<html><head><style type='text/css'>p { margin: 0; }</style></head><body><div style='font-family: Arial; font-size: 12pt; color: #000000'><P>John/All,</P>
<P> </P>
<P>It is a useful reminder I guess, but I have to assume that this is</P>
<P>something that folks on this list are familiar with, No? State is</P>
<P>more complicated for parallel work and less deterministic. As</P>
<P>these qualities accumulate in the processing of any workload</P>
<P>a stochastic outcome (timings in this case, but it can also effect</P>
<P>precision as anyone that has compare vector to scale results</P>
<P>remembers) is the result.</P>
<P> </P>
<P>This effect is another manifestation of the behavior demonstrated</P>
<P>by Galton in the late 1800s when he dropped English pence</P>
<P>through a pinned board (the pins are the identical cores, the pence</P>
<P>parrallel processes) ... he got a gaussian distribution. As the article </P>
<P>suggests you can reduce the variance by limiting the non-determinism</P>
<P>in process and that is what eXludus (a Montreal based software scheduling</P>
<P>company) is doing at the job and process level. As the number of</P>
<P>of cores and/or number of processes (virtual or otherwise) grows</P>
<P>so does the variance in outcomes. This is a on-die manifestion of</P>
<P>job skew is it not, and another of the second law of thermodynamics ...</P>
<P> </P>
<P>rbw<BR><BR><BR>----- Original Message -----<BR>From: "John Hearns" <hearnsj@googlemail.com><BR>To: beowulf@beowulf.org<BR>Sent: Tuesday, December 23, 2008 6:54:39 AM GMT -05:00 US/Canada Eastern<BR>Subject: [Beowulf] Not all cores are created equal<BR><BR></P>
<DIV>I'm surprised this has not been flagged up yet. Shamelessly passed on from Slashdot:</DIV>
<DIV><A href="http://www.gcn.com/online/vol1_no1/47765-1.html" target=_blank>http://www.gcn.com/online/vol1_no1/47765-1.html</A></DIV><BR>_______________________________________________ Beowulf mailing list, Beowulf@beowulf.org To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf </div></body></html>