Thoughts on Managing Hundreds of Lab Machines

February 25th, 2008 · No Comments

I have recently been looking into several products to help automate the management of about 150 lab machines that I am tasked with keeping well maintained. The issue with trying to maintain a bunch of lab machines without some sort of automated solution is that you have to go around and touch each machine and spend a lot of time on it when someone wants something done, like a new software product installed or Windows Updates put on.

Enter the automated solution products. These can range from hardware solutions such as Kace Technologies KBOX or software only solutions such as LanDesk or Altiris. I am not going to discuss any specific product just some of my complaints about these types of products in general and their lack of easy to use interfaces.

When I set out to try to find a way to automate the management of the labs, I had an ideal test case in my mind that I wanted to see how each product could do to attain success on that test case. My test case is fairly simple. This test case involves all tasks being applied to specific subset of my network that I can specify (such as “Lab 1″). First, I want a wake-on-lan packet sent to every machine at 2 a.m. or so in the morning. Then, after each machine is booted up I want them to be unfrozen via a script (our lab machines use something called Deep Freeze which prevents changes from being made by users - any changes made are reverted on the next reboot unless the machine is ‘unfrozen’). After they reboot from being unfrozen, I want Firefox 2.0 to be installed on all machines in the Lab. Then I want Windows Updates and Anti-virus definitions updated. Finally, I want the machines to be told to refreeze on the next boot and to shut down.

Now, I don’t think any part of my request is wholly unreasonable, and I also do not think that any part of that request should be overly hard to program given that all the products i’ve looked at maintain an ‘agent’ on the system at all times anyway so they can see what is going on and perform tasks and inventory the system periodically. The issue is that none of the products i’ve looked at so far can do that test case above in a completely automated fashion from one large script or “task”. I’d love to see an interface where I have a series of scripts or software distributions that I want to be able to do and I just drag them into my task in the order in which I want them performed, perhaps inserting “wait” timers for 5 minute periods or other dependencies that verify a registry key or a file exists before moving onto the next step.

All the solutions i’ve been given so far involve scheduling multiple events at different time intervals, or scheduling a wake-on-lan for a particular time period and then having some other script run, followed by some other script etc. It’s just far too complex for what should be a relatively simple task.

At this point i’m almost to where I want to mess around and write my own agent and management piece just to accomplish what i’m trying to do! Problem is, knowing the way software patents are now in the U.S. i’m sure one of these companies has patented the wazoo out of any sort of process for remote management so i’d probably get my pants sued off if I wrote my own product, even if I wasn’t selling it!