SC is the International Conference for
 High Performnance Computing, Networking, Storage and Analysis

SCHEDULE: NOV 13-19, 2010

Techniques for Managing Data Distribution in NUMA Systems

SESSION: Research Poster Reception

EVENT TYPE: Poster

TIME: 5:15PM - 7:00PM

AUTHOR(S):Alexander Merritt, Kevin Pedretti

ROOM:Main Lobby

ABSTRACT:
MPI is the dominant programming model for distributed memory parallel computers, and is often used as the intra-node programming model on multi-core compute nodes. However, application developers are increasingly turning to hybrid models that use threading within a node and MPI between nodes. In contrast to MPI, threaded models often do not require application developers to deal explicitly with data locality. Data locality therefore has to be identified at runtime, using memory allocation policies such as first-touch or next touch, or specified statically by the application user at launch time (e.g., via the numactl tool on Linux). This poster evaluates several of these existing policies using micro-benchmarks on an AMD "Magny-Cours" system with 24 cores distributed among 4 NUMA domains. Additionally, we implement and evaluate a dynamic runtime migration scheme that leverages a novel pagetable replication scheme to gather per-NUMA domain memory access traces.

Chair/Author Details:

Alexander Merritt - Georgia Institute of Technology

Kevin Pedretti - Sandia National Laboratories

Add to iCal  Click here to download .ics calendar file

Add to Outlook  Click here to download .vcs calendar file

Add to Google Calendarss  Click here to add event to your Google Calendar

   Sponsors    IEEE    ACM