Home Sitemap Contact 中文 CAS
 
Navigation
  • HOME
  • About Us
  • Research
  • People
  • International Cooperation
  • News
  • Education & Training
  • Join Us
  • Publications
  • Papers
  • Resources
  • Life at ICT
  • Links
  • Location:Home>News>Upcoming Events
    Treating Data Like Grocery and Machines Like Stores
    Author:
    ArticleSource:
    Update time: 2009-06-20
    Close
    Text Size: A A A
    Print

    Speaker:Dr. Xiaosong Ma
    Time: 9:30am, 2009.6.23
    Place: Room 440, Institute of Computing Technology, Chinese Academy of Sciences

    Abstract:
    Today's data-intensive services (such as video-on-demand) bring more challenges to storage servers that handle many concurrent requests. Meanwhile, the advances in multi-core and many-core computers force us to examine new or expanded computation models to fully explore the potential of powerful nodes.
    In this talk, I will describe two recent projects conducted in our PALM (Parallel AppLications and MIddleware) group at North Carolina State University. In the first one, we explored applying supply chain management techniques to data prefetching in storage systems. With this approach, data requested by different client access streams are treated as different grocery items. For each stream, we adjust the prefetching level based on the customer consumption rate, using inventory theory algorithms. In the second project, we assess the feasibility of aggressive volunteer computing, where active nodes, rather than idle nodes, are used for running foreign workloads. We performed preliminary evaluation using the pairwise combination of 5 representative foreign workloads and 6 native ones. Our results indicate that by piggybacking additional workloads on active PCs, we can obtain significant energy savings compared to exisiting alternatives such as traditional volunteer computing, and in many cases the performance interference between the concurrent workloads is quite limited.

     

    Address :No.6 Kexueyuan South Road Zhongguancun,Haidian District Beijing,China
    Postcode :100190 Tel : (8610)62601166 Email : office@ict.ac.cn