Toward a Proof of Concept Implementation of a Cloud Infrastructure on the Blue Gene/Q

Abstract

Conventional cloud computing architectures may seriously constrain computational throughput for HPC applications and data analysis. The traditional approach to circumvent such problems has been to map such applications onto specialized hardware and co-processor architectures. These shortcomings have given rise to calls for software that can provide richer environments for implementing clouds on these HPC architectures. It was recently reported that a proof of concept cloud computing system was successfully embedded in a standard Blue Gene/P HPC supercomputer. This software defined system re-arranged user access to nodes and dynamically customized features of the BG/P architecture to map clouds systems and applications onto the BG/P. This paper reports on efforts to extend the results achieved on the BG/P to the newer BG/Q architecture. This work demonstrates a potential for a cloud to capitalize on the BG/Q infrastructure and provides a platform for developing better hybrid workflows and for experimentation with new schedulers and operating systems within a working HPC environment.