<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=Windows-1252">
</head>
<body style="word-wrap: break-word; -webkit-nbsp-mode: space; -webkit-line-break: after-white-space;">
<div>
<div style="margin: 0px; font-size: 14px; font-family: Monaco; color: rgb(245, 245, 245); background-color: rgb(0, 0, 0); position: static; z-index: auto;">
ior/bin/IOR.mpiio.mvapich2-2.0b -h</div>
</div>
<div><br>
</div>
<div>
<div style="margin: 0px; font-size: 14px; font-family: Monaco; color: rgb(245, 245, 245); background-color: rgb(0, 0, 0); position: static; z-index: auto;">
<span style="color: #000000; background-color: #f5f5f5">-t</span> N transferSize -- size of transfer in bytes (e.g.: 8, 4k, 2m, 1g)</div>
</div>
<div><br>
</div>
<div>IOR reports it in the log :</div>
<div><br>
</div>
<div>
<div style="margin: 0px; font-size: 14px; font-family: Monaco; color: rgb(245, 245, 245); background-color: rgb(0, 0, 0); position: static; z-index: auto;">
Command line used: /home/aik/lustre/benchmark/git/ior/bin/IOR.mpiio.mvapich2-2.0b -v -a MPIIO -i5 -g -e -w -r -b 16g -C -t 8k -o /mnt/lfs/admin/iotest/ior/stripe_2/ior-testfile.ssf</div>
</div>
<div style="margin: 0px; font-size: 14px; font-family: Monaco; color: rgb(245, 245, 245); background-color: rgb(0, 0, 0); position: static; z-index: auto;">
...</div>
<div>
<div style="margin: 0px; font-size: 14px; font-family: Monaco; color: rgb(245, 245, 245); background-color: rgb(0, 0, 0); position: static; z-index: auto;">
Summary:</div>
<div style="margin: 0px; font-size: 14px; font-family: Monaco; color: rgb(245, 245, 245); background-color: rgb(0, 0, 0); position: static; z-index: auto;">
<br>
</div>
<div style="margin: 0px; font-size: 14px; font-family: Monaco; color: rgb(245, 245, 245); background-color: rgb(0, 0, 0); position: static; z-index: auto;">
api = MPIIO (version=3, subversion=0)</div>
<div style="margin: 0px; font-size: 14px; font-family: Monaco; color: rgb(245, 245, 245); background-color: rgb(0, 0, 0);">
test filename = /mnt/lfs/admin/iotest/ior/stripe_2/ior-testfile.ssf</div>
<div style="margin: 0px; font-size: 14px; font-family: Monaco; color: rgb(245, 245, 245); background-color: rgb(0, 0, 0);">
access = single-shared-file, independent</div>
<div style="margin: 0px; font-size: 14px; font-family: Monaco; color: rgb(245, 245, 245); background-color: rgb(0, 0, 0);">
pattern = segmented (1 segment)</div>
<div style="margin: 0px; font-size: 14px; font-family: Monaco; color: rgb(245, 245, 245); background-color: rgb(0, 0, 0);">
ordering in a file = sequential offsets</div>
<div style="margin: 0px; font-size: 14px; font-family: Monaco; color: rgb(245, 245, 245); background-color: rgb(0, 0, 0);">
ordering inter file=constant task offsets = 1</div>
<div style="margin: 0px; font-size: 14px; font-family: Monaco; color: rgb(245, 245, 245); background-color: rgb(0, 0, 0);">
clients = 32 (8 per node)</div>
<div style="margin: 0px; font-size: 14px; font-family: Monaco; color: rgb(245, 245, 245); background-color: rgb(0, 0, 0);">
repetitions = 5</div>
<div style="margin: 0px; font-size: 14px; font-family: Monaco; color: rgb(245, 245, 245); background-color: rgb(0, 0, 0);">
xfersize = 8192 bytes</div>
<div style="margin: 0px; font-size: 14px; font-family: Monaco; color: rgb(245, 245, 245); background-color: rgb(0, 0, 0);">
blocksize = 16 GiB</div>
<div style="margin: 0px; font-size: 14px; font-family: Monaco; color: rgb(245, 245, 245); background-color: rgb(0, 0, 0); position: static; z-index: auto;">
aggregate filesize = 512 GiB</div>
</div>
<div><br>
</div>
<div>Here we have xfersize 8k, each client of 32 writes 16GB, so the aggregate file size is 512GB.</div>
<div><br>
</div>
<div>I would expect records size to be ~1MB for our workloads.</div>
<div><br>
</div>
<div>Best regards, Alex.</div>
<br>
<div>
<div>On Apr 29, 2015, at 11:07 AM, Scott Nolin <<a href="mailto:scott.nolin@ssec.wisc.edu">scott.nolin@ssec.wisc.edu</a>> wrote:</div>
<br class="Apple-interchange-newline">
<blockquote type="cite">Ok I looked up my notes.<br>
<br>
I'm not really sure what you mean by record size. I assumed when I do a file per process the block size = file size. And that's what I see dropped on the filesystem.<br>
<br>
I did -F -b <size><br>
<br>
With block sizes 1MB, 20MB, 100MB, 200MB, 500MB<br>
<br>
2, 4, 8, 16 threads on 1 to 4 clients.<br>
<br>
I assumed 2 threads on 1 client looks a lot like a client writing or reading 2 files. I didn't bother looking at 1 thread.<br>
<br>
Later I just started doing 100MB tests since it's a very common file size for us. Plus I didn't see real big difference once size gets bigger than that.<br>
<br>
Scott<br>
<br>
<br>
On 4/29/2015 10:24 AM, Alexander I Kulyavtsev wrote:<br>
<blockquote type="cite">What range of record sizes did you use for IOR? This is more important<br>
than file size.<br>
100MB is small, overall data size (# of files) shall be twice as memory.<br>
I ran series of test for small record size for raidz2 10+2; will re-run<br>
some tests after upgrading to 0.6.4.1 .<br>
<br>
Single file performance differs substantially from file per process.<br>
<br>
Alex.<br>
<br>
On Apr 29, 2015, at 9:38 AM, Scott Nolin <<a href="mailto:scott.nolin@ssec.wisc.edu">scott.nolin@ssec.wisc.edu</a><br>
<<a href="mailto:scott.nolin@ssec.wisc.edu">mailto:scott.nolin@ssec.wisc.edu</a>>> wrote:<br>
<br>
<blockquote type="cite">I used IOR, singlefile, 100MB files. That's the most important<br>
workload for us. I tried several different file sizes, but 100MB<br>
seemed a reasonable compromise for what I see the most. We rarely or<br>
never do file striping.<br>
<br>
I remember I did see a difference between 10+2 and 8+2. Especially at<br>
smaller numbers of clients and threads, the 8+2 performance numbers<br>
were more consistent, made a smoother curve. 10+2 with not a lot of<br>
threads the performance was more variable.<br>
</blockquote>
<br>
</blockquote>
<br>
<br>
_______________________________________________<br>
lustre-discuss mailing list<br>
<a href="mailto:lustre-discuss@lists.lustre.org">lustre-discuss@lists.lustre.org</a><br>
http://lists.lustre.org/listinfo.cgi/lustre-discuss-lustre.org<br>
</blockquote>
</div>
<br>
</body>
</html>