Ask Your Question

How to use all cores available in the cluster to a single VM??

asked 2017-02-28 09:42:45 -0500

vathanlal gravatar image

updated 2017-02-28 09:51:56 -0500


I have an OpenStack newton cluster and Iam interested in Testing a MPI application in this cluster. In my cluster I have 3 compute nodes each with 64 cores thus total 192 cores. I have one instance with a MPI application. How can I run this instance with MPI application using all this 192 cores?? Can I create a flavor with 192 cores and allocate that flavor to my MPI application instance and run it?? Whether this will work?? or is there any other solution so that I can run a single instance using all the cores available in the cluster even though they are in separate compute nodes??

edit retag flag offensive close merge delete

2 answers

Sort by ยป oldest newest most voted

answered 2017-02-28 15:34:05 -0500

An instance needs to be scheduled to a single host. If you want your application to use all cores, it needs to be implemented on several instances.

But I always thought MPI is a communication protocol for running applications that are distributed over several, usually many hosts?

edit flag offensive delete link more

answered 2017-02-28 17:22:27 -0500

Instance can run on one host only. So, if all 192 cores are not on one physical node (which I don't believe :)), it won't work. Creating flavor with 192 cores will just result in scheduling failure. You can implement some sort of map-reduce mechanism and distribute it over nodes, but I can't think of anything else.

edit flag offensive delete link more

Your Answer

Please start posting anonymously - your entry will be published after you log in or create a new account.

Add Answer

Get to know Ask OpenStack

Resources for moderators

Question Tools

1 follower


Asked: 2017-02-28 09:42:45 -0500

Seen: 183 times

Last updated: Feb 28 '17