Next: Applications Categories
Up:
Characteristics of HPC
Previous: Applications Enhancement by
To summarize the conclusions of Working Group 2, some means of
investigating long term economic, social, legal, technical, viable models for
HPCC evolution in the USA must be found and a solution identified. We
note:
- the current incentive model for vendors, ISVs, and industries is incomplete;
- in Working Group 2 only two ISVs were parallelizing codes;
- the business case for HPC(C) uptake in a industry is typically inadequte;
- the use of MPPs in gas and oil industry is a counter-example and we need
to examine why this has been successful.
- the example of government intervention in Europe to provide product
development and advance customer uptake for HPC(C) may be worth following.
Working Group 2 therefore proposes the following actions:
- As an HPCC community we must identify platform independent software
standards scaling from workstations to MPP with tools using these
standards. This requires continued support and involvement in forums
like those for HPF and MPI but also requires great efforts to broaden
the user base who provide input to such forums.
- Industry and commercial users must be more actively
involved in HPCC - academics must find a way of becoming more involved
in end-user symposia as a first step to this. Currently, the tail
(academic developers) is wagging the dog (end-user communities).
In particular, can we identify big enough industrial markets
upon whose requirements to base viable HPCC standards (eg SMP,
distributed systems, WWW) and upon whom to focus the new enterprise model first?
- HPCC developers must find collaborators to broaden the application base
outside traditional science and engineering areas and into applications such as:
- integrated manufacturing (eg car body design);
- Low-latency (eg real-time systems);
- Data intensive (eg crisis management systems);
- Business (eg financial instrument simulation);
- Event driven modeling and simulation (eg defense and road traffic simulations).
Next: Applications Categories
Up:
Characteristics of HPC
Previous: Applications Enhancement by
Geoffrey Fox, Northeast Parallel Architectures Center at Syracuse University, gcf@npac.syr.edu