IMFSSC: An In-Memory Distributed File System Framework for Super Computing

摘要

Supercomputing has been widely implemented in theoretical physics, theoretical chemistry, climate modeling, biology simulation and medicine research for high-performance and energy-efficient computing. Many of scientific applications are I/O sensitive and users have to tolerate high latency when supercomputing center storage processes thousands of I/O requests. In this paper, IMFSSC, an in-memory distributed file system framework for super computing is proposed, which is supposed to improve latency performance for I/O sensitive applications and relieves the congestion when I/O requests burst. IMFSSC consists of modules for multiple master-slave supports, load balance among huge amount computing nodes, and uses memory space to store data to minimize the latency. Additional features will be added in the near future and it is supposed to provide better support for large scaled computer systems. Finally, the performance is tested under the framework for evaluation, which shows high scalability and good I/O performance.