| Organizing Committee: | Lenore Mullin, William Harrod, Sonia Sachs (ASCR DOE) William Dally (Stanford), Peter Kogge (Notre Dame), Shekhar Borkar (Intel), Richard Murphy (Sandia National Laboratories) |
It has been approximately two years since DOE and DARPA completed their “ground breaking” exascale reports. In July 2010, the DARPA UHPC program was started. Four teams were funded to research and develop ExtremeScale designs. Various computer vendors have claimed that they are investigating the designs of exascale computers (IBM, HP and SGI). Several DOE labs are also investigating exascale designs (LBL, Sandia). The purpose of this workshop is to address the hardware and software architecture challenges for developing exascale computers. It would be very informative to revisit the exascale studies and explore what has been learned and to discuss research gaps. To achieve this goal we will meet and discuss the lessons learned from these various activities.
This workshop will address the following questions:
- What are the lessons learned?
- What are the unexpected technical challenges?
- What are the major risks and risk reduction plans for developing an Exascale computer in 2020?
- What can be said about the software challenges? This discussion should include both the system software and development environment.
- Can we develop a new program model that provides the developer with “ease of programming” to fully achieve their Exascale goals (performance, power, space, locality, etc.) or should we investigate how to extend existing languages and their programming models?
- Will the HPC users, funding agencies, and HPC market space support an ecosystem that is required for a new programming model and language?
- What technological advances are needed to achieve an Exascale computer and optimal run time system, or what is so unique about Exascale that innovation is required?
- Are there existing technologies that can be leveraged to provide the innovations needed? What are they?
- Is it realistic that a 20MW Exascale computer can be delivered in 2020? This is assuming that the system has a minimal memory configuration. The actual power requirements of a deployed system will be larger.
- Assuming that an Exascale computer needs to be delivered and operational in 2020, when do these investments need to be initiated? Is this goal achievable?
During the general discussion session, the attendees will develop a partially ordered, reverse time line that will indicate possible development paths, including a variety of technical challenges that lead towards a 2020 delivery date of an exascale computer.
A report that summarizes the workshop discussion and provides recommendations on gaps in exascale research will be generated by the workshop committee. The report will be completed within two weeks of the workshop date and outlined by the committee on the last day of the workshop.
The workshop will be held over 2 days: Agenda is subject to revision.
- Tuesday, August 2, the workshop will be held in the Paul G. Allen Building, Room 101, located at 420 Via Palou Mall, Stanford, CA 94305
- Wednesday, August 3, the workshop will be held in the Packard Electrical Engineering Building, Room 204, located at 350 Serra Mall, Stanford, CA 94305
- Please follow room directional signs in each building.
Copyright 2011, Oak Ridge Associated Universities
Privacy Security Notice | ORISE | ORAU