locked
New site question RRS feed

  • Question

  • So I am a little new to SCCM 2012 and have been tasked with rolling out a fresh infrastructure to about 20k devices. I've had the product up in a lab for about 5 months now and know the ins and outs pretty well (I also have used SMS 1.2, 2.0 and 2003 previously).

    So my question is (and I know this might be ambiguous), what is the ideal infrastructure for a site my size? Not all 20k devices are in one location, so I was going to do something like this:

    • 1 primary site (SQL on separate box) -- Primary site would run a Distribution Point, Management Point and Reporting Services Point
    • I would then put 1 DP in all the sites with both large amount of clients or slow WAN links.
    • I would like to roll the client out using WSUS

    Now the catch here is that we have an existing SCCM environment up that is already published so automatic client site assignment won't work for my site. And no, the intent is that the environments will remain separate (internal political issue).

    With all of that, is my infrastructure layout going to be efficient or do I need to expand management points to the larger sites for console access or some such?

    Thanks in advance for any guidance.

    Thursday, December 4, 2014 2:04 PM

All replies

  • Hi,

    It is really hard to say without knowing more, WAN links, number of DPs, e.tc. but in theori it should work just fine. I would probably run the SQL on the same server depending on the hardware available to gain performance and keep the environment easier to manage.

    For the console question, the console connects to the SMS Provider and not the Management Point so from that perspective there is no need for more MP's, also keep in mind that all clients will evaluate all MP's and choose one randomly.

    Regards,
    Jörgen


    -- My System Center blog ccmexec.com -- Twitter @ccmexec

    Thursday, December 4, 2014 2:19 PM
  • The admin console uses the SMS Provider, not the MP.

    It's difficult to answer your question, because there are so many details missing. Only someone that knows the environment can design it correctly...

    If you have 20k clients you may want a few MPs and DPs. When you say that not all clients are on the same location? Are the locations connected with high bandwidth? If not look into secondary sites (depending on the bandwidth, # clients, etc).

    You can install WSUS & SUP on the primary site server. For the amount of clients perhaps, host the DP & MP (a couple at least) on remote systems. I would prefer seeing the DB on the primary site instead of DP/MP. Thsi is assuming that your server can handle the load.

    Look at the requirements for the DP/MP to see how many clients they can support - i'm feeling lazy :) They are easy to install.

    Thursday, December 4, 2014 2:20 PM
  • Yeah I knew this would be a tough thing to answer. Especially since I don't even have it set up yet.

    I had to spin off the SQL part due to some internal limitations .. is performance going to take that much of hit?

    Thank you for your reply.

    Thursday, December 4, 2014 2:26 PM
  • Thank you for your reply. I knew this would be difficult to answer but wanted to get a general idea if I could.

    The remote sites with a large client base have large WAN pipes so I was thinking that just using a DP would suffice. Worse case I figure if that doesn't work I can always put in a secondary site.

    The sites that I have on slow WAN links don't have a high client count (100 or less)... is a DP sufficient for those?

    As far as client requirements, each MP can do 25K clients... but for performance would I be better off putting a MP on the larger remote sites or is that overkill?

    Thursday, December 4, 2014 2:30 PM
  • It's OK to host the DB on a remote system. It's just easier to host it on the site server. Keep in mind that the Primary Site server will need to be local admin on the DB server and it is going to install some services on it.

    As long as the bandwidth between the DB server and the Site Server are good, you shouldn't see an impact. Make sure that the hardware meets and surpasses the minimum requirements. SQL performs better with lots of RAM (e.g. 16GB+).

    For 20k clients you should have adequate hardware on your servers... You're not running a tine environment.

    Thursday, December 4, 2014 2:31 PM
  • Thank you for the response. the SQL box I have has a lot of horsepower which is why I broke it off to it's own server.

    Thanks again, based off of your response I think hardware wise I am in good shape.
    Thursday, December 4, 2014 2:38 PM
  • if you have adequate WAN links, that DPs are better than Secondary Sites. Make sure that you don't have more clients than supported by the DP.

    With the DP, you can throttle and schedule content replication, but not upwards flow. This is where the secondary site fills the gap if the WAN links were small.

    Can't say if 100 clients across a slow link would be best with a DP or a secondary site. It depends on how much content is flowing up to the primary. Normally the biggest concern is the initial 10-20 mb software updates catalog info (incremental afterwards). I think that you should be fine. You may want to look into throttling/scheduling those remote DPs to ensure that they don't take up all the bandwidth especially during work hours.

    There is no point on putting MPs in the remote sites as they are randomly assigned (unlike DPs). You could create bigger issue, where clients form the fast primary site get assigned to an MP in a remote site!

    Only install MP in a remote site if using secondary sites (mandatory).


    Thursday, December 4, 2014 2:40 PM
  • Brilliant .. this is exactly what I needed to see.

    What I think I will do then is start with DPs. If that turns out to not be adequate I can always upgrade those servers to secondary sites as needed.

    And I have looked in to throttling and such and have a plan for that. Thank you for the suggestion ... that should in theory do the trick. If not, then I'll stand up a secondary site.

    Thursday, December 4, 2014 2:43 PM
  • Thank you for the response. the SQL box I have has a lot of horsepower which is why I broke it off to it's own server.

    Thanks again, based off of your response I think hardware wise I am in good shape.

    There are also recommendations that you may want to consider when configuring the SQL server/DB, considering the number of clients you have. Look into these, there are blogs online:

    - set the min/max memory for the SQL Server

    - add additional files to the: tempdb (data+logs), CM db (data+logs). 

    http://blog.coretech.dk/kea/system-center-2012-configuration-manager-sql-recommendations/

    Thursday, December 4, 2014 2:44 PM
  • Thank you for the response. the SQL box I have has a lot of horsepower which is why I broke it off to it's own server.

    Thanks again, based off of your response I think hardware wise I am in good shape.

    Keep in mind that Remote SQL will make thing a ton harder to troubleshoot.

    As a general rule, I add 15 days to any project when SQL is remote from CM12 primary site.


    Garth Jones | My blogs: Enhansoft and Old Blog site | Twitter: @GarthMJ

    Thursday, December 4, 2014 3:40 PM
  • Now the catch here is that we have an existing SCCM environment up that is already published so automatic client site assignment won't work for my site.

    One idea might be to learn from what is there already and shrink it using 2012's new capabilities.

    So for instance making Secondary sites, DP's instead.

    Thursday, December 4, 2014 4:17 PM