ARM Architecture Clusters
OLCF currently maintains two small clusters based on the 64-bit ARM architecture instead of traditional x86-based architecture. These systems are available to support computer science research projects aimed at exploring the ARM architecture. Appropriate research topics could include, but are not limited to, functionality testing of research software and prototypes, power-efficiency evaluations, and efforts to improve the system software, runtime support and the programming environment for ARM architecture based systems.
To access the computers, users must have an XCAMS account. (Click here to create one). Once you have your XCAMS user name, send an email to firstname.lastname@example.org. Include your XCAMS user name and which system(s) you want access to. An OLCF staff member will reply shortly.
Arm1 users must also have an NDA with Cray on file. Send an email to email@example.com for help establishing up the NDA. (There are no NDA requirements for access to Wombat.)
All published work related to these systems should include the following acknowledgement statement:
This research used the resources of the Oak Ridge Leadership Computing Facility, located in the National Center for Computational Sciences at the Oak Ridge National Laboratory, which is managed by UT Battelle, LLC for the U.S. Department of Energy, under the contract No. DEAC05-00OR22725.
ARM1 is a single rack Cray Envoy cluster with 16 compute nodes and 80TB in-rack storage. Each compute node consists of two 48-core Cavium ThunderX1 processors and 128 GB RAM (eight DDR4 DIMM’s). All compute nodes are interconnected through a 10 Gigabit Ethernet network.
Wombat is a single rack cluster from HPE. It has 15 compute nodes, two of which have two AMD GPU accelerators attached (four GPU’s total in the system). Each compute node has two (pre-production) 28-core Cavium ThunderX2 processors, 256 GB RAM (16 DDR4 DIMM’s) and a 480 GB SSD for node-local storage. Nodes are connected with EDR InfiniBand (~100Gbit/s).
Wombat hasn’t been installed yet. We hope to allow access to users by the end of September 2017.