I really don t know how many at once. But I know that if you launchs them in a single task, be carefull. Don't start again the task if all the computers don't really finished their provisioning task. The reason is that you will add a new provisionning task on a running provisionning computer. I prefers create copies of scheduled task and add a computer per task. And I ran about 10 tasks, so 10 provisioning deployment.
Hi Giles yes we are aware of that - its caught a few people out here!
I have provisioned up to 30 machines at once. The console can become extremely slow at times but bearable, unless you have the image on the core server. If that is the case I wouldn't do more then 10.
I think the best way is to have an images servers. We are using Norton Ghost to deploy OS images, and we have a different PXE server. So a computer (we will call it PC-1) starts on a PXE server (we will call it SRV-a), then it loads the Windows PXE CD image (which is on the PXE server then do not use LD-Core-server ressources).
At this time, the Computer PC-1 has a windows PE loaded on is own resources, and by using a network drive on the Image file server, it deploys a windows OS partition, without using any core server ressources. Then it will prepare the OS before restarting. At this level, it will begin to use some core server ressources.
Then the computer restarts, launchs its sysprep, then restarts.
Then the computer installs the LD clients and then all the softwares. At this level it uses all the packages it needs, and so it uses core server ressources.
That's the way we use the OS deployment without loosing some core server ressources. We have only one Core server..
I suggest you to control how much CPU is used console-side by provschedule.exe. There was an issue with that component in the 8.8 version that I think was not fixed cause "it works as desired".
If you schedule more than 5-10 different tasks (as you usually do if you launch provisioning selecting a template in WinPE) provschedule.exe hangs; no problem if you schedule, i.e., 100 machines in a single tasks.
In the past I solved this issue writing a service that create a task every 5 minutes with all the machines that make a request during this time window.
In your case please also note that a PXE Rep should be placed in every subnet to avoid PXE image transfer in WAN.
I don't think there is a real-world maximum - I have imaged over 200 computers simultaneously in 5 minutes with LANDesk Provisioning, all pulling from a single Preferred Server @ 6700 megabit throughput. Granted, we had 10-gig fiber to all 52 buildins, the image server had a 10-gig fiber NIC, and every PC was on 1-gig.... but as long as the network and Preferred Servers can handle it, the sky is the limit.
I don't work at that place any more, but our network made LANDesking REALLY fun - SATA hard drives were the bottleneck of all software distribution and Provisioning tasks!
@Jary I actually dont use PXE at all.. I have Dual boot images with LANDesk Provisioning for windows 7 created at build time and a .Net program that will take an XP machine and Edit the boot record, copy the boot files from the preferred server then reboot into LANDesk Provisioning.
Saves us having 500 PXE servers and 1000 for resiliency!
@JSMCPN Now that does sound like fun!! That is a serious network....