The Research Computing (RC) Team is excited to announce the new High Performance Computing (HPC) cluster, “Explorer,” which will supersede the current Discovery Cluster.
The Discovery HPC cluster has served the Northeastern University community extremely well over the last several years, and the Explorer Cluster will chart the path of a new journey in research and learning for the Northeastern community — with a new, more efficient operating system (Rocky Linux 9.3) and state-of-the-art GPU resources for compute-intensive computational science applications (NVIDIA H200).
What's New on Explorer?
Several new and exciting components of the Explorer HPC cluster expand the research capabilities for our community.
- The Explorer HPC cluster includes 32 powerful new NVIDIA H200 GPU chips with 140 GB of memory per GPU on the gpu and multigpu partitions. Here are a few useful materials:
- Explorer uses Rocky Linux 9.3, a current-generation operating system, which improves compatibility with current-generation software.
- A new version of Open OnDemand (OOD) with improved and standardized Open OnDemand applications that significantly enhances user experience.
- For many widely used scientific applications, compilers, and libraries, we have already built software modules and made them available to our users. If you need a particular module to be built for use on Explorer, please contact us at rchelp@northeastern.edu
How do I Access Explorer?
If you have used the Discovery Cluster recently, you should be able to access Explorer and run jobs. If you are facing issues (outlined below), please submit an Access Request ticket so the Research Computing team can create your HPC account on the Explorer cluster.
If you were not part of a /work directory on the Discovery Cluster or if you encounter one of the following errors when you try to submit a job on the Explorer Cluster, please submit an Access Request ticket.
- “sbatch: error: Batch job submission failed: Invalid account or account/partition combination specified”
- “srun: error: Unable to allocate resources: Invalid account or account/partition combination specified”
How do I Get Started on Explorer?
If you’re ready to get started on Explorer but need more information, the RC Team has written a quick start guide to the Explorer HPC cluster. The RC team has worked hard to ensure that this transition would be smooth and easy for our HPC users, so the Explorer HPC cluster uses the SLURM job scheduler, the same as the Discovery HPC cluster, and has the same user directories for /scratch and /home. The RC team has transitioned /work directories to /projects directories, but your already existing /work directories are present under /projects.
Principal Investigator Information
If you have a PI-owned partition on the Discovery Cluster and you would like to be moved to the Explorer Cluster, please send an email to rchelp@northeastern.edu, and we will arrange the transfer of your nodes to the new cluster. Eventually, all machines will be moved to the Explorer Cluster; following the positive experience of moving the course-related computing to the Explorer Cluster this Spring Semester.
Questions?
If you have any questions, please reach out to the RC team and we will be happy to help guide you through this transition. RC is happy to connect with you in the way that works best for you:
- Office Hours: RC Office Hours are a great way for you to connect with the RC team for short (~10-15 min) consultations. Office Hours are held every Wednesday from 3 – 4 p.m. ET and Thursday from 11 a.m. – 12 p.m. ET.
- Consultations: Schedule a consultation with one of Research Computing’s staff members to get personal, one-on-one assistance for your research computing and data storage needs. Consultations are available to any Northeastern student, faculty, or staff member.
- Ticket: To submit a ticket in ServiceNow, select from the RC ServiceNow catalog. You may need to sign in with your Northeastern username and password to view the catalog.
- Email: You can send the RC team an email at rchelp@northeastern.edu. This will also automatically generate a ticket in ServiceNow.
