Improving compute node performance using virtualization

Modified variants of Linux are likely to be the underlying operating systems (OSs) for future exascale platforms. Despite the many advantages of this approach, a subset of applications exist in which a lightweight kernel (LWK)-based OS is needed and/or preferred. We contend that virtualization is capable of supporting LWKs as virtual machines (VMs) running at scale on top of a Linux environment. Furthermore, we claim that a properly designed virtual machine monitor (VMM) can provide an isolated and independent environment that avoids the overheads of the Linux host OS. To validate the feasibility of this approach we demonstrate that given a Linux host OS, benchmarks running in a virtualized LWK environment are capable of outperforming the same benchmarks executed directly on the Linux host.

[1]  Srilatha Manne,et al.  Accelerating two-dimensional page walks for virtualized systems , 2008, ASPLOS.

[2]  Courtenay T. Vaughan,et al.  Application Performance under Different XT Operating Systems. , 2008 .

[3]  Peter A. Dinda,et al.  VNET/P: bridging the cloud and high performance computing through fast overlay networking , 2012, HPDC '12.

[4]  Jack Dongarra,et al.  Introduction to the HPCChallenge Benchmark Suite , 2004 .

[5]  Brian Kocoloski,et al.  A case for dual stack virtualization: consolidating HPC and commodity applications in the cloud , 2012, SoCC '12.

[6]  Peter A. Dinda,et al.  An Introduction to the Palacios Virtual Machine Monitor—Release 1.0 , 2008 .

[7]  Peter A. Dinda,et al.  Palacios and Kitten: New high performance operating systems for scalable virtualized and native supercomputing , 2010, 2010 IEEE International Symposium on Parallel & Distributed Processing (IPDPS).

[8]  Peter A. Dinda,et al.  Minimal-overhead virtualization of a large scale supercomputer , 2011, VEE '11.

[9]  Peter A. Dinda,et al.  An Introduction to the Palacios Virtual Machine Monitor— Version 1.3 , 2011 .

[10]  Alex Landau,et al.  ELI: bare-metal performance for I/O virtualization , 2012, ASPLOS XVII.

[11]  Ron Brightwell Why nobody should care about operating systems for exascale , 2011, ROSS '11.

[12]  Jack J. Dongarra,et al.  The LINPACK Benchmark: An Explanation , 1988, ICS.

[13]  Karsten Schwan,et al.  High performance and scalable I/O virtualization via self-virtualized devices , 2007, HPDC '07.

[14]  Jan Stoess,et al.  A light-weight virtual machine monitor for Blue Gene/P , 2011, ROSS '11.

[15]  Kamil Iskra,et al.  Performance and Scalability Evaluation of ‘Big Memory’ on Blue Gene Linux , 2011, Int. J. High Perform. Comput. Appl..

[16]  Susan Coghlan,et al.  Operating system issues for petascale systems , 2006, OPSR.