home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.sys.isis
- Path: sparky!uunet!wupost!usc!rpi!batcomputer!cornell!ken
- From: ken@cs.cornell.edu (Ken Birman)
- Subject: Re: Forwarded comments Re: Notes on IP Multicast option for Isis
- Message-ID: <1992Dec24.170241.28587@cs.cornell.edu>
- Organization: Cornell Univ. CS Dept, Ithaca NY 14853
- References: <1992Dec24.162926.27546@cs.cornell.edu>
- Date: Thu, 24 Dec 1992 17:02:41 GMT
- Lines: 66
-
- In article <1992Dec24.162926.27546@cs.cornell.edu> ken@cs.cornell.edu (Ken Birman) writes:
- >From: gthaker@hannibal.atl.ge.com (Gautam H Thaker)
- >
- >Here are my two cents worth of comments...
- >
- > When will IP mcast be used?
- > - I plan to require that all (or most) machines using Isis support
- > the IP Multicast standard, which was developed by Steve Deering at
- > Stanford and has become part of Solaris 2.1; Steve's code is also
- > available for many other platforms, but perhaps not as a standard
- > offering from the vendor.
- >
- >Requiring that most machines using isis support IP Multicast seems
- >a bit strong. I don't know if HP, DEC, IBM, Lynx etc. support this
- >as well as SOlaris 2.1, be cause we worry about using ISIS from all
- >these other platforms.
-
- Well, I really prefer to have the standards argument to point to.
- based on Werner's comments, sounds like most of this actually do support
- the interface, and if Isis does something of its own, we lose the
- strong advantage of being "standard compliant". If, say, Lynx lacks
- support for this -- a published IP standard -- why not lean on them
- to offer it? More appealing than porting some wierd Isis solution to
- a zillion driver-level interfaces...
- >
- > - You will need to have an authorization for "ipmcast" in your isis.rc
- > file.
- >Eventually, multicast should be folded in with price of standard ISIS
- >to keep ordering process simple, but no big deal here.
-
- Anyone who has Isis-All gets this as part of the package.
-
- > - Only messages sent in diffusion mode (to the full set of members and
- > clients) will be multicast. However, you can overlap groups and in this
- > way can build up other patterns pretty easily.
- >Not sure about this. Does this mean that if I have peer group of 10 members
- >and I do a normal cbcast multicast won't get used?
-
- Actually, what it really wants is that "the destination includes all members
- and all clients". So, for your case, it would be used. You don't need
- to specify "D" (diffusion mode) explicitly, but you do need the right
- destinations...
-
- >Please consider publishing a "performace brief" for ISIS, including
- >numbers for multicast perf. I realize that you and your staff does not
- >have infinite time, but perhaps slowly something can be built up.
-
- Actually, we had been holding off mostly because of this multicast issue.
- I didn't want to compare Isis with things like Ameoba if there would
- be some glaring unfairness, like them using hardware multicast and us
- using point-to-point communication... now that there is a standard I
- can at least quote comparable numbers (and, I am confident, the numbers
- will be in the same ballpark as any others, or better -- because if they
- aren't, we'll tune until they are!) (That is, for stuff running over UNIX).
-
- >Also, how do people find Solaris 2.x overall? Do things work seamlessly,
- >or is some amount of effort needed to move to Solaris?
-
- Don't know yet. Others who read this probably can comment but I only
- have worked with Solaris to do the initial Isis port -- that went smoothly,
- by the way. We are waiting for Solaris threads and will start to use
- the Solaris system much more in 1993 once we are on the threads package.
- --
- Kenneth P. Birman E-mail: ken@cs.cornell.edu
- 4105 Upson Hall, Dept. of Computer Science TEL: 607 255-9199 (office)
- Cornell University Ithaca, NY 14853 (USA) FAX: 607 255-4428
-