II-NEW: Distributed Computing Laboratory for Large Scale System Modeling and Analysis

点击翻译按钮获取中文摘要

This award is to acquire an 8 node multicore hybrid CPU+GPU system in support of research in use of multicore-based, GPU-based and hybrid high performance computing applications in the areas of large and complex system modeling and their solution methods, computational fluid dynamics, high performance data and stream processing, and dependable and high integrity methods on hybrid multicore and GPU methods. The proposed configuration also includes one front node and one storage server node with 80 TB disk capacity interconnected by an existing Infiniband network for multi-tiered, hierarchical parallel computing. Each compute node consists of Quad AMD new Interlagos processor 6220 with 8-core operating at 3.0GHz (can go up to 3.6GHz) and 512 GB of RAM, and a dual 448-core Tesla Fermi GPUs with 6 GB memory. The system will be used to support CISE centric interdisciplinary high performance computing research.This project establishes a computational instrument deploying multiple multicore hybrid CPU+GPU systems to support applications requiring hybrid high performance computing machinery. The instrumentation facilitates research in multicore- and GPU-based high performance computing in the context of the diverse set of applications at the frontiers of research: modeling of large and complex systems and their solution methods, such as high performance data and stream processing, dependable and high integrity methods on hybrid multicore and GPU, and computational fluid dynamics.