1994 MIT SWSC Debate - Affirmative Position
Resolved: That to exploit advances in parallel processing technology,
Fortune 500 companies should invest in networks of distributed workstations
instead of parallel computers.
See the illustrated summary of the debate (slow to load, but worth it).
In the first affirmative round, Dawson Engler summarized empirical evidence showing:
A more detailed summary is available.
- MPPs are unreliable (KSR, Paragon, CM-5). Their manufacturers (TMC, KSR) are also unreliable.
- Networks for distributed workstations are fast and getting faster.
- The University of Arizona successfuly put C* on a network of HP Snakes.
- MPPs don't have anything that workstations don't have, except for a huge price tag, more system adminstrators, and pretty lights.
In the second affirmative round (slides available in HTML and Postscript), Ellen Spertus emphasized the
following advantages of distributed workstations over MPPs:
- cost (entry cost, price/performance, ability to use legacy
- scalability (prototyping, expanding systems, meeting needs of
- system availability (reliability, ability to be physically distributed
and to be linked to other types of computers); and
- software availability (applications and development tools).
1994 MIT Student
Workshop on Scalable Computing