darrell@sdcsvax.UUCP (02/13/87)
Marty Fouts brought up several good points about Distributed
systems. I agree with the idea of a user interface to allow users to
write distributed applications. What I think Marty was saying was a set of
routines that users can use that may or may not send data off to another
system to be processed. This should all happen with out the user knowing
it is happening and knowing which machine it is going to.
There are numerous other options that can take this approach. Virtually
any application you are currently using could be written to take
advantage of this user interface. Ideally, you would want to design
the distributed interface so that existing applications could be linked
with a different library and not know the difference.
This is not trivial especially in a heterogeneous environment.
But, with this type of interface and the underlying support you should be
able to achieve some sense of load balancing and fault tolerance.
It would not be a "real" distributed system, but I think for most
general cases it would be an acceptable solution.
I just wish I had time to work on it :-).
One problem I seem to have with some of my colleagues and other managers
is how to let them "see the light". They are of the opinion if you need
to have several applications share information an the current machine
won't handle the load, you buy a bigger computer. Also They think
that distributed processing is not applicable to "their" applications.
I have been trying to educate them on the advantages of distributed
processing/systems but it is slow going. Anybody have any ideas on
how to "enlighten" them?
-----------------
Bill Stoltz
Texas Instruments
Process Automation Center
P.O. Box 655012, M/S 3635
Dallas, TX 75243
UUCP: {uiucdcs!convex!smu, {rice, sun!texsun}!ti-csl}!tifsie!bill
DECNET: TIFSIE::bill
Voice: (214) 995-2786
darrell@sdcsvax.UUCP (02/15/87)
[superstitious line-eater food] I have been checking out this discussion periodically and I decided it was time to say something. There is currently a distributed system project hosted at Rome Air Development Center (Griffiss AFB) which is being performed by the folks at BBN (Bolt Beranek and Newman) called CRONUS. I am currently researching this system for inclusion into a project that I am working on there. The nice things about this distributed system is that the routines are all incorporated into the system libraries and tag on to the system calls (read, write, open,close etc.) directly. The nice thing about that is that a person can port an application to the distributed environment rather easily. The system uses TCP/IP as a platform for communications. It has currently been tested over ARPA-net, Satelite link and directly over broadband ethernet. The current integrations (that I know of) are on UNIX (4.[23]BSD), VMS, Symbolics (don't know the status on this). Anyone from BBN/CRONUS care to comment? --------- My opinions are my own and anybody else who shares them is probably insane!