Calc was originally introduced in 2013 as DPB/DGE's single computational Linux server, and was recently upgraded and expanded in 2018 with vGPU nodes. This system is provided to allow users to run their code that require more computational resources, such as memory, CPU, or GPU, than a general desktop computer. Calc now has a login node and 4 compute nodes. Resources are allocated for each job submission and automatically re-niced, depending on the system load and prioritization.
With the vGPU expansion, code that specializes in graphical analysis, computation, or output can be given the appropriate resources that can only be provided by a GPU. Thus, codes that rely on Tensorflow or CUDA can be run on Calc.
While Calc boasts vGPUs, four additional compute nodes, and overall more computational resources than before, a computing cluster this size is perfect for computational newcomers to job submissions and resource allocation. It can function as a starting point to eventually build up to larger scale jobs that are usually run in larger clusters like Carnegie's Memex.
Features & Specs
- 1 login node and 4 compute nodes
- 40 vCPUs (8 for login node, 8 per compute node)
- 640 GB Memory (*128 for login node, 128 per compute node) *8GB Memory limit per user on login node
- 5 Tesla M60 vGPU (1 for login node, 1 per compute node)
- 25TB Dedicated NFS Scratch
- Connected to Homes and Data