A fault-tolerant strategy for virtualized HPC clusters

Abstract

Virtualization is a common strategy for improving the utilization of existing computing resources, particularly within data centers. However, its use for high performance computing (HPC) applications is currently limited despite its potential for both improving resource utilization as well as providing resource guarantees to its users. In this article, we systematically evaluate three major virtual machine implementations for computationally intensive HPC applications using various standard benchmarks. Using VMWare Server, Xen, and OpenVZ, we examine the suitability of full virtualization (VMWare), paravirtualization (Xen), and operating system-level virtualization (OpenVZ) in terms of network utilization, SMP performance, file system performance, and MPI scalability. We show that the operating system-level virtualization provided by OpenVZ provides the best overall performance, particularly for MPI scalability. With the knowledge gained by our VM evaluation, we extend OpenVZ to include support for checkpointing and fault-tolerance for MPI-based virtual server distributed computing.

OpenCluster: A Flexible Distributed Computing Framework for Astronomical Data Processing

“…However, in general, MPI is rarely selected for developing real-time data processing systems because it does not provide standardized fault tolerance interfaces and semantics. Although extensive research (Rodrguez et al 2007, p. 153;Walters & Chaudhary 2009;Hursey et al 2011) has been conducted in this area, few available tools exist to help parallel programmers enhance their applications with fault tolerance support. Moreover, the exploitation of MPI is impeded by difficulties in software development.…”

A resiliency model for high performance infrastructure based on logical encapsulation

“…The resiliency of HPC platforms in the face of component failure has been a topic of significant work [4][5][6][7][8][9][10]. Although an abundance of checkpointing research has been performed on virtualized computing platforms [6,[8][9][10][11], these methods have not addressed heterogeneous systems including both virtual machines and externally hosted physical storage volumes.…”

“…Although an abundance of checkpointing research has been performed on virtualized computing platforms [6,[8][9][10][11], these methods have not addressed heterogeneous systems including both virtual machines and externally hosted physical storage volumes. In this paper, we offer an approach to resiliency that is applicable to these aggregated complex resources.…”