From owner-freebsd-hackers  Tue Apr 29 18:08:32 1997
Return-Path: <owner-hackers>
Received: (from root@localhost)
          by hub.freebsd.org (8.8.5/8.8.5) id SAA21909
          for hackers-outgoing; Tue, 29 Apr 1997 18:08:32 -0700 (PDT)
Received: from sendero.i-connect.net (sendero-ppp.i-Connect.Net [206.190.143.100])
          by hub.freebsd.org (8.8.5/8.8.5) with SMTP id SAA21889
          for <freebsd-hackers@freebsd.org>; Tue, 29 Apr 1997 18:08:28 -0700 (PDT)
Received: (qmail 10777 invoked by uid 1000); 30 Apr 1997 00:46:52 -0000
Message-ID: <XFMail.970429174652.Shimon@i-Connect.Net>
X-Mailer: XFMail 1.1-alpha [p0] on FreeBSD
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 8bit
MIME-Version: 1.0
In-Reply-To: <199704290644.XAA02193@rah.star-gate.com>
Date: Tue, 29 Apr 1997 09:08:24 -0700 (PDT)
Organization: iConnect Corp.
From: Simon Shapiro <Shimon@i-Connect.Net>
To: Amancio Hasty <hasty@rah.star-gate.com>
Subject: Re: New Features - invitation to comment
Cc: freebsd-hackers@freebsd.org
Sender: owner-hackers@freebsd.org
X-Loop: FreeBSD.org
Precedence: bulk


Hi Amancio Hasty;  On 29-Apr-97 you wrote: 
> 
> Well, 
> 
> 
> This sort of thing reminds me of  VMS clusters merged with NFS.

Sort of.  Maybe.  We are focusing now on block/raw devices.  File systems
are to be done later.  Not a priority yet.

> At first glance , I see the need for redundancy / mirroring . Also,
> the access time or weight for a given host block. Ability to migrate
> work load to different system on demand.

This is good. Does anyone have any access to the Mariposa project at
Berkeley?

> Does the system implements caching : write behind or read ahead?

Open issue.  not at the moment as first application does synchronous
writes (commit logs).

> Also I  would like to hear a little bit more on the distributed
> lock manager and how does it handle failure cases or what happens
> if the host where the application is running dies.

More details will be forthcoming.  Every ``owner'' has a co-owner.
Every pair holds mutual dead-man switches and boradcast the death of 
a co-owner.

> Perhaps, is best if the system address "virtual hosts/block" so if
> a system goes down requests can continue on a mirroring system.

Yes.  See above.

Simon