imaging a large group via NetBoot- Casper

[#1]

We are using Casper with a Netboot server and only one Casper server for
the packages. It seems to be taking a very long time to image MacBooks
out of the box. What kind of numbers has anyone else been getting? If I do just 1 or 2 it will complete in about 45 minutes, if I step it up
to more than 7 the time goes to over 3 hours. If I add more servers for
the packages will that help?

Any ideas, suggestions or hints would be greatly appreciated.

Thanks, Shawn

Liverpool Central School District
Computer Services
Liverpool, NY 13090
(315) 622-7150

[tlarkin]

I am getting read to do a mass image of about 6,000 Macs over the
Summer. At my previous job last year I helped reimage 10,000 machines
(mostly windows boxes) over our Summer break. I wrote a simple shell
script with Zen Imaging (Novell) that ran at netboot and automated a lot
of the steps needed to reimage a machine.

I have found it best practice if you are going to do this, just go ahead
and buy a few big "dumb" switches, with no management on them at all. Hook your machines up to the "dumb" switch and start to netboot. With
things like Port Fast and Spanning Tree I think you definitely get
overhead from your managed switches. We were unicasting everything
since we could never get multicasting to quite work right (mainly due to
the managed switches making it take longer).

We didn't have time to fuss with it so we just unicasted machines. 10,000 over a 2.5 month period isn't too shabby.

This year I plan on doing the same. Getting an area where I can run
power and a dumb switch and reimage like 50 at a time with very simple
non managed networking.

Also, if you are using a lower spec machine that is unicasting say, 20+
connections pushing out a 6+ gig image it is going to take a while. Since each connection will take up resources from the netboot server.

Thomas Larkin
TIS Department
KCKPS USD500
tlarki at kckps.org
cell: 913-449-7589
office: 913-627-0351

[#3]

We image 8 or so at a time in about 45 minutes for all of them. It doesn't
seem to increase time that much when we add more computers b/c it seems
that the network is the bottle neck as no single client is maxing out the
afp copy pull when it is in file copy mode. We have a JSS on one server
and the main CasperShare on a second server and the CasperShare is on a
RAID. Both servers are on a gigabit switch and the clients are off that at
100 mb. Block copy can be up to 40 - 50 Mb/s and then file copy drops
down. We Netboot off a 3rd server that has lots of RAM on the same gigabit
switch.

I'm curious to see if a "dumb" switch would change our time, that is
curious to me. Maybe I'll try it this summer.
-Nathaniel

Nathaniel Lindley

++++++++++
Learning Systems Specialist
Educational Technology
Saint Paul Public Schools
Saint Paul, Minnesota
nathaniel.lindley at spps.org
phone: 651-603-4929

[#4]

The only problem I've found at times using an un-managed switch is the amount of time it takes a link to come alive without PortFast. I've seen NetBoot fail because the switch did not activate the link fast enough. I've used the HP ProCurve 1400 switch before with great success.

Steve Wood
Director, Information Technology
swood at integerdallas.com

The Integer Group | 1999 Bryan St. | Ste. 1700 | Dallas, TX 75201
T 214.758.6813 | F 214.758.6907 | C 940.312.2475

[#5]

What's the infrastructure? switches? servers?

[tlarkin]

I have seen this exact thing, if you boot holding down the opt key
instead of N, it will give you the netboot server option to boot from,
but you may need to wait an extra 10 to 15 seconds.

Thomas Larkin
TIS Department
KCKPS USD500
tlarki at kckps.org
cell: 913-449-7589
office: 913-627-0351