Next: Working Group 2
Up:
NPAC Technical Report
Previous: Government Intervention Scenarios
The group identified a number of enhancements in applications that
would result from improvements in systems software.
- Better debuggers would speed up code development.
- Better code profilers and performance monitoring tools would help
identification of areas where HPCC could really improve applications.
- More stable operating systems releases for HPCC
platforms would enhance user confidence and increase chance of HPCC
uptake by industry, commerce and other mission critical software needs.
- Better collaboration software (and compatibility with existing systems
such as the world wide web) would lead to less time spent reinventing
and redeveloping algorithms and multiple unmaintained versions of utilities
that do almost the same thing.
- More resource-efficient library codes (in terms of storage capacity,
processing speed as well as bandwidth) would substantially enhance the capability
of existing HPCC systems.
- Since isolationism and commercial forces do not always produce the optimal
solutions - forums to introduce some standards from end-user
requirements (independent of the suppliers) would be powerful. HPF
and MPI forums are examples.
- Code browsing and analysis tools to aid in code migration to parallel
systems, whether it be in the form of libraries, or other software
engineering tools, would aid uptake of HPCC.
- Reduction of the latencies due to system software due to
under-engineered communications libraries would allow better use of existing
HPCC hardware and greater confidence in it.
Geoffrey Fox, Northeast Parallel Architectures Center at Syracuse University, gcf@npac.syr.edu