allbery@ncoast.UUCP (Brandon Allbery) (06/23/86)
Expires: Quoted from <323@lewey.UUCP> ["Re: Re: Favorite operating systems query"], by wje@lewey.UUCP... +--------------- | > ... DEC has over the years developed an excellent | > layered system of hardware and software that allows the addition of more | > machines without putting one bunch of users off by themselves on a separate | > machine. In fact to the users, Vax-clustering makes many machines look like | > one. There's no graceful way with UNIX to add a machine and not find that | > you have to segregate some users from others. | > | > Steven A. Minneman (Fujitsu America Inc, San Jose, Ca) | | The above is not really true with respect to UNIX any longer. | Several vendors offer working distributed file systems for UNIX. | For example, NFS from SUN and TRFS from Integrated Solutions both | allow easy expansion to new machines without segregating users. | Also, Todd Brunhoff's RFS is included with 4.3 BSD, and has been | distributed to the network. It works with both 4.2 BSD and 4.3 BSD. | Even AT&T will eventually release their RFS for System V Release 3. | UNIX today, especially BSD UNIX, is much more sophisticated in this | area than it was a few years ago. +--------------- Not to mention Plexus NFS, which is a separate product from Sun NFS and has been available for (at least) two years, running under System III and now System V; it is Plexus's major product (look up ``plexus'' someday in the dictionary). Many computer companies now offer networking (Yes, Plexus NFS supports an rlogin-type interface; it's called ``vtty''.) I do not work for Plexus; in fact, I'm currently having an argument with them. I only mention this to show an example of network abilities other than the major ones (Sun, 4.xBSD). There are no doubt other companies with this kind of product. --Brandon -- ihnp4!sun!cwruecmp!ncoast!allbery ncoast!allbery@Case.CSNET ncoast!tdi2!brandon (ncoast!tdi2!root for business) 6615 Center St. #A1-105, Mentor, OH 44060-4101 Phone: +01 216 974 9210 CIS 74106,1032 MCI MAIL BALLBERY (part-time)
wmartin@brl-smoke.ARPA (Will Martin ) (06/24/86)
This discussion brings to mind the sort of environment I have long maintained as what we should have (for an Army activity which used UNIX-based workplace automation tools for our internal office automation and support work) -- some sort of amorphous group of machines running the same UNIX, where the user is not restricted or assigned to any particular machine, and where all files are available to all machines. The user connects (either via a local network command or some sort of dialup or network connection) and gets put automatically on whichever machine has the lowest load (either at that moment or over the past arbitrary time period). He can login and be put in his login directory, and has full access to the file structure (with whatever security is appropriate). If one of this group of machines needs work, the users on it can be transferred to others and that machine drops out of the group -- the net effect on the user community is not some arcane error message but instead merely longer response time, as the load is shared among fewer machines. I had hoped that the VAXCluster concept would someday work under UNIX and provide something like this; now, I don't know, given the seeming disinterest in ever getting VAXCluster to work for UNIX and also for the government-related concern of vendor independence (we never know, when we go out on a new procurement, just what we will end up getting, so we cannot specify "VAX" for example; only performance equivalent to <some specifications>). Any gurus out there think this dream is ever likely to become reality? Will Martin UUCP/USENET: seismo!brl-smoke!wmartin or ARPA/MILNET: wmartin@almsa-1.ARPA
devine@vianet.UUCP (Bob Devine) (06/27/86)
In article <1632@brl-smoke.ARPA> wmartin@brl-smoke.ARPA (Will Martin ) writes: >I had hoped that the VAXCluster concept would someday work under UNIX >and provide something like this; now, I don't know, given the seeming >disinterest in ever getting VAXCluster to work for UNIX and also for the >government-related concern of vendor independence (we never know, when >we go out on a new procurement, just what we will end up getting, so we >cannot specify "VAX" for example; only performance equivalent to <some >specifications>). > >Any gurus out there think this dream is ever likely to become reality? While I hardly qualify as a guru on the topic of DEC market direction, I have a feeling that DEC will one day offer Ultrix on a VAX cluster. Why? Simply because they will be able to sell much more of their boxes. We've just seen the result of an enormous R&D effort with the introduction of the 8x00 line. Judging from past introductions, it will be several years before another big jump is done. In the mean time, they have to provide something in the unending MIPS race. What other direction can they go? Bob