Reducing the Costs of Teaching CUDA in Laboratories while Maintaining the Learning Experience Quality

Abstract

Graphics Processing Units (GPUs) have become widely used to accelerate scientific applications; therefore, it is important that Computer Science and Computer Engineering curricula include the fundamentals of parallel computing with GPUs. Regarding the practical part of the training, one important concern is how to introduce GPUs into a laboratory: installing GPUs in all the computers of the lab may not be affordable, while sharing a remote GPU server among several students may result in a poor learning experience because of its associated overhead. In this paper we propose a solution to address this problem: the use of the rCUDA (remote CUDA) middleware, which enables programs being executed in a computer to make concurrent use of GPUs located in remote servers. Hence, students would be able to concurrently and transparently share a single remote GPU from their local machines in the laboratory without having to log into the remote server. In order to demonstrate that our proposal is feasible, we present results of a real scenario. The results show that the cost of the laboratory is noticeably reduced while the learning experience quality is maintained.