<div dir="ltr">I'm doing 1 second samples and the rates are very steady. The reason I ended up at this level of testing was I had done a sustained test for 2 minutes at about 5MB/sec and was seeing over 500MB/sec going to the disk, again sampling at 1-second intervals. I'd be happy to provide detailed output and can even sample more frequently if you like.<div>
<br></div><div style>from my shorter test I was experimenting looking at some of the XFS data with collectl and recorded this, it if help at all:</div><div style><br></div><div style><div><font face="courier new, monospace">segerm@az1-sw-object-0006:~$ collectl --import xfs</font></div>
<div><font face="courier new, monospace">waiting for 1 second sample...</font></div><div><font face="courier new, monospace">#<--XFS Ops--><-----------XFS Logging----------><------Extents------><------DirOps-------><----Trans---><-</font></div>
<div><font face="courier new, monospace"># Write Reads Writes WrtKBs NoRoom Force Sleep ExtA BlkA ExtF ExtF Look Cre8 Remv Gdnt Sync Asyn Empt</font></div><div><font face="courier new, monospace"> 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0</font></div>
<div><font face="courier new, monospace"> 0 0 3 768 0 3 3 0 0 0 0 0 0 0 0 3 0 0</font></div><div><font face="courier new, monospace"> 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0</font></div>
<div><font face="courier new, monospace"> 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0</font></div><div><font face="courier new, monospace"> 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0</font></div>
<div><font face="courier new, monospace"> 0 0 1 256 0 1 1 0 0 0 0 0 0 0 0 1 0 0</font></div><div><font face="courier new, monospace"> 38 88 95 24320 0 95 96 54 54 54 54 115 76 76 154 95 473 0</font></div>
<div><font face="courier new, monospace"> 339 776 968 247816 0 968 978 484 484 479 479 1011 675 671 1351 967 4087 0</font></div><div><font face="courier new, monospace"> 321 748 929 237806 0 929 935 450 450 453 453 967 645 647 1287 930 3847 0</font></div>
<div><font face="courier new, monospace"> 279 637 810 207360 0 810 811 391 391 390 390 838 559 558 1118 810 3324 0</font></div><div><font face="courier new, monospace"> 209 482 610 156160 0 610 610 286 286 289 289 627 417 420 834 610 2451 0</font></div>
<div><font face="courier new, monospace"> 0 0 3 768 0 3 3 0 0 0 0 0 0 0 0 3 0 0</font></div><div><br></div><div style>I can say for a fact I was doing about 300 wtrites/sec whcih the write numbers seem to support, though I don't know what the read numbers are measuring. you can also see from the logging data that was 250MB/sec going to disk.</div>
<div style><br></div><div style>Are there other numbers that are meaningful that you'd like to see? All it takes is adding a couple of print statement as what you're seeing above only took a hour or so to throw to together.</div>
<div style><br></div><div style>I can collect as much (or little as you like) and actually can save the complete contents of /proc/fs/xfs/stat every second in a file for later playback.</div><div style><br></div><div style>
-mark</div></div><div style><br></div></div><div class="gmail_extra"><br><br><div class="gmail_quote">On Fri, Jun 14, 2013 at 8:17 PM, Nathan Scott <span dir="ltr"><<a href="mailto:nathans@redhat.com" target="_blank">nathans@redhat.com</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div class="im"><br>
<br>
----- Original Message -----<br>
> actually I have since found a decoder ring here -<br>
> <a href="http://xfs.org/index.php/Runtime_Stats" target="_blank">http://xfs.org/index.php/Runtime_Stats</a> and have been incorporating a lot of<br>
> the data so I can look at things in real time. I'd still love to know why<br>
<br>
</div>Ah, good stuff.<br>
<div class="im"><br>
> writing 1000 1K files results in 200MB/sec of disk I/O though. clearly<br>
<br>
</div>For how many seconds? (or fractions thereof?) You may need the level of<br>
detail that only tracing will provide to explore further, although off the<br>
top of my head I'm not sure exactly which events you should focus on (log,<br>
inode creation, or space allocation at a guess).<br>
<br>
cheers.<br>
<br>
--<br>
Nathan<br>
</blockquote></div><br></div>