https://jira.whamcloud.com This file is an XML representation of an issue en-us 9.4.14 940014 05-12-2023 https://jira.whamcloud.com/browse/LU-3769 Lustre <p>server: lustre-2.4.0-2.6.32_358.11.1.el6_lustre.gef18aa1.x86_64_ga5e72ed.x86_64 on CENTOS 6.4<br/> client: lustre-client-2.4.52-3.9.9_302.fc19.x86_64_gfdd4844.x86_64 on FC19</p> <p>the client is mounted with flock parameter:<br/> 192.168.2.104@o2ib:/lustrefs on /lustre type lustre (rw,flock)</p> <p>I have run some micro-benchmark like mdstat/iozone/IOR without problem with good results.<br/> Customer asks me to run compilebench benchmark. I have downloaded the compilebench rpm from <a href="http://build.whamcloud.com/job/toolkit/" class="external-link" target="_blank" rel="nofollow noopener">http://build.whamcloud.com/job/toolkit/</a></p> LU-3769

compile bench real slow

Bug Minor Closed Won't Fix WC Triage Gabriele Paciucci Fri, 16 Aug 2013 07:06:23 +0000 Mon, 2 Sep 2013 10:45:12 +0000 Mon, 2 Sep 2013 10:45:04 +0000 Lustre 2.4.0 0 2 <p>I have set this parameters on the client:<br/> lctl set_param osc.*.checksums=0<br/> lctl set_param osc.*.max_rpcs_in_flight=256</p> <p>Compilebench results against a local disk:<br/> compilebench -D /tmp/ -i 2 -r 2 --makej<br/> using working directory /tmp/, 2 intial dirs 2 runs<br/> native unpatched native-0 222MB in 1.00 seconds (222.37 MB/s)<br/> native patched native-0 109MB in 0.31 seconds (353.78 MB/s)<br/> native patched compiled native-0 691MB in 0.51 seconds (1356.07 MB/s)<br/> create dir kernel-0 222MB in 3.06 seconds (72.67 MB/s)<br/> create dir kernel-1 222MB in 3.14 seconds (70.82 MB/s)<br/> compile dir kernel-1 680MB in 0.89 seconds (764.76 MB/s)<br/> compile dir kernel-0 680MB in 0.61 seconds (1115.79 MB/s)<br/> read dir kernel-1 in 21.46 52.44 MB/s<br/> read dir kernel-0 in 20.26 55.55 MB/s<br/> read dir kernel-1 in 0.66 1705.12 MB/s<br/> delete kernel-1 in 1.36 seconds<br/> delete kernel-0 in 1.04 seconds</p> <p>run complete:<br/> ==========================================================================<br/> intial create total runs 2 avg 71.75 MB/s (user 0.40s sys 0.79s)<br/> no runs for create<br/> no runs for patch<br/> compile total runs 2 avg 940.28 MB/s (user 0.08s sys 0.50s)<br/> no runs for clean<br/> no runs for read tree<br/> read compiled tree total runs 3 avg 604.37 MB/s (user 0.56s sys 1.16s)<br/> no runs for delete tree<br/> delete compiled tree total runs 2 avg 1.20 seconds (user 0.18s sys 0.35s)<br/> no runs for stat tree<br/> no runs for stat compiled tree</p> <p>Compilebench results on the same machine against the lustre fs:<br/> compilebench -D /lustre -i 2 -r 2 --makej<br/> using working directory /lustre/GABRIELE/, 2 intial dirs 2 runs<br/> native unpatched native-0 222MB in 37.65 seconds (5.91 MB/s)<br/> native patched native-0 109MB in 10.89 seconds (10.07 MB/s)<br/> native patched compiled native-0 691MB in 5.40 seconds (128.07 MB/s)<br/> create dir kernel-0 222MB in 36.36 seconds (6.12 MB/s)<br/> create dir kernel-1 222MB in 37.58 seconds (5.92 MB/s)<br/> compile dir kernel-1 680MB in 6.94 seconds (98.07 MB/s)<br/> compile dir kernel-0 680MB in 6.61 seconds (102.97 MB/s)<br/> read dir kernel-1 in 32.43 27.84 MB/s<br/> read dir kernel-0 in 31.67 28.51 MB/s<br/> read dir kernel-1 in 3.96 228.03 MB/s</p> <p>the delete process was too long to complete....</p> <p>So is it slow compared to before, or is it slo because it was always slow on lustre (and lustre is not very good at these workloads).</p> <p>(note we run compile bench regularly in testing so you can track history of performance on those).</p> <p>Hi Oleg,<br/> I have no experience with compilebench... Could you please give some suggestions how to run in the right way this tool? I have found benefits avoiding sync and drop cache (-n).</p> <p>You probably need to ask somebody like Yu Jian for advice on this.</p> <p>Thanks Oleg</p> Development Epic/Theme Performance Rank 1|hzvy3r: Rank (Obsolete) 9709 Severity