[comp.os.mach] MIG and netmsgserver vs. NIDL and location brokers

caswell@hplabsb.UUCP (Deborah Caswell) (09/23/89)

Does anyone know if there's been a written comparison of
writing distributed applications using Mach's mig and the
network message server versus using NCS's nidl and 
location brokers?  I would appreciate pointers to any
such comparison.

Thanks,
Debbie Caswell
caswell@hplabs.hp.com

caswell@hplabsb.UUCP (Deborah Caswell) (09/24/89)

We already have ISIS running on various machines in our lab on HP-UX.
What does ISIS have to do with Mach?  Is there a version of ISIS which
is implemented using Mach ports?

Debbie

chewy@apple.com (Paul Snively) (09/24/89)

In article <28300001@hplabsb.UUCP> caswell@hplabsb.UUCP (Deborah Caswell) 
writes:
> Does anyone know if there's been a written comparison of
> writing distributed applications using Mach's mig and the
> network message server versus using NCS's nidl and 
> location brokers?  I would appreciate pointers to any
> such comparison.

You might wish to add ISIS to your list; I just received reprints of all 
of the papers, as well as a bound manual (the manual is $15.00) from The 
ISIS Project.  If you want the same things, just send mail to 
schiz@gvax.cs.cornell.edu with the subject line "I want ISIS" and ask for 
the papers and, if you don't mind receiving an invoice for $15, the 
manual.  Just tell 'em what mailing address to send everything to.  The 
ISIS software itself is available from Cornell as well, either on tape or 
via anonymous FTP; The ISIS Project will tell you how to get it once you 
receive your written materials.

__________________________________________________________________________
Just because I work for Apple Computer, Inc. doesn't mean that they 
believe what I believe or vice-versa.
__________________________________________________________________________

ed@mtxinu.COM (Ed Gould) (09/26/89)

>You might wish to add ISIS to your list; I just received reprints of all 
>of the papers, as well as a bound manual (the manual is $15.00) from The 
>ISIS Project.  [Further description of obtaining ISIS.]

Also, ISIS will be included as user-contributed software on the mt Xinu
distribution of Mach and other things due out in January.

-- 
Ed Gould                    mt Xinu, 2560 Ninth St., Berkeley, CA  94710  USA
ed@mtxinu.COM		    +1 415 644 0146

"I'll fight them as a woman, not a lady.  I'll fight them as an engineer."

chewy@apple.com (Paul Snively) (09/27/89)

In article <28300002@hplabsb.UUCP> caswell@hplabsb.UUCP (Deborah Caswell) 
writes:
> We already have ISIS running on various machines in our lab on HP-UX.
> What does ISIS have to do with Mach?  Is there a version of ISIS which
> is implemented using Mach ports?
> 
> Debbie

I can't commend about whether they're using ports or not, but a Mach 
implementation is being beta-tested.

__________________________________________________________________________
Just because I work for Apple Computer, Inc. doesn't mean that they 
believe what I believe or vice-versa.
__________________________________________________________________________

ken@gvax.cs.cornell.edu (Ken Birman) (09/27/89)

In article <28300002@hplabsb.UUCP> caswell@hplabsb.UUCP (Deborah Caswell) writes:
>We already have ISIS running on various machines in our lab on HP-UX.
>What does ISIS have to do with Mach?  Is there a version of ISIS which
>is implemented using Mach ports?
>

We recently ported ISIS onto MACH, although not using Mach ports (what
advantage would this give us?  Not much, as far as we can tell).  This
port is available as part of ISIS V1.3, and will also be included on the
standard Mt. Xinu distribution of MACH next January (along with Camelot,
a system for doing transactional shared memory).

Anyone using ISIS at HP is probably using an old version, since ISIS
V1.3 won't really be "officially" out until next week.  That version
didn't run under Mach except on the NEXT machine.

I take it that OSF has made a decision to go with Mach as their OS.
We submitted ISIS to OSF under their recent RFT for distributed
computing environments.  However, OSF is a bit of a black box and there
is no way to predict whether they will actually include ISIS into their
offering.  Whether they do or not, my group is planning to introduce a
number of ISIS-based application programs next year, including a 
platform for dealing with sensors and actuators in realtime, a file
system that extends NFS to support transparent replication and fault
tolerance, and software for what we call "distributed application management"
(namely, the problem of supervising/scheduling/auto-restarting a
distributed but not especially robust application system).  And, we
will certainly have a version of ISIS running on the OSF system.

Regarding the original call for a "comparison", I am obviously too 
biased to comment on this.  But, I would like to observe that ISIS
is unique in addressing the issue of "consistency" as it arises in
a distributed system subject to failures and concurrency.  We believe
that if you don't start by tackling this fundamental problem, you
simply won't end up with a robust server.  On the other hand, once
you DO address this problem, you get a more uniform and consistent
environment, and application programs are far simplified.  It seems
that with enough elbow grease (mostly on my part), the ISIS system
can perform as well and scale as well as any other technology.
(ISIS applications generally don't see these issues.)

If you want papers on ISIS, contact me or my secretary, Helene Croft
(croft@cs.cornell.edu).  If you want a copy of ISIS V1.3, you can
copy it using anonymous FTP or via UUNET, and Helene can explain
how.  Its free and, with this latest version, seems remarkably robust.
We also have a news group (comp.sys.isis), mostly to announce releases
and bug fixes, and to answer questions about the system.  Not a
very active group, so far.  In addition to Mach and HP UX, we run on
DEC systems (Vax, 3100), the MIPS box, the Apollo 10.1 UNIX release,
IBM PC/RT's under AIX, the NeXT machine, and Gould's 9800 series.
Our system has interfaces to C, GCC, ANSI C, C++, Lucid and Allegro
Common LISP, Fortran, and (in a limited way) C-Prolog.  We are 
hoping to see ports to the Iris, Cray's UNIX, and the 3B series of
machines soon, and a port to Bull's equipment exists but I don't
have a copy of it yet.

Ken Birman