Reader Jay Blair asked me to pass along this info on goings on DICE Alliance Day at the beginning of May in Ohio. Both the workload management best practices and the PFLOPS workload panel look pretty interesting:
(Jay is) organizing a set of sessions on workload management for the DICE Alliance Day (http://www.avetec.org/dice/diceAlliance/DA07_overview.htm).
Survey of Current Government and Industry Workload Management
Jay Blair, ASC MSRC / DoD HPCMP
Lee Ward, Sandia National Laboratory / DoE
Deb Graico, Pacific Northwest National Laboratory / DoE
Dr. Dan Duffy, Goddard Space Flight Center / NASA
Ron Bewtra, GFDL – National Oceanic and Atmospheric Association (NOAA) / DoC
Jim Escoe, Procter & Gamble (P&G)
Brian Mitchell, GE Global Research
Rapid advancements in hardware technology and changes in user requirements (i.e., time sensitivity and interaction) have compounded scheduling and queuing requirements. Now, virtualization computing has reemerged as an approach to deliver greater use and flexibility of HPC resources. Current practices are resulting in most HPC centers are experiencing inefficiency and delays that prevent them from getting application runs completed productively. Learn about current HPC centers best practices and where industry leaders believe the future leads for scheduling HPC jobs in an increasingly complex environment.
Industry Panel: Workload Management of Tomorrow
Bill Nitzberg, GTO Grid Technologies, Altair / PBS Pro
Songnian Zhou, CEO, Platform / LSF
Waiman Chan, Senior Technical Staff, IBM / Load Leveler
Fritz Ferstl, Director of Grid Engineering, SUN / N1
David Jackson, CTO, Cluster Resources / MOAB
Ming Xu, HPC Program Manager, Microsoft / CCS
Srinidhi Varadarajan, CTO, Evergrid
The recognized industry leaders of Workload Management will participate on a panel session related to the current and future issues surrounding workload management as the state of the practice marches towards the Peta-Flop workload. Some topics that will be discussed are:
- What does the PFlop workload “look” like?
- Improving resource availability in massive environments / Metrics for WLM
- Holistic scheduling orchestrating storage, network, and compute resources
- WLM Simulation / Heuristics
- Paradigm shifts (multi-core / VMs)
- Specialized Processors (FPGA, GPU, Co-Processors, Cell)
- Topology aware scheduling
- Targeted resource delivery (QOS/SLA) and response time
- Adaptive computing/workload-aware event handling
As far as I can tell, this is the first time that these companies have come together on a single stage to discuss the future of the Peta-Flop Workload.