[comp.parallel] SURVEY - Interconnection Topologies

tbridg@iuvax.cs.indiana.edu (Tim Bridges) (09/03/90)

                    Interconnection Topology Survey

I am conducting a survey of the largest instantiations of various 
interconnection topologies. This work is part of an attempt to analyze 
scalability factors for parallel architectures. If you can provide a candidate 
architecture that may provide the largest example of any of the listed 
topologies, please respond with the size description and a reference to:

        tbridg@iuvax.cs.indiana.edu

As an example, I would submit the following:

        Hypercube - 4096 nodes (12 dimension), CM-2, Thinking Machines Corp.,
        and a reference to a TMC technical report or Hillis' book. This system
        has actually been implemented.


There may be bigger hypercube architectures built or proposed. If so, I would 
like to hear from someone out there.

I am solliciting entries for the following connection topologies. If you have
a submission that doesn't fit one of these categories, please submit and give
a description. This list is not meant to be exhaustive. No constraint is made 
as to SIMD or MIMD systems. I am attempting to collect data on physical 
systems as well as proposed systems. If in doubt, submit and provide 
descriptive text.

        Bus - # of processors
        2-D Grid
        Hypercube
        Omega or Banyon
        Crossbar
        Tree

In addition, I would like to hear about the biggest (# of words) RAM 
implementations on uni-processor and shared memory multi-processor systems. 
This data provides a good baseline.

Thank you in advance for your help. I will, of course, summarize and post
the survey results.

Timothy Bridges
tbridg@iuvax.cs.indiana.edu