The Department of Statistics maintains about a dozen linux servers for faculty and graduate students, two of which have "Nvidia K20" graphics processing units(GPU) that can be used to run CUDA code. The linux servers are typically used to run R or MATLAB jobs, but are increasingly being used for julia, python, mpi, cuda, etc, by both students and faculty alike. In addition, about another dozen servers are maintained for faculty research, and/or for use by their group members. The linux servers are a mix of Ubuntu, RedHat, and Scientific Linux operating systems. The oldest server has 32GB RAM, and a 6 year old Xeon processor, while the newest servers have 256GB RAM, and the latest Xeon processors.
In 2012, the Department of Statistics acquired a 384-core (now 512) high performance computing (HPC) cluster which uses the SLURM resource scheduler. The cluster is mostly used for running array jobs, but is also capable of fully parallel operations. More recently, the department has built an experimental Hadoop cluster using parts from retired servers. It is mostly used for class related work, but will be expanded for general use in the future. In addition to the resources housed within the department, specialized or large-scale computing is sometimes done using cloud-based resources such as Amazon's EC2 and Elastic MapReduce systems.
Access to the servers is automatically granted upon entering the program, however you must contact department IT staff, or stop by 4206 MSB with your Photo-ID, to set an initial password.
For more information on these systems, hostnames, job submission, etc, please reference the systems page. (loginID required)