To CAS or not to CAS....

I have been designing and deploying configuration manager for 12 years. In recent years I have been lucky enough to be working on some large scale SCCM projects.  So what is large scale to me? over 40,000 seats goes into the "large scale" category for me.  The CAS is nothing new to SCCM in 2007 we had the idea of the central admin site and in many ways it acted in the same way as the CAS. The CAS or Central Admin Server is basically a central admin location where you can manager multiple primary sites in one console.

When SCCM 2012 was released the CAS had to be installed during the initial install phase and as a result a lot of people deployed a CAS as a just in case design.  The CAS then received a lot of negative press because it implemented a layer of complexity that many didn't understand.  I am going to cover the physical things to do when deploying a CAS in another blog but in this blog we are just going to look at some of the design decisions you may go through when choosing to deploy a CAS.

A single Primary Site (PS) can support up to 100k clients and I have never deployed a SCCM environment with 100k seats but have installed a CAS 3 times so was I wrong to do that?  So let me give you a profile of 1 install.  The customer was a retail customer with a total of about 80k seats.  The client was expecting about 5% growth over the next 3 years and they have a support goal of never going past 80% of the Microsoft supportable limit and so we needed a CAS to support more than 1 PS.

My next install was for 45k and that is well within the 100k limit with a growth rate of about 15% over the next 3 years, with this deployment we still went for a CAS and 3 PS servers.  So if the CAS includes more complexity and more databases and servers then why chose a CAS?

It came down to the following points;

If I have a single PS server supporting say 50k clients and it goes down then we can add no new software, clients or OS images.  No new policies can sent to the clients and this means its a pretty high single point of failure for a large org.

In this example we have 1 CAS and 3 PS, the 3 PS were in the US, EU and APAC.  With this design we could loose the CAS and 2 PS servers and still have no issues with the last PS working.

There is no doubt that with the changes to SCCM SP1 and the ability to add a CAS after installing the first PS that for smaller sites its going to be more likely that more smaller sites will not go for a CAS to begin with unless you really need one, and who should decide that ... well you.

In SP1 there are also some good changes to how we can view what is happening from a site replication standpoint from within the monitoring component.