Will the Glass Walls Shatter when Big Iron Shifts

1
Will the Glass Walls Shatter when Big Iron Shifts?

• Arne Laukholm
• Director of IS, University of Oslo
• http//www.uio.no/laukholm

2
(No Transcript)
3
(No Transcript)
4
Opening remarks

• New technology will open new opportunities
• but
• the best technology may not always win.
• Success requires the right technology,
  organizational skills, market and timing.

5
A brief history lesson
6
Is there a point her somewhere?

• The cost of a true supercomputer, capable of
  sustaining the needs of any single Norwegian
  research group except for 3 or 4 can be purchased
  within the budget of any Norwegian University
• Should this affect the way we organize our
  business?
• I think not

7
Some unsubstantiated remarks

• Any internationally recognized research group
  will find the compute power they need somewhere
• A significant amount of the total research at our
  institutions is done by young, unestablished
  researchers
• Those researcher will adapt their research to
  whatever facilities they are given access to, be
  it PCs, local servers or supercomputers.

8
A consequence of this

• The need that we should focus on is then the
  requirements of the young researchers, in
  particular those that are not part of a well
  established research group
• Most of those needs volume computing rather than
  extreme peak speed.

9
A quote from the Mid-term Evaluation Report

• The spread in technology and in geographical
  location has been instrumental in attaining
  cost-effective computing for a heterogeneous
  group of users. Furthermore, the user demand is
  primarily for volume of CPU-cycles rather than
  peak performance. A continued strategy with
  several computing providers is therefore
  recommended.

10
Which are the options

• Discontinue the HPC program altogether
• Keep up, or increase research support staff, but
  only limited HW funding
• Have one center, like a downsized SDSC
• Four centers, different technologies, competition
  as well as teamwork

11
How should we choose?

• The choice should be the option that provides
  most added value to the research community within
  the available budget
• or perhaps
• the option that can attract the most outside
  funding?

12
What are the most important mechanisms to provide
added value?

• Access to people familiar with the tools and
  techniques of a given research field, ie active
  scientists within key fields working at the
  center
• Access to equipment that can sustain diverse
  requirements
• huge amounts of storage (Memory and disk)
• databases

13
From the Mid-term Evaluation Report

• The level of high performance computing
  expertise in the user community should be
  strengthened through a pool of support personnel
  from which users can apply for support time.

14
The DNMI Challenge

• The DNMI has no direct contribution to the
  overall funding of the HPC program, but has
  strong influence on operational patterns and
  technology directions.
• Should the HPC funding for the research community
  be split from the funds for operational weather
  forecasts?

15
A closing remark

• We are entering a period where changes will
  happen within the university community faster
  than we like to think
• So whatever plans we make need to be flexible
• They must be adaptive to changes in budgets,
  organization and marked needs
• Can we then afford to put all our eggs in one
  basket?

16
Thank you for your attention!