Re: [9fans] process migration

[Charles Forsyth <forsyth@caldo.demon.co.uk>]

[-- Attachment #1: Type: text/plain, Size: 917 bytes --]

some years ago, with an early Plan 9, when i was still
at the university, an undergraduate looked at process migration
as his 3rd year and MEng project, using Plan 9
running on several loosely-coupled 68030 cards in an Eltec VME crate, with
a PC running Plan 9 providing some form of monitoring
and control.  as a project it worked out quite well, and he'd
got it to the stage where he could shift processes from one
card to another.  the student made any necessary kernel changes
himself even though he hadn't touched any operating system before,
with relatively little advice from me.  (I'd done the Eltec port for
an earlier frame-grabbing robot control project.)

my view of OS-based process migration (outside rather specialised cases) before
the project would have been quite well summed up by ron minnich's comment;
i don't think it has changed much since then (outside rather specialised cases).

[-- Attachment #2: Type: message/rfc822, Size: 3270 bytes --]

From: ron minnich <rminnich@lanl.gov>
To: 9fans@cse.psu.edu
Subject: Re: [9fans] process migration
Date: Thu, 18 Sep 2003 07:38:26 -0600 (MDT)
Message-ID: <Pine.LNX.4.44.0309180735450.22490-100000@maxroach.lanl.gov>

On Thu, 18 Sep 2003, Kamal R. Prasad wrote:

> does it allow processes from moving from cpu server A to cpu server B
> after executing from sometime on server A? how is this different from
> UNIX?

it's not, neither Plan 9 nor Unix support this in a general way. There are
systems that support this in a limited way, however.

People keep wanting migration across OS instances to work the way that it
works on CPUs in an SMP. It doesn't, and never has, except in very special
cases. It's easy to get the easy parts right, hard to get the hard parts
right, and impossible to get the impossible parts right. Most people stop
at easy and call it a day.

ron