[Beowulf] fast and resilient scatter-gather in the cloud

[INKozin]

What do people use in the cloud (e.g. AWS) for low latency scatter-gather
operations as a closest but resilient alternative to MPI?
The amount of data in question is about 100 MB so we are looking at ~ 1 sec
over 10 Gbps. Assuming 20 cores per node running a threaded code, we will
need 20-40 nodes. Ideally the overall runtime should be ~ 10 sec or even
less.

Thanks for your thoughts and suggestions!

Igor
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.beowulf.org/pipermail/beowulf/attachments/20151013/82a60464/attachment.html>