The University’s Office of Research supports the Department of Applied and Computational Mathematics and Statistics (ACMS) and other programs through its Center for Research Computing. ACMS also has access to several major computer cluster resources:
- ICSB grad cluster: This Beowulf cluster consists of 24 cores with 1 GB of memory per node. This essential resource has real-time access for the development of new parallel algorithms. This contrasts with the normal queued method for “production” code and allows for parallel optimization prior to running the code for long periods of time.
- The ACMS has priority access to 720 cores in the Center for Research Computing (CRC) cluster. These were purchased with funds from the College of Science. CRC is configured to handle the diverse ACMS requirements that range from solving multiscale models to Molecular Dynamics simulations to constructing genome databases. The cluster consists of over 7032 available cores, with a minimum of 1GB of RAM per core. Access to the resource is using a queuing system, ideal for long production code runs with multiple head nodes. In addition, there are several “large memory” queues for applications that require it. Interconnections are over QDR Infiniband or GB Ethernet, and the cluster is connected to the campus backbone through 10 Gbps optical fibers. Access to the cluster is granted to users through stringent security protocols. The CRC cluster is connected to 300T of permanent storage, accessed by users via the Andrews File System (AFS), and 64T of high-speed “scratch” storage for shared program access. The permanent storage is backed up to an offsite Sony PetaScale storage silo.
- Prof. Sommese’s group has priority access to 680 cores with over 1.2 TB RAM in the Center for Research Computing (CRC) cluster. These were purchased with external funds and funds from the Vincent J. and Annamarie Micus Duncan Chair.
- Prof. Hauenstein’s group has priority access to over 1000 cores with over 2 TB RAM in the Center for Research Computing (CRC) cluster. These were purchased with external funds.
In addition to the clusters, ACMS has a dedicated server for hosting websites pertaining to current research, allowing dissemination and sharing of valuable information (e.g. biomath.math.nd.edu). ACMS envisages that some projects will incorporate a portal into their dedicated websites to allow approved parties to actually run developed code on ACMS clusters. The server also houses the ACMS Wiki for enhanced collaboration between members of ACMS and project partners.