IMAGINE, which is an acronym of Innovative Materials discovery Accelerated through Genomics and INformatics Engine, is a high performance cluster that is devoted to support computational materials science research at Virginia Tech.  IMAGINE is currently developed by Mr. Roger Link (Physics IT Staff) and Dr. Shengfeng Cheng (Physics Faculty) in the Department of Physics.

IMAGINE is a hybrid 444-core cluster currently with 19 DELL RX620 PowerEdge nodes, each of which contains 2 octa-core Xeon E5-2650v2 processors, giving 16 cores per node. IMAGINE also has 8 other nodes (Exxact Quantum TXR411/TXR430) equipped with NVidia Tesla GPUs, enabling GPU computation capability. These nodes have either 16 or 20 CPU cores as well. The inter-node communication is through infiniband adapters and a high-speed infiniband switch (36-port Mellanox Infiniband SX6036 QSFP Switch), which provides a high throughput and low latency interconnect solution. As a result, multi-core, multi-thread, and multi-node jobs can run very efficiently on IMAGINE. The headnode of IMAGINE has a 30T storage capability using a RAID system, which allows long-term safe storage of critical data.

IMAGINE has the potential to grow and to accommodate 8 more computing nodes, each of which can be either a traditional mutli-core CPU node (e.g., DELL RX620), or a node equipped with GPUs. Contributing nodes to IMAGINE is welcome, which will guarantee full access to the cluster for the contributing parties.

The main features of IMAGINE are: (1) High parallelization efficiency of multi-node parallel jobs; (2) Flexibility: jobs can run for a period desired by researchers; (3) Strong personalized support: code development, debugging, and testing are easy and the cluster software environment can be adapted to ensure the full capacity of a given software package; (4) GPU capability: IMAGINE provides a flexible platform to develop, test, and run codes that utilize GPU computations to speed up materials modeling and simulation.