The primary purpose of Tukey is to analyze and visualize data produced on Mira. Equipped with state-of-the-art graphics processing units (GPUs), Tukey converts computational data from Mira into high-resolution visual representations. The resulting images, videos, and animations help users to better analyze and understand the data generated by Mira. Tukey can also be used for statistical analysis, helping to pinpoint trends in the simulation data. Additionally, the system is capable of preprocessing efforts, such as meshing, to assist users preparing for Mira simulations.
Tukey shares the same software environment, network and file systems as Mira, enabling direct access to Mira-generated results.
Tukey has a total of 96 compute nodes; each node has 16 CPU cores and two NVIDIA Tesla M2070 GPUs. Aggregate GPU peak performance is over 98 teraflops double precision, and the entire system has a total of 6 terabytes of system RAM and 1.1 terabytes of GPU RAM. Access to Tukey is provided by two login nodes, which provide compilation and job submission capabilities. Job scheduling is provided by the Cobalt job scheduler. All Mira users receive a default allocation of 8000 core hours on Tukey.
Tukey Node Configuration
- Architecture: AMD x86_64
- Processors: Two 2GHz AMD Opteron 6128 per node (8 cores per CPU, 16 cores total)
- GPUs: Two NVIDIA Tesla M2070 GPUs per node
- Memory/node: 64GB RAM per node, 6 GB GPU RAM per node
- QDR Infiniband interconnect
- 450GB local scratch space