O.S. Evalutation

Kyriakos Lakkas (klakkas@cs.ucsd.edu)
Thu, 18 May 2000 04:11:53 -0700

Thorsten von Eicken, David E. Culler, Seth C. Goldstein, and Klaus E.
Schauser, "Active Messages: a Mechanism for Integrated Communication and
Computation," Proceedings of the 19th International Symposium on Computer
Architecture, May 1992, pp. 256-266.

This paper presents the idea of active messages. The authors identified that
the two aproaches of the time serve the two edges of distributed execution;
message passing machines focus on processing, while message driven
architectures devote their resources to fast communication. They are
proposing a more balanced machine, that will allow communication between
processors in the same time with actual computation, minimizing the
communication overhead. To support their idea, they compare it with several
actual distributed systems of the time.

The notion of active messages seems pretty good. Especially nowadays, that
we have enough transistors to use inside the processor's chip, the
implementation of a separate communication co-processor that will only deal
with the active messages seems like a great idea. Furthermore, the
modifications proposed in both the processor and the network design do not
seem hard to implement. I don't know whether active message dominate today's
distributed systems, but if not, I am really curious to find out why...