From owner-freebsd-questions  Thu Nov 16 11:30:28 2000
Delivered-To: freebsd-questions@freebsd.org
Received: from corpmx1.ess.harris.com (corpmx1.ess.harris.com [130.41.65.49])
	by hub.freebsd.org (Postfix) with ESMTP id 01F3737B479
	for <freebsd-questions@freebsd.org>; Thu, 16 Nov 2000 11:30:25 -0800 (PST)
Received: by corpmx1 with Internet Mail Service (5.5.2650.21)
	id <VM8WS6XL>; Thu, 16 Nov 2000 14:30:22 -0500
Message-ID: <95B669A7D872D41182A600508BDFFB8C12D631@mlbmx7.ess.harris.com>
From: "Potts, Ross" <rpotts@harris.com>
To: 'Bruce Bauman' <bbauman@wgate.com>, freebsd-questions@freebsd.org
Subject: RE: NFS swap questions
Date: Thu, 16 Nov 2000 14:30:14 -0500
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2650.21)
Content-Type: text/plain
Sender: owner-freebsd-questions@FreeBSD.ORG
Precedence: bulk
X-Loop: FreeBSD.ORG

Here's my thought.  Haven't acted on it mind you, but, here goes.  Get a hard
drive for each machine expressly for swap(just large enough for the 1.5 - 2
times memory rule.  I think  multiple swap files being accessed at the same time
would really thrash the partition if it was one disk.  What if you used VINUM in
such a way as to allow striping across multiple disks comprising one partition
at the same time.  I believe you can set vinum to either append each volume as
needed or just use each one at the same time.  If only you can allocate which
volume to use for each machine in the single partition.  

Here's where a board like ABIT BE6-II would come in handy, as you can populate
it with eight ide drives(4 dma66, 4 dma33).  Of course if $ is no object, then
go SCSI.

Please if I'm way off, let me know.

        PC 1    PC 2      PC 3    PC 4
            \          |           |         /
             \         |           |        /
              \        |           |       /
   VINUM STRIPED SWAP PARTITION (defined on and controlled by server)
               /       |           |       \
              /        |           |        \
_______/_____|______|______\________
|disk one  | disk two | disk 3    |disk4       |
|              |              |              |              |
|swap 1/4 |swap 2/4 |swap 3/4 |swap 4/4 |
|------------|-----------|------------|------------|
                              |
                        SERVER

If you want to go real fast, install solid state hardware for the swaps


	-----Original Message-----
	From:	Bruce Bauman [SMTP:bbauman@wgate.com]
	Sent:	Wednesday, November 15, 2000 2:49 PM
	To:	freebsd-questions@freebsd.org
	Cc:	bbauman@wgate.com
	Subject:	NFS swap questions

	I want to build a medium-sized cluster of diskless FreeBSD 4.X-stable
	machines, all booting from a single FreeBSD server. I want them to swap
	over NFS to a single partition on the server.

	Because of the number of clients, I'd like the clients to allocate their
	swap files dynamically during boot (e.g. dd if=/dev/zero
	of=/nfsswap/swapfile.<ip-address> bs=1M count=256). I intend to do this
by
	slightly hacking the rc.diskless scripts. How do I then enable swapping
on
	these files? The available documentation isn't very clear.

	Second question: Are there any recommended options for the filesystem on
	which the swap files will live? Should it be async and noatime? Any
other
	performance tuning tricks?

	Thanks for any insight.

	-- Bruce


	To Unsubscribe: send mail to majordomo@FreeBSD.org
	with "unsubscribe freebsd-questions" in the body of the message


To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-questions" in the body of the message