Skip to main content


Introduction

The SDSC Comet system is a ~2 petaflops (PFLOPS ) Dell Linux Cluster that was put into production in April, 2015. Comet's diverse set of nodes caters to a wide variety of use cases. Most importantly, the fact that even standard compute nodes have relatively high memory and SSDs makes the system ideal for data-intensive applications, which most super computers are not as well equipped to handle. This makes Comet both a fairly general system for HPC, and one of the go-to systems for data or I/O intensive applications within the XSEDE community.

Used with the permission of SDSC
Panoramic photograph of 1 of 2 server rows
Photo courtesy of SDSC

However, Comet is not the best system to use if you are in the very small percentage of researchers who have extremely large, compute-intensive jobs (e.g. more than ~2000 standard cores), or if your code is better suited to take advantage of the Intel Xeon Phi architecture rather than GPGPU programming. There are other systems which may be appropriate in this case, and the Stampede Environment module covers a system which fills these niches.

This module is intended to insure that all users have the basic knowledge to use Comet efficiently. It contains the information needed to access the system, transfer data and code to the system, compile/link code, and submit jobs to the batch system. It also covers several examples that are well suited as Comet use cases.

Brandon Barker
Cornell Center for Advanced Computing

March 2017