Date: Wed, 4 Mar 1998 10:34:04 -0500 (EST) From: "Ron G. Minnich" <rminnich@Sarnoff.COM> To: Niall Smart <njs3@doc.ic.ac.uk> Cc: Tom <tom@sdf.com>, "Ron G. Minnich" <Sarnoff.COM!rminnich@minas-tirith.pol.ru>, Alex Povolotsky <tarkhil@minas-tirith.pol.ru>, hackers@FreeBSD.ORG Subject: Re: Cluster? Message-ID: <Pine.SUN.3.91.980304102946.27680D-100000@terra> In-Reply-To: <E0yACqr-0005la-00@oak2.doc.ic.ac.uk>
next in thread | previous in thread | raw e-mail | index | archive | help
> On Mar 3, 11:04pm, Tom wrote: > I think a good approach to transparent clustering is through distributed > shared memory. However, the coherencey schemes used for high reliability yes, this has been an active area too. DSM is great, we build 'em, we use 'em, but it's not the panacea. It's just good for some things, poor for others. > I'd be more than interested in adding some DSM support to FreeBSD > to support transparent fall-over clustering. well, it's a nice goal, the trick is getting the VM fixed, and other pieces fixed, so that it will work. I tried to get this to happen in 1994, when I offered FreeBSD MNFS (see web page for info on MNFS), but the reception was muted, although some projects in some places did use MNFS. Note that Chuck Cranor's UVM fixes the problems, but it is only in OpenBSD and NetBSD. ah well. In terms of crash resilience: while not perfect, MNFS was as crash resilient as NFS, and in fact we used it to support DSM for migrating Condor processes. Condor processes migrate by sending themselves a signal and dumping core, and then restarting the corefile elsewhere. ron To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hackers" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.SUN.3.91.980304102946.27680D-100000>