WIOSCA-07 Final Program


 8:30 - 8:55   Breakfast
 8:55 - 9:00   Welcome


 9:00 - 10:00   Keynote Address 1: Burton Smith

      Presentation Slides
    Speaker:   Burton Smith
    Bio:           http://www.microsoft.com/presspass/exec/techfellow/Smith/default.mspx

    Dr. Burton J. Smith, Technical Fellow for Microsoft Corporation, works with various groups within the company to help expand efforts in the areas of parallel and high performance computing. He reports directly to Craig Mundie, chief technical officer and senior vice president for Advanced Strategies and Policy. Burton is recognized as an international leader in high performance computer architecture and programming languages for parallel computers. Before joining Microsoft, he served at Cray Inc., formerly Tera Computer Company, as chief scientist and a member of the board of directors from its inception in 1988 to December 2005, and was its chairman from 1988 to 1999. Prior to founding Tera Computer Company in 1988 Burton spent six years with Denelcor, Inc. and three years with the Institute for Defense Analyses. From 1970-1979 he taught at the Massachusetts Institute of Technology and the University of Colorado. In 2003, Burton received the Seymour Cray Computing Engineering Award from the IEEE Computer Society and was elected to the National Academy of Engineering. He received the Eckert-Mauchly Award in 1991 given jointly by the Institute for Electrical and Electronic Engineers and the Association for Computing Machinery and was elected a fellow of each organization in 1994. Burton attended the University of New Mexico, where he earned a BSEE degree, and the Massachusetts Institute of Technology, where earned SM, EE, and Sc.D degrees.

    Title:        Many-core Operating System Challenges and Opportunities
    Abstract:  

    The advent of many-core chips will transform the operating system landscape. Parallel computing makes non-traditional demands on resource allocation and scheduling for both processors and memory, chiefly because strongly non-uniform thread progress makes it hard to reason well about parallel performance. New schemes for resource management will emerge and need new kinds of architectural support. Added to this is the need to handle different kinds of processors, e.g. graphics and I/O processors. Finally, traditional OS responsibilities for I/O such as DMA and interrupt handling may enjoy more direct support from hardware as parallelism acquires first-class status.

 10:00 - 10:30   Coffee Break


 10:30 - 11:00   Desktop Workload Characterization for CMP/SMT and Implications for Operating System Design

      Paper   Presentation Slides
  • Sven Bachthaler, Simon Fraser University
  • Fernando Belli, Simon Fraser University
  • Alexandra Fedorova, Simon Fraser University


 11:00 - 11:30   The Cost of IPC: an Architectural Analysis

      Paper   Presentation Slides
  • Isaac Gelado, Universitat Politecnica de Catalunya
  • Javier Cabezas, Universitat Politecnica de Catalunya
  • Lluis Vilanova, Universitat Politecnica de Catalunya
  • Nacho Navarro, Universitat Politecnica de Catalunya


 11:30 - 12:00   Re-architecting VMMs for Multicore Systems: The Sidecore Approach

      Paper   Presentation Slides
  • Sanjay Kumar, Georgia Institute of Technology
  • Himanshu Raj, Georgia Institute of Technology
  • Karsten Schwan, Georgia Institute of Technology
  • Ivan Ganev, Georgia Institute of Technology

 12:00 - 1:30   Lunch Break


 1:30 - 2:30   Keynote Address 2: Beng-Hong Lim

      Presentation Slides
    Speaker:   Beng-Hong Lim
    Bio:          

    Beng-Hong Lim is a Senior Director of R&D at VMware where he is responsible for the architecture and advanced development of VMware's virtualization technologies. He was a key developer of the initial versions of VMware's Workstation and ESX Server products. His research interests include system software, computer architecture, and distributed computing. He received a PhD in Computer Science from MIT.

    Title:        Virtualizing Beyond the Box: Architectural Support for a Distributed, Virtualized Compute Fabric
    Abstract:  

    Virtualization technology has moved beyond its initial application of enabling a single physical machine to host multiple virtual machines. Today, a distributed form of virtualization can manage multiple, networked physical machines as a pool of resources for hosting virtual machines. This talk will highlight several key features of such a distributed, virtualized compute fabric and discuss its implications on computer architecture and operating systems.



 2:30 - 3:00   Managing Shared L2 Caches on Multicore Systems in Software

      Paper   Presentation Slides
  • David Tam, University of Toronto
  • Reza Azimi, University of Toronto
  • Livio Soares, University of Toronto
  • Michael Stumm, University of Toronto


 3:00 - 3:30   Base Vectors: A Potential Technique for Microarchitectural Classification of Applications

      Paper  
  • Dan Doucette, Simon Fraser University
  • Alexandra Fedorova, Simon Fraser University

 3:30 - 4:00   Coffee Break


 4:00 - 5:30   Panel: Challenges and Opportunities for System Software (OS/VM) in the Multi-core Era
    Moderator: Emmett Witchel (UT-Austin)
    Panel:

          Konrad Lai (Intel) Slides
          Beng-Hong Lim (VMware) Slides
          Chuck Moore (AMD) Slides
          Burton Smith (Microsoft)
          James Smith (Wisconsin) Slides
          Michael Swift (Wisconsin) Slides
          Yuanyuan Zhou (UIUC) Slides

 5:30 - 5:35   Closing Remarks