[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: freeze/unfreeze processes
---------- Forwarded message ----------
Date: Mon, 04 Aug 1997 22:52:53 -0700
From: Jim Dennis <jimd@starshine.org>
To: Ira Abramov <ira@scso.com>
Cc: jimd@starshine.org
Subject: Re: freeze/unfreeze processes
> Hi there Jim! we were sitting at the same table at the SVLUG meeting a
> month ago and you told me this is actually being looked into by someone,
> and I know people in the Hebrew U back here in Jerusalem have been doing
> something similar for a while... do you remember some Email addresses I
> can connect people with know-how to?
Unfortunately I don't have the references. I know that this
was a feature of the Sprite research project at UC Berkeley
and Apollo's Domain OS (before Apollo was purchased by HP).
One friend of mine is involved in some work on that which is
under NDA (and I'm therefore not at liberty to discuss his
work). It's not available and would be done under Linux or
FreeBSD in any event.
I remember coming across some references in the collection of
USENIX conference proceedings that I have here. Here's a list
of some references that I can dig up:
USENIX Conference Proceedings
New Orleans, LA, Jan. 1995
libckpt: Transparent Checkpointing Under Unix, p 213
by James S. Plank, Micah Beck, Gerry Kingsley
University of Tennessee; Kai Li, Princeton U.
plank@cs.utk.edu
beck@cs.utk.edu
kingley@cs.utk.edu
li@cs.princeton.edu
San Francisco, CA, Jan. 1992
Supporting Checkpointing and Process Migration Outside
the Unix Kernel, p 283
by Michael Litzkow and Marvin Solomon, Univ. of
Wisconsin, Madison
mike@cs.wisc.edu
solomon@cs.wisc.edu
San Diego, CA, Jan. 1989
Job and Process Recovery in a UNIX-based Operating System
by Brent A. Kingsbury, John T. Kline,
Cray Research Inc.
brent@yafs.cray.com
jtk@hall.cray.com
Dallas, TX, Feb. 1988
Process Migration in Unix Networks, p. 357
by K. I. Mandelberg, and V. S. Sundcram, Emory Univ.
gatech!emory!km
gatech!emory!vss
A Process Migration Implementation for a Unix System, p. 365
by Rafael Alonso, and Kriton Kyrimis, Princeton U.
alonso@princeton.edu
kyrimis@princeton.edu
Process Cloning: A System for Duplication Unix Processes, p.373
by Chad Hunter, MITRE Corp.
harvard!linus!chad
Note: Naturally I don't know if any of these e-mail address
are valid -- particularly the bang paths in the earlier years.
You could certain "domain-ize" them (gatech.edu, harvard.edu)
and try addresses at linus.harvard.edu and gatech.emory.edu or
emory.gatech.edu)
I hope these help. I also did a quick Yahoo! search and
came up with the following URL's.
Hibernator II: Checkpoint/Restart & Process Migration for UNIX
http://softway.com.au/softway/products/hibernator.html
Preemptive Process Migration in Networks of UNIX Workstations
http://docs.dcs.napier.ac.uk/DOCS/GET/bernard92a/document.html
And at jckyau's Bookmarks
http://www.cs.hku.hk/~jckyau/bookmarks.html
... I found these references:
Amoeba WWW Server
http://www.am.cs.vu.nl/
MOSIX
http://www.cs.huji.ac.il/mosix/txt_whatis.html
Load Balancing and Process Migration
http://web.soi.city.ac.uk/homes/james/loadrefs.html
Load Balancing and Process Migration:
References to Online Documents
http://www.cs.tu-bs.de/~petri/pgmigrefs.html
Warp Web on checkpointing
http://warp.dcs.st-and.ac.uk/warp/systems/checkpoint/
> ---------- Forwarded message ----------
> From: ury@memco.co.il
> To: Ira Abramov <ira@scso.com>
> Cc: Gilad Ben-Yossef <gby@gtekil.com>, linux-il <linux-il@linux.org.il>
> Subject: Re: killer net app
>
> On Thu, 31 Jul 1997, Ira Abramov wrote:
>
>> ok, the idea is this... what if you could detach an entire process, along
>> with memory allocation and open files, freeze it into a file and thaw it
>> on a different machine?
>
> It's been on the MF area for many, many years.
>
> For UNIX, similar thing was Done about 10 years ago, at least in one
> place - HUJI. Also the was the Pheonix project.
>
> I worked in that project untill about 5 years ago. We then did it for BSDI
> on Pentiums. Today you can login to a 64 200Mhz Pentium Pro machines,
> which behave as ONE unix. The diffrence is that we didn't freeze the
> process, just migrated it to another machine.
>
> To freeze a process ? This is difficult. There are numerous data
> structures in the kernel that you have to keep. Just the process table can
> be a mess. I don't think UNIX's design fits for that. Doing that on JavaOS
> is much easier.
>
> If you want to start a techinical discussion of how-to-do-that, pliz mail
> me.
>
>> uses? countless... move a compilation to a faster machine after it
>> started, maybe even on the fly (reminds anyone in huji of Prof. Barak's
>> monster?). freeze your sessions in the X windows and continue at home at
>> leisure. freeze a process at a specific breakpoint, then thaw it at that
>> point again and again for various break points (worth gold for beta
>> testing uses), send someone a program's demo as a started process without
>> a binary he can use and reinstall, and probably a thousand crazy uses in
>> client-server load balancing circus acts...
>>
>> for all I know there may be someone working on something like that as we
>> speak, I think it's worth donating effort into.
>>
>> -------------------------------------------------------------
>> Ira Abramov <ira@scso.com> Scalable Solutions
>> POBox 3600, Jerusalem 91035, Israel Tel (972)2-642-6822
>> http://www.scso.com/~ira Check out: http://www.linux.org.il
>>
>
> Ury Segal | Ate quando agosto dois ???
> 052-651629 | ------------------------- janeiro oito
> http://www.cs.huji.ac.il/~ury
> Hiroshima '45 Tschernobyl '86 Windows '95
--
Jim Dennis, info@mail.starshine.org
Proprietor, consulting@mail.starshine.org
Starshine Technical Services http://www.starshine.org
PGP 1024/2ABF03B1 Jim Dennis <jim@starshine.org>
Key fingerprint = 2524E3FEF0922A84 A27BDEDB38EBB95A