Talapas Supercomputer

talapas serversThe Talapas supercomputer is a High Performance Computing (HPC) cluster capable of performing over 250 trillion floating point operations every second. The acquisition and deployment of Talapas marks a new era for research computing at the University of Oregon. A true research cluster, Talapas is designed to accommodate a wide variety of workflows, from classical MPI based distributed computing to GPU accelerated simulations to large memory applications. This heterogeneous design delivers a single cluster that can effectively serve the unique needs of the diverse computational research community at the University of Oregon.

Located in the state of the art Allen Hall Data Center, Talapas is currently in late stage open beta testing and is available for use by all researchers at the University of Oregon. We also welcome applications for use from our sister institutions in Oregon.

Talapas takes its name from the Chinook word for coyote, who was an educator and keeper of knowledge. The RACS team consulted with the Confederated Tribes of the Grand Ronde to choose this name, following a Northwest convention of using American Indian languages when naming supercomputers.

Citation and Acknowledgement

Please use the following language to acknowledge Talapas in any publications or presentations of work that made use of the system:

"This work benefited from access to the University of Oregon high performance computer, Talapas."

Specifications

The following are the specifications of Talapas as deployed. The cluster continues to grow through ongoing condo purchases. 

Compute Hardware

The Talapas HPC cluster contains three categories of nodes: standard nodes, GPU nodes, and large memory nodes. 

qtynode typeprocessors (total cores)memorylocal storagenetworkingaccelerator
96Standard Nodesdual E5-2690v4 (28 cores)128GB200GB SSDSingle Port EDR InfiniBandN/A
24GPU NodesDual E5-2690v4 (28 cores)256GB200GB SSDSingle Port EDR InfiniBandDual NVIDIA Tesls K80
8Large Memory NodesQuad E7-4830v4 (56 cores)1TB, 2TB, or 4TBdual 480GB SSDSingle Port EDR InfiniBandN/A

Storage Hardware

All compute resources are connected to our DDN GRIDScaler 14k storage appliance via the EDR InfiniBand interconnect. The DDN appliance runs GPFS and provides over 1.5 PB of usable storage.

applianceenclosuresdrivesfilesystemusable space
DDN GS14k5 SS8462 84-slot enclosures10 800GB Mixed Use SSD Drives (metadata)GPFS1,579 TiB
  362 6TB 7200 RMB 12Gb/s SAS 4Kn drives  
  21 800GB Mixed Use SSD Drives (fast tier)  

Interconnect

All compute nodes and the DDN storage controllers are connected via a high-speed EDR InfiniBand network providing 100Gbit/s throughput. The network is arranged in a "fat-tree" topology with compute nodes connected to leaf switches and all leaf switches connected to core switches. Currently, the network is configured with a 2:1 overall blocking ratio, i.e. there are twice the number of connections up from the nodes to the leaf switches as there are from the leaf switches to the core switches. This allows us to scale economically while providing unblocked communication between the 24 compute nodes that share a common leaf switch. The two DDN storage controllers each have dual connections to the InfiniBand core switches.

interconnect