Site Review of the Ultra Project, September 1987

During the course of the days presentations we heard described an
overall review of the project a summary of the software development, a
description of the programming environment for parallel processing, a
description of the operating system, the existing hardware, plans for
the Ultra III, applications run on the current machine, algorithm work
in linear algebra and xxxxxxxxx xxxxxx computer science relating to
parallel processing.  The overall presentation had the feel of a review
rather than a new proposal for research.

The existing machine Ultra II working and available for
experimentation.  The machine itself is rather uninteresting in terms
of a parallel processor.  The kind of machine that is in place can be
purchased from a number of computer vendors.  In fact the commercially
available machine provides for many more processors and faster
interconnection machines.  This in itself is not a critisizm since when
the project was started such commercial machines were not available.
The same statement will most likely be true of the program and cannot
xxxxx xxxxx Ultra III in three years time.  The interest in this
project is primarily centered  on the construction of a working
combining network operating system.

The work on the operating system appears to have progressed.  Although
the absence of IBM as a research partner was conspicuous.

Parallel Computing Environment:

Work in this area is desperately needed.  The effort within Ultra at
this stage is small without a clear focus for future development.

Numerical Applications:

The fact that a number of applications have been ported to Ultra is
interesting.  The current system is too small to adequately draw
conclusions.  These same experiments could have been performed on say
a Sequent, Encore or Alliant.  The underlying mathematics behind the
implementation is of more general interest and not specific to Ultra.
This work is really independent of the construction of Ultra III and
could go on independently.

Bottom Line

The project's interest is mainly centering on building and experimenting
with a combining network.  Building the processor complex should not be
difficult given the experience with the existing machine.

Recommendation  

Reevaluate the project after a year to determine progress on the
network.

IBM should have more of a part of his project and cost xxxx xx
the instruction of the combining network.

Budget 

.