The Institute for Cyber-Enabled Research facilitates advanced computational research for the Michigan State University research community and additional affiliate organizations. ICER provides computational resources, such as the High-Performance Computing Center (HPCC), as well as training, user support, and consulting. Please see our About pagefor more information.
In addition to access, support, and training to utilize the High-Performance Computing Center, ICER provides a growing list of computational services for faculty, staff, and students engaged in research on campus. Please see For Usersfor more information.
Anyone pursuing research on MSU’s campus can get a free account to use the HPCC, if sponsored or requested by a faculty member. Faculty may request user accounts for themselves and for visiting faculty, graduate students, and undergraduates who are working with them. Many services are available at no cost.
Most services at ICER require an HPCC user account. For convenience, we utilize your MSU NetID and password once your HPCC account is active. Regular faculty members may request accounts for themselves. Students and staff must have their PI request their accounts. Click here to request an account.
If you use the HPCC for computation or consult with ICER, please acknowledge us in your publications. For example: "This work was supported in part by Michigan State University through computational resources provided by the Institute for Cyber-Enabled Research."
The HPCC team is the group within ICER that is responsible for building, maintaining, and supporting MSU's high-performance computing clusters and infrastructure. The HPCC infrastructure includes large, high-speed file systems, collaborative storage space, user access mechanisms, and all of the networking, electrical power, and the cooling needs that are required for a high-performance computing environment.
In general, think of ‘computation’ as something mathematical done on your computer such as a simulation of a real-world process, calculating statistical values, processing media (e.g. video or audio streams), comparing DNA, etc. Our laptop and desktop computers can handle millions of computations over a few hours, which is adequate for most things. There are cases where the number of computations is so high that the user may be waiting weeks, or longer, to get the results. High-performance computing is doing the same computations but on systems much larger than your typical computer. Some analysis for research requires billions of computations on hundreds of computers working together on large amounts of data or requires weeks of data. In a High-performance computing environment, these computations can be performed much more quickly compared to your average computer.
The HPCC maintains a number of computing clusters on the MSU campus. A computing cluster is a collection of high-speed computers for intense computational functions and data processing. The HPCC systems include several thousand processors, called cores, that can work together on very large problems. More information, and a complete description of the current systems at the HPCC can be found by visiting the User Documentation. (See [https://www.youtube.com/watch?v=CN0KP15r8A0](link is for a light-hearted and easy introduction to HPC).
A cluster is a set of nodes connected with high-speed networking to allow for massively parallel work, or simply to share among many users. The MSU HPCC has several clusters available. Users run their programs on these clusters by submitting a batch script to our queue and waiting for the scheduler to identify open resources. Please see the HPCC documentation for current details of our systems’ configuration. Users who buy-in to the cluster gain priority access to nodes they purchase and don’t have to wait as long.
The term 'node' in the context of high-performance computing is a single server computer with a special job. Each node has several cores and can run several processes at once. Some nodes have a special purpose, for example to allow access to our system (‘gateway nodes’), for writing and testing your programs (‘development nodes’), or for performing the main job of computational research work (‘compute nodes’), among others.
MSU is exemplary in how it funds and structures the HPC resources. Any MSU researcher has access to high-performance computing for themselves and their workgroup simply by requesting an account. Access to shared resources is via our HPC scheduling system. Those who request and need significant computational resources are required to wait for system resources to become available. While researchers can purchase nodes, even those nodes are still available for short tasks by the general user population when they are available. That is to say: We are good at sharing and nothing goes under-utilized.
ICER has a buy-in program where faculty may purchase nodes in the HPCC clusters. Faculty and research groups have priority-access to these purchased nodes when submitting programs for batch or interactive use. Please see the section on "What is a node" and our Buy-In program for details. However, when your buy-in nodes are not being used, they are available for short-run programs (typically less than 4 hours) by the other HPCC users. If you submit a job while another user has been placed on your buy-in node, then your job will be the next one to run. You do not need to install, provision, or maintain nodes you purchase.
In general, all users submit batch programs to our scheduling system and wait for parts of the cluster to become available for them to use. The main HPC cluster is not immediately available for users to log-in and run a program. Instead, users write their programs and then submit a request, also called a job, to run on the cluster. This job request specifies how much time and how much of the computing resources they will need. Based on these criteria, and how long the job has been waiting to run, the scheduling system then assigns these jobs to nodes when they are ready. Users are limited by the number of jobs and the maximum amount of resources that they may run at any given time. If these limits are reached, then their pending jobs are held until one of their currently running jobs ends. This way a single user cannot overutilize the system.