This content has been marked as final. Show 2 replies
While I'd like to see you running Oracle Solaris on those blades and using Oracle Solaris Studio to build and optimise your applications, unless you specifically need to
build a high availability solution there is no requirement to deploy Oracle Solaris Cluster (http://www.oracle.com/technetwork/server-storage/solaris-cluster/documentation/solaris-cluster-businesscontinuity-168285.pdf).
Depending on your latency and bandwidth requirements for inter-node communication you may chose between gigabit, 10Gig and Infiniband fabrics, and typically your HPC
application will manage the "clustering" aspects of your compute task.
Since you're deploying an HPC type of cluster, the cluster software being suggested might be our Oracle Grid Engine product. It is a distributed resource manager (DRM) that provides a platform to schedule users compute jobs against the cluster and have them automatically load balanced and managed.
If you happen to be going to Oracle Open World, there are a few sessions specifically on deploying a compute grid:
The other cluster software, Solaris Cluster, mentioned is for providing traditional HA for your applications, but is not intended for compute grids. Again, Grid Engine also isn't required, but you should definitely consider it, if you don't yet have a DRM platform to manage the jobs.
Just in case, it may also be worth considering our Switched 10GbE NEM if you need to move a lot of data between the compute nodes.