word cloud about submit

Annual SubMIT review

by

On May 22, SubMIT had its annual review where the project team presented to the steering committee a summary of the cluster’s performance and recent upgrades. SubMIT started on December 9, 2021 and has now grown to 975 users, with consistently 50-80 weekly active users. From its inception, SubMIT’s goal has been to allow Physics researchers to access high-performance resources, enabling easy access for newcomers while supporting advanced customization for experienced users.

What happened in the past year?

A few of this past year’s key realizations:

  • Upgraded the operating system on all machines from CentOS7 to AlmaLinux 9
  • Changed the /data file system from gluster to cephFS
  • Integrated 6 machines (75 to 77, 82, 95 and 96) to provide hundreds of additional cores and increase the storage space by 3 times so we can continue to accomodate new users
  • Installed OpenMPI and added a Globus endpoint
  • Hosting A2rchi directly on the system
  • Added network links to reach 100 Gbps
  • Set up loose restrictions to improve the overall stability, e.g. by enforcing users to access the GPU nodes via slurm rather than direct login

User support with SubMIT

We analyzed all user tickets (emails sent to submit-help@mit.edu) collected within the last two years. The banner picture visualizes the relative abundance of topics in the email subject, where topics that appear more often are enlarged.

Number of user tickets opened by SubMIT users per month, as of May 15 2025.

Furthermore, we are happy to say that almost all tickets, including their possible followups, are resolved in less than 2 days!

Allocation of resources

By introducing Linux control groups we were also able to optimize the resources on our cluster and provide more computing power for slurm use.

Current allocation of CPUs on SubMIT. Furthermore, several thousand CPUs are available through HTCondor (not shown).

Looking ahead

We aim to continue providing stable operations, improve computational efficiency, the network speed, the amount of storage and tailored support to all our current and future users!

Leave a Reply