Body
What is a Supercomputer/HPC Cluster?
Calling something a “supercomputer” can bring to mind images and ideas of something futuristic and almost alien, but the reality of supercomputers is much easier to understand. The “super” in supercomputer is parallelism, meaning that it can take a very large job that would otherwise take a huge amount of time to compute is split into many pieces that all run at the same time, even using multiple compute nodes. These “nodes” are just computers. They have many CPU cores, lots of RAM, and very high-performance networking, but they are not fundamentally different from an everyday computer. A supercomputer/HPC cluster uses many such nodes, connected using high-performance networking, and a scheduler like Slurm that manages the resource requests and jobs that users submit to run on the set of resources that make up the cluster.
What is the Mill?
The Mill is the current iteration of Missouri S&T’s supercomputer, following the decommission of the Foundry. To get access to the Mill, please fill out the account request form (https://missouri.qualtrics.com/jfe/form/SV_e5n4I8cvzo77jr8). We ask that you cite the Mill’s DOI on any publication where the Mill’s resources were used (https://doi.org/10.71674/PH64-N397). This both helps us keep track of the value that having the cluster adds to campus and helps us justify the budget to continue supporting and improving the supercomputing program. If you need more immediate access to resources than the queue provides, or more storage than the default 50GB home directory, we also provide leases on multiple types of compute nodes and storage.
Who Can Benefit from Using One?
Not all work can benefit from the parallelism of a supercomputer. Work with many independent tasks, meaning tasks that do not require input from, or the results of the other tasks, will parallelize very well and benefit greatly from a supercomputer. Work that consists of many steps that each depend on the result of the previous step or require processes to update one another frequently will not benefit as much from parallelism, but they may still benefit from the large amount of memory that is available on the compute nodes. If you have questions about whether your workflow is suitable for HPC feel free to request a consultation (https://tdx.umsystem.edu/TDClient/48/Portal/Requests/ServiceDet?ID=1096), and we will help find the most suitable computing resources for you.