High performance computing has become an important element of research on campus. Many groups have wanted to build clusters, but do not know where to start. IMSS can help from the design, building, implementation, and system administration. We have built clusters from 7 nodes all the way up to 1024 nodes. Depending on the needs of the group, we can implement high speed interconnects, parallel filesystems, or simple gigabit based clusters with NFS
We are running clusters on a variety of types of hardware and even in the cloud. We do prefer to use specific vendors and hardware types, as we have found that this approach helps us with standardizing and monitoring of the clusters we maintain.
Depending on your preference, requirements, and budget, we can deploy different schedulers and resource managers. Some small clusters opt to operate without a scheduler, but running a schedule will give you better efficiency of your cluster and simplify use of the resources.
If you need help getting tarted with a cluster, need computational resources and don’t know where to go, or just have some questions, let us know.