> No, that's indeed what happens on most supers I've seen. Even if some<BR>
> codes are capable of running to the (almost) full scale of the<BR>
> machine, such large runs are quite marginal. They often take place<BR>
> before the system enters production, to demonstrate its capabilities,<BR>
> > Along similar lines, has Google or Amazon ever published their batch job<BR>
> > capacity (it must be in petaflops...)<BR>
> That means high latencies and grumpy MPI. <BR>
Grumpy MPI. Now there's a name for a new open source project, if I have ever seen one.<BR>
Let's start it now - a new MPI implementation which is intended for clouds and geographically<BR>
separated machines with high latencies between them. The MPI implementation will cope with <BR>
Network interruptions of a long duration, but will of course produce copious grumpy error messages.<BR>
And we can have mailing lists which are famed across the Internet for their spikiness and flame wars.<BR>
I vote we invite Victor Meldrew to be the project's new figurehead.<BR>
<a href="http://en.wikipedia.org/wiki/Victor_Meldrew" target="_blank">http://en.wikipedia.org/wiki/Victor_Meldrew</a> <BR>
The contents of this e-mail are confidential and for the exclusive use of the intended recipient.
If you are not the intended recipient you should not read, copy, retransmit or disclose its contents.
If you have received this email in error please delete it from your system immediately and notify us either by email or telephone.
The views expressed in this communication may not necessarily be the views held by McLaren Racing Limited.
<br />McLaren Racing Limited | McLaren Technology Centre | Chertsey Road | Woking | Surrey | GU21 4YH | UK | Company Number: 01517478