Sccm 2012 hierachy consideration

Question

The system consists of 10 data-centers, 200+ remote (geology) sites and has about 50,000 clients in total, with a single forest. 
Now, when come into the plan/design a sccm2012 hierarchy for this system, I have some questions regarding site systems placement in data-centers:
1. Do I need to place a primary site in each of the data-centers? This will result in 10 primary sites with a CAS.
2. Can I place only one primary site in one (the largest) data-center, and place a MP/DP in each of the rest data-centers? This may lead to a stand-alone primary site.
3. Can I place two primary sites in two major data-centers with a CAS, and place a MP/DP in each of the rest data-centers?
4. In case placing one or two primary sites, can I place just a DP instead of MP/DP in each of the rest of the data-centers?
For remote sites with 50+ clients I will consider place a DP over there.
Thanks in advance for your help.

Answer 1

I concur with Narcoticoo (lots of folks agreeing on this thread :-) ) that the consultant didn't know what he was talking about -- I'd even go so far as saying you should ask for your money back.
For the question of "is a secondary preferable for higher bandwidth or a low/narrow bandwidth?", low bandwidth or high client count is where secondary sites are best used. This is because they compress, throttle, and schedule most client
to MP traffic destined for the primary site including state and status messages and hw and sw inventory. The overhead caused by this traffic is quite small though in normal scenarios which is why they aren't strictly required/recommended unless the bandwidth
between the clients at a location and the MP will be over-utilized. This is a subjective standard though because it truly depends on the network involved as well as the client count.
For the question "As a secondary in sccm2012 also has a database (express though), does it cause latency of replication when bandwidth is not good enough?" the answer is not black and white. If the links are truly terrible, volatile, and/or have
extreme latency, like say VSAT connections to roaming ships, then yes, the replication between the secondary site and primary site is an issue. In these extreme cases, the use of third-party, complementary tools is a valid option. Generally though, the trade-off
of less client chattiness combined with the scheduling, throttling, and compression of upward client traffic is beneficial where bandwidth is a concern.
Jason | http://blog.configmgrftw.com | @jasonsandys