Difference between revisions of "Openstack at MIT Overview"

From GlueXWiki
Jump to: navigation, search
 
(6 intermediate revisions by the same user not shown)
Line 1: Line 1:
 
__TOC__
 
__TOC__
  
At MIT there are a few people working on a cluster designed to allow generic users to process data with custom software running on virtual machine images instead of directly installing the software on the nodes like a typical cluster.  This is geared toward a flexible computing model via virtualization and the technology being used is called [https://www.openstack.org Openstack].  I've installed GlueX software on a VM and deployed it on this cluster which has shown good results thus far (<5% different from running on "bare metal").
+
At MIT there are a few people (led by Jan Balewski) working on a cluster designed to allow generic users to process data with custom software running on virtual machine images instead of directly installing the software on the nodes like a typical cluster.  This is geared toward a flexible computing model via virtualization and the technology being used is called [https://www.openstack.org Openstack].  I've installed GlueX software on a VM and deployed it on this cluster which has shown good results thus far (<5% different from running on "bare metal").
  
== Reuse cluster (current) specs ==
+
== "Reuse" cluster specs ==
* 22 Dell blades with 8 cores, 16 GB RAM, 100 GB disk   
+
* 27 Dell blades with 8 cores, 16 GB RAM, 100 GB disk   
 
* Almost all hardware is recycled from other clusters through MIT's "reuse" program
 
* Almost all hardware is recycled from other clusters through MIT's "reuse" program
  
 
== Data Challenge 2 ==
 
== Data Challenge 2 ==
* Would like to include this in the upcoming data challenge
+
* Would like to include in the upcoming data challenge
** From the MIT perspective this would be the first demonstration of this cluster's performance and hopefully generate some additional interest
+
** From the MIT perspective, this would be the first demonstration of this cluster's performance and hopefully generate some additional interest
** From the GlueX perspective, would add to the total compute resources (expect to have exclusive access to ~200 cores) and explore virtualization
+
** From the GlueX perspective, this adds to the compute resources (expect to have exclusive access to ~160 cores) and explore virtualization

Latest revision as of 07:28, 21 March 2014

At MIT there are a few people (led by Jan Balewski) working on a cluster designed to allow generic users to process data with custom software running on virtual machine images instead of directly installing the software on the nodes like a typical cluster. This is geared toward a flexible computing model via virtualization and the technology being used is called Openstack. I've installed GlueX software on a VM and deployed it on this cluster which has shown good results thus far (<5% different from running on "bare metal").

"Reuse" cluster specs

  • 27 Dell blades with 8 cores, 16 GB RAM, 100 GB disk
  • Almost all hardware is recycled from other clusters through MIT's "reuse" program

Data Challenge 2

  • Would like to include in the upcoming data challenge
    • From the MIT perspective, this would be the first demonstration of this cluster's performance and hopefully generate some additional interest
    • From the GlueX perspective, this adds to the compute resources (expect to have exclusive access to ~160 cores) and explore virtualization