<div>thank you for your detailed analysis. your guess is right .my outdir is set to a place,not local to the nodes.i know the reason of my slow job,although i did not know it through. <br>as for how to use a local directory as outdir ,it seems another new thing for me.
</div>
<div>would it be possible to tell me about that ?</div>
<div>thank you in advance <br> </div>
<div> </div>
<div> </div>
<div>your sincerely </div>
<div> xuyuehua</div>
<div> </div>
<div><span class="gmail_quote">2007/12/13, Lorenzo Paulatto <<a href="mailto:paulatto@sissa.it">paulatto@sissa.it</a>>:</span>
<blockquote class="gmail_quote" style="PADDING-LEFT: 1ex; MARGIN: 0px 0px 0px 0.8ex; BORDER-LEFT: #ccc 1px solid"><br>On Thu, December 13, 2007 13:42, xu yuehua wrote:<br>> yes,maybe once i set disk_io="none" ,when something goes wrong ,like
<br>> power<br>> is off suddenly,you have to start from scratch .from this point ,it is<br>> wasting time in deed.<br>> but is there an better idea ?<br>> i am wondering<br><br>I think it is highly unlikely that you have more RAM than disk space, so
<br>disk space should be no issue. Furthermore it should not be so slow, as<br>long as you are leaving wfc_collect=.false. which is the default.<br><br>From these assumptions I can guess that you are using a network share as
<br>temporary directory for your calculation; in other words your outdir is<br>set to some place, probably in you home directory, which is not local to<br>nodes but stays on a server and is shared via NFS (network file system).
<br><br>If this is your case, every time wavefunctions are written the data has to<br>pass trough the network, it is somehow equivalent to collect_wfc=.true.<br>but done externally by the operating system. Furthermore if you are using
<br>the same network for NFS and mpi when the traffic from NFS suddenly bursts<br>it is likely to conflict with mpi and slow it down a lot.<br><br>To avoid this problem you must use a local directory as outdir (like<br>/tmp), but you to have to remember that if you want to restart a job you
<br>have to use exactly the same node you used before in order for them to<br>find their wavefunctions. Otherwise you may have to recollect the data by<br>hand, which is probably not trivial (I have never tried).<br><br>I hope what I wrote was not already known to you, goodbye.
<br>--<br>Lorenzo Paulatto<br>+39 040 3787 511<br><a href="http://people.sissa.it/~paulatto/">http://people.sissa.it/~paulatto/</a><br><br><br>----------------------------------------------------------------<br>SISSA Webmail
<a href="https://webmail.sissa.it/">https://webmail.sissa.it/</a><br>Powered by SquirrelMail <a href="http://www.squirrelmail.org/">http://www.squirrelmail.org/</a><br><br>_______________________________________________<br>
Pw_forum mailing list<br><a href="mailto:Pw_forum@pwscf.org">Pw_forum@pwscf.org</a><br><a href="http://www.democritos.it/mailman/listinfo/pw_forum">http://www.democritos.it/mailman/listinfo/pw_forum</a><br></blockquote></div>
<br><br clear="all"><br>-- <br>Xu Yuehua<br>physics Department of Nanjing university<br>China