Networked Peripherals

[Gary Smith]

In article <1991Apr4.022904.11142 at nas.nasa.gov>, ciotti at nas.nasa.gov (Robert B. Ciotti) writes:
|> Is anyone using HPPI interfaces for networked peripherals or thinking
|> of doing this? What other things are people using them for?
|>  
|> Bob
|> --------------------------------------------------------------------
|> Robert B. Ciotti                         HSP Systems Development
|> Numerical Aerodynamic Simulation N258-5  (415) 604-4408
|> NASA Ames Research Center                ciotti at orville.nas.nasa.gov
|> Moffett Field, CA 94035-1000
|> --------------------------------------------------------------------

Bob,

I've had numerous internal (and with Ultra, NSC, Maximum Strategy,
etc.) conversations regarding such an idea over the past two years.
We (system programming group) implemented "networked mass storage"
for Cyber systems two decades ago at UT Austin.  Of course, the
"network" was quite homogeneous and the LAN rather hefty: Cyber
systems only communicating via ECS.  ONE of the minor problems with
the "heterogeneous networked peripherals" idea is the lack of stan-
dard data formats, not to mention other "minor" problems such as
little vs big endian byte ordering, etc.  UNICOS inodes don't look
much like SunOS inodes.  Even if they did, CRAY floating point num-
bers ain't very IEEE, etc.  Just use XDR some say.  Just let the
receiver make it right say others.  Of course I'm still talking
about problems at a rather "high level."  There simply isn't time
here to visit some of the lower level problems, such as specific
function code/timing sequences, etc.  You get the idea.  Networked
peripherals in the sense that many folks want them are HARD TO DO.

-- 
Gary Smith <gary at chpc.utexas.edu>
Systems Group, Center for High Performance Computing
The University of Texas System
Commons Building 1.151C, Balcones Research Center
10100 Burnet Road
Austin, TX 78758
(512) 471-2411