[Sandia National Laboratories]





Return to
Home Page




Second Workshop on Distributed Virtual Supercomputers
Sheraton Charleston -- Charleston, SC
March 30 - April 1, 1998

DRAFT Agenda, Version VI


Monday, March 30

8:00 Continental breakfast and registration
9:00 Discussion of workshop agenda, goals, procedures, and expectations
     Ralf Gruber
     Bill Camp
     Ken Kliewer
9:30 Overview of activities and current plans
     James Tomkins, SNL
     Swiss-Tx Ralf Gruber
     Martin Frey
     Pierre Kuonen
10:30 Break
11:00 Overview continues
     Bill Camp and Art Hale, SNL
     Al Geist and Ken Kliewer, ORNL
12:00 Lunch
1:00 State-of-the-art presentations
1:00 DEC - Alpha and interconnect plans
1:45 Tandem - interconnect plans
2:30 Giganet - VIA
3:15 Break
3:45 Organize workshop groups

Interconnects
     Leaders: Anton Gunzinger and Art Hale

Suggestion: a strategy for incorporating VIA and ST into
future XANs (SAN/LAN/WAN). At a minimum this group should
produce a list of gaps and preferably some strategies for
filling them.

Distributed computing infrastructure (conferencing, electronic
notebooks, security, scheduling, middleware, fault tolerance,
checkpoint, accounting, monitoring, ...=>resource management)
     Leader: Al Geist

Distributed storage and data management
     Leader: unknown

Suggestion: explore the roles of DFS, NFS, and HDF in detail.
Identify common name space strategies. Backup.

Applications
     Leaders: Grant Heffelfinger, Tim Sheehan

Suggestion: applications requirements group draw up as specific
a possible set of requirements as they can. The requirements
should include at a minimum compilation, job launch (interactive
and batch), debugging, performance monitoring, libraries (math
and system), I/O needs, and checkpointing. In each case the group
should be challenged to think about what is different in serial
vs. MPP vs. distributed systems.

Visualization
     Leader: Unknown

Suggestion: develop a proposal for some standard standard
interfaces between visualization services (rendering, etc.)
and data whether it be meshes, experimental data, or simulation
data. The interface should allow for data to be on tape, disk,
or in memory and allow for efficient pre-fetch and caching over networks.

General questions to be addressed in all groups:

  • Where are we?
  • What are our targets?
  • What is needed to get from a to b?
  • Define specific tasks and time tables.
5:30 Adjourn

Tuesday, March 31

7:30 Breakfast
8:30 Workshop groups continue
10:30 Break
11:00 Workshop groups continue
12:30 Lunch
1:30 Reports from working groups
1:30 Interconnects
2:00 Distributed computing infrastructure
2:30 Distributed storage and data management
3:00 Break
3:30 Applications
4:00 Visualization
4:30 General discussion of all the reports: emphases correct?
5:30 Adjourn

Wednesday, April 1

7:30 Continental breakfast
8:30 General discussion of planned systems in their entirety
     Capabilities
     Scalability
     Impact
     Attractive features
     Problems
     Specific systems for specific niches?
10:00 Break
10:30 Summary of system discussion
     David Greenberg
11:00 Decisions on tasks for the future
     Ralf Gruber and Bill Camp
12:00 Lunch
1:00 Adjourn



[Mail to:] Lilia G. Martinez

Last modified: March 6, 1998


Back to top of page

Questions and Comments || Acknowledgment and Disclaimer