if my perf. stats are up > to the mark or not. Any comments are very appreciated. I was a sysadmin nearly 2 decades ago. So I cannot comment on current benchmarks. But ... > Server side bonnie++ > http://dl.getdropbox.com/u/118481/io_benchmarks/bonnie_op.html Block input is 243M/sec, output is 137M/sec > > Client side bonnie++ > http://dl.getdropbox.com/u/118481/io_benchmarks/bonnie_op_node25.html Block input is 21M/sec or about 1/10 the server's performance. Block output is 87M/sec. The output is 4 times faster than the input. I guess this makes sense, because the writing is cached, while the reading requires the disk block to be accessed (assuming the cache was emptied). NFS writing seems fast. NFS reading will depend on RAM and caching on the server. A benchmark won't tell you this. If you want to optimize NFS performance, there are the nfsstat values. This gives server and client stats. Sun published books years ago that talked about tuning NFS performance and other kinds.. It's by Adrian Cockcroft http://www.amazon.com/Sun-Performance-Tuning-Sparc-Solaris/dp/0131496425/ref=sr_1_4?ie=UTF8&s=books&qid=1253933881&sr=1-4 There is a newer edition of this book. I don't know if it pertains to Linux. I tried to find out the info he gave on nfsstat interpretation. I was not too successful. If I have time, I will look it up. Perhaps here are some suggestions: http://regions.cmg.org/regions/mspcmg/.../LISA_Sol_Linux_Perf.ppt http://www.slideshare.net/adrianco/solaris-linux-performance-tools-and-tuning http://regions.cmg.org/regions/mspcmg/.../LISA_Cap_Plan_Free.ppt I suppose the best thing to do is 1) Benchmark the different disks and compare. 2) keep an eye on the server and stats and do server tuning the best you can. (This is an art) |