GPGPU Based High Perfor- mance Computation Support for Client-Server Framework

Abstract

In today's era of computing where applications can require upto terabytes of data for processing, virtually all microprocessor vendors have switched to models where multiple processing units, referred to as processor cores, are used in each chip to increase the processing power.

This brings into focus the Graphics Processing Units (GPUs) which emphasize on the many-core paradigm of data processing rather than the multi-core approach supported by most of today's modern Central Processing Units (CPUs). With the advent of General Purpose GPUs (GPGPU), even those applications that are not directly associated with graphics operations can still harness the computing capabilities of a GPU.

The objective of this project is the development of a platform-independent client-server framework to support GPGPU based high performance computation with transparency to underlying hardware and operating systems.

The system would provide extensibility in terms of the number and type of tasks or jobs that the client can submit for processing at the remote server connected to the GPGPU. The processed results would then be returned back to the client.