High Performance Computing

HPC Cluster

The High-Performance Computing (HPC) Cluster is available to researchers, faculty, students, and staff in need of large-scale computing. Tufts Linux HPC Cluster is comprised of hundreds of CPUs and GPUs.  Nodes are interconnected via a 100Gbps ethernet network.  Storage is provided by a parallel filesystem.
Tufts HPC is located in the Massachusetts Green High Performance Compute Center. The MGHPCC is an intercollegiate high-performance computing facility whose computing and storage resources serve a user community of more than 20,000 faculty and student researchers and educators, supported by on-campus research teams.

Please complete the Request Access form to gain access to Tufts HPC.

HPC Researcher Contribution Model

In addition to the significant shared resources provided by the university, individual researchers can buy their own nodes.  The Tufts HPC Researcher Contribution model lets researchers buy compute nodes for their projects, which are integrated into the Tufts HPC cluster by TTS Research Technology. This system removes admin responsibilities from node owners and expands HPC capacity for everyone. Owners have priority access to their nodes and set access policies, while unused capacity is shared with the Tufts community. Non-owner jobs are preempted quickly when owners need their nodes. While all ongoing operational and overhead costs are covered by TTS, there is a nominal charge to the researcher to cover the install of the nodes.

Specific details of the program can be found on our HPC policy page.

HPC Cluster Storage

The High-Performance Computing (HPC) Cluster storage offers a reliable, scalable, high-performance solution for distributed computing. It features a multi-petabyte all-flash storage system.  Users can access the system locally and remotely through ssh clients and Globus file transfer service. Tufts also provides web-based access via the OnDemand (OOD) web portal software. 

Faculty sponsorship is needed for access to HPC Cluster Storage. For more information, visit Research Data Storage.