Our only Distribution point is saturated, will adding another locally help?

[powellbc]

From what the documentation says, multiple distribution points are useful in scenarios where you have remote offices, etc. My question is, if we have a single DP on our campus, will adding another in the same geographic location one help distribute the load? One area regularly images 20–30 machines at a time and network utilization on the server is often running at 70–90%. This impacts others trying to do one off jobs as well as slowing the entire process down.

My hope is adding another DP will allow large numbers of machines to be imaged with less slowdown.

[dpertschi]

Definitely, there are a number of ways you could deal with it, splitting the load by network segments, and defining which segments use which DP.

With mass imaging being a concern, I'd consider reserving a high bandwidth DP (10K drives, gigE to the desktop) for imaging, on a separate switch from the general population.

[powellbc]

Thanks for the info.

We have 2 Netboot Servers which are not doing anything else. Does it sound like a good or bad idea to use one of those?

[Chris_Hafner]

I suppose that depends on your network. Are you maxing out the connection from that specific "geographic location"? If you're server is putting it's back out and not saturating the uplink then sure. However, if you're maxing out what your network can handle then no.

Here we are a little overpowered for our environment. I run two Mac pro's with Areca hardware RAID cards with 3 stripped Cheeta 15k7 SAS drives. Each array hosts a netboot and DP. My main JSS, is the secondary (load balanced) NetBoot, but is the DP. My secondary server is the Primary NetBoot, secondary DP as well as being the SUS. Both units have their enet ports trunked into our core switch.

With that said, it's not uncommon for us to be imaging 50 plus computers at a time while running a few hundred machines installing various items from the DP's via self service. Everything runs at the speed of the network at that point. I'll be adding another DP... but mostly because I was born paranoid.

[powellbc]

My biggest concern is as of now the only machines we have available to work as DPs are Mac minis. I am not sure if they are up to the task.

If we simply add more DPs, how do clients pick which to latch onto? Is it random or do you have to use network segments, etc. and assign them?

[Chris_Hafner]

Ton's of places run Mac Mini's and swear by them. As for load balancing there are many many options. I'll not say anything about automatic load balancing at the moment (going for my CJA in three weeks so I don't know the best practices). However, you can set all of your DP's, NetBoot servers and SUS's by network segment, department, building or pretty much whatever else you can think of. From those I talk to here, it's certainly more common that my solution. I'll admit that in a few years time I'll probably shift the JSS to a VM somewhere and run Mac Mini's as DP's.

[bentoms]

It all depends on the numbers & network architecture I suppose.

We use one x Mac mini server to upto 100 clients.

You can setup distribution point failover & load balancing. So that if a DP is busy or down, it will use the other.

Works well in my experience.

[powellbc]

Thanks everyone for the great info. I think I am going to set up a second DP on one of the netboot servers, and set it as the failover from the Master DP.

I'll report what I see in practice.

[powellbc]

The Sync has been going on for 2 days. It appears that the files are being routed from the Master DP to the other through my local machine. What????

[nessts]

if it were me and the servers are OS X servers or linux servers i would learn how to use rsync between the two servers and not use Casper Admin to rsync the servers.

rsync -av --progress --delete --inplace /Volumes/Data/CasperShare distributionpointDNSname:/Volumes/Data/CasperSahre

and then make a cron or launchdaemon depending on what OS you are using...

[mm2270]

FWIW, we have almost all Mac minis as our distribution points around the globe (about 6) and we're serving between 4 - 5k clients, so yeah, they may be up to the task in your situation. Granted, we don't use Casper Imaging, so large imaging related files aren't being pulled from them (nor NetBoot) so that accounts for a few things.

[powellbc]

The 2 machines are Windows and OS X respectively.

Next time I am going to run Casper admin from the OS X server to see if that speeds things up.

[powellbc]

Actually, I am going to try Rsync to pull the files from the Windows share. The Casper Admin initiated sync took over 2 full days to complete, and running it again seems to want to resync everything again, so it simply is not a practical option.

Should I sync all folders in the Master DP share, or can I skip the Casper Data folder? I noticed the sync form Casper Admin did not move that over.