Scalable Computing Support Center
The Scalable Computing Support Center is a service and support center that connects Duke University researchers to hardware, software, educational and personnel resources, both local and global, to enable novel computational science. We will leverage the parallel computing facilities already in place at Duke, help build out the computational infrastructure to handle future work-loads, foster the development of scalable applications and assist in the training of parallel-aware researchers.
Duke Shared Cluster Resource
The SCSC works with OIT to manage and maintain the Duke Shared Cluster Resource -- a high-performance Linux cluster with nearly 680 machines and 2700 CPU-cores.
The DSCR is based on a sharing model where interested researchers may purchase machines and add them to the existing cluster. Once such a purchase has been made, they then get high-priority access to their own machines as well as lower-priority access to the spare cycles on all other machines in the cluster, including machines contributed (i.e. purchased) by other researchers.
The SCSC works with OIT to provide for the system administration, power, cooling and management of all of the machines in the DSCR. Technical documentation on the DSCR may be found on the SCSC wiki system.
Training and Consulting
SCSC staff are available to help with code optimization and parallelization -- enabling your existing programs to run faster or allow the use of larger data-sets. Through OIT's training program, we provide free classroom, group or department-specific classes on parallel computing topics and software engineering principles.
Regular training topics in our training program include the Sun Grid Engine batch scheduling system, Intro to Parallel Programming Models, OpenMP, Pthreads, MPI, CUDA, and Program Optimization. These classes are offered through OIT's Training Services group. A current listing of all OIT classes may be found on the OIT Training Web page. If you would like to schedule a group- or department-level seminar on a topic of interest, please contact us to set up a time. A list of all classes can be found on the training section of the SCSC wiki.
The SCSC also offers consulting services, led by staff with high-performance computing expertise in areas including scalable computing, debugging, code optimization and parallel programming models. We can sit down with you, one-on-one, to work directly on your program or workflow, software engineering process, tools and other project components.
Contact Us
Questions about cluster software or configuration, requests for access to the DSCR, and other inquiries regarding large-scale computational needs should be directed to scsc@duke.edu.