RFS vs. NFS (really Locus and AIX)

[Richard Mathews]

> In article <10206 at ulysses.homer.nj.att.com>, ekrell at hector.UUCP (Eduardo Krell) writes:
> > Again, you're considering execution semantics, which clearly don't
> > belong in the File System. On Locus (now part of IBM's AIX), you could
> > exec any binary on any node and it would be executed on an appropriate
> > cpu. eg, exec'ing a Vax binary from a Sun would make it run on a Vax
> > CPU if available, fail otherwise.

Steve Dyer writes:
> I was not aware that AIX incorporated any such semantics from the Locus
> project; it wasn't mentioned as part of IBM's AIX family definition or of
> their "DS"--Distributed Services, in a briefing I attended in Austin last week.
> "Locus" the company bears little resemblance to the UCLA project; it shares
> the name and a few principals.

(For clarity, I refer to the company as LCC (Locus Computing Corp.)
and the product as LOCUS)

Mr. Krell is correct.  The above statement by Mr. Dyer is misleading
at best.  While the LOCUS functionality is not part of the AIX Family
Definition, it is part of AIX on the 370 and PS/2 Model 80.  I suppose
it's not in the Family Definition because these features have not been
announced for all machines which support AIX.  (I could be wrong about
all of this -- someone else here says it is in the Family Definition,
but I don't see it in that part of the announcement).

The reason that LOCUS was not mentioned as part of DS is because DS
is a totally separate product.  It is my understanding is that it is
NFS-like in that it provides a transparent file system capability,
though with more function.  I'd actually be interested in any
available description of DS vs. NFS vs. anything else comparable to DS.
Version 3.2 of NFS is also included in the AIX Family Definition.

The AIX port for the 370 and PS/2(80) is based on work done at LCC,
specifically the LOCUS Distributed Operating System first developed
at UCLA.  The LOCUS project is the oldest and largest of LCC's
development teams.  It is certainly not true that the company shares
only "the name and a few principals[sic?]" with the UCLA project.

Some info about an older version of the LOCUS O/S can be found in the book
	The LOCUS Distributed System Architecture
	Gerald J. Popek and Bruce J. Walker (eds.)
	MIT Press, 1985
The book describes some of the features of the system, as well as the
protocols used to implement them (a few years ago, that is).

I don't want to make this into a commercial, but in the interest of setting
the record straight, here's part of the IBM announcement (number 288-130,
Advanced Interactive Executive/370 (AIX/370), March 15, 1988):

        Transparent Computing Facility(*) (TCF) allows the distribution of
        data and processes among processors in a TCF cluster....  Within
        the TCF cluster, location of data and processes is transparent to
        both application programs and end users....

	    (*) Transparent Computing Facility is based on work done at
		the Locus Computing Corporation.

        Transparent File Systems: There is a single distributed
        hierarchical file system in a TCF cluster.  A file may be accessed
        from any node in the cluster....  TCF provides a method of
        optionally replicating files and directories on multiple cluster
        nodes to increase availability and performance of the file system.
        The multiple copies ar kept up to date by the system....

        Process Transparency: Allows a user to execute commands and run
        processes on any node in the TCF cluster.  Work may be routed
        either implicitly or explicitly to the most appropriate node in the
        TCF cluster....

        Process Migration: Allows a user to move a process in execution
        from one node to another of the same architecture....

The above is, of course, just my own rambling (except where I have quoted
others, hopefully accurately).  I am a member of the LOCUS O/S development
team and coauthor of a chapter in the book, but I am not in charge of
anything.  With any luck, I've worded everything carefully enough that I
will still have a job in the morning.

AIX and lots of other things are trademarks of IBM, LCC, AT&T, and lots
of other letters.

Richard M. Mathews
Locus Computing Corporation		       lcc.richard at CS.UCLA.EDU
			 {ihnp4,ucla-cs,trwrb}!lcc!richard