Hi Guys,
I have been spending a LOT of time researching SCOM 2012 in all areas and aspects. I have carried out various sizings, looked at best practises, referenced architecture etc and believe I am getting some good ground and putting initial designs together. The design as I look at it is pretty excessive in terms of number of server roles and placements. To give you an idea of my sizing, I am looking at 1,500 windows servers, around 40 UNIX servers and potentially 1,000-2,000 network devices located in multiple locations globally. Sizing tools calculate around 3.4 TB of DB space required for both DBs.
ATM, my design incorporates..
Primary
4 x DB servers (2 for Ops DB, 2 for DW)
2 x Management Servers (Wintel)
2 x Management Servers (Networks)
2 x Management Servers (Unix)
1 x Reporting SRS Server
2 x Gateway Servers (Sit in my DMZ)
DR Site
1 x DB Server (Ops DB)
1 x Management Server (Wintel)
1 x Management Server (Networks)
1 x Management Server (Unix)
This totals 19 servers/VMs for the solution with DR which looks a tad excessive to me and I will probably struggle to justify it. I have some questions in several areas I would be greatful for some clarification/advice on.
1) Is it wise to store the Ops DB and the DBWH on the same SQL cluster, or should I be looking to split these out into their own clusters.
2) I have read somewhere in one place that if monitoring network and unix devices, that these management servers write only to the DW database. I have not read this anywhere else and can't confirm this. Is this true, or do all devices write to the Ops DB (and DW DB simultaneously for historic purposes)?
3) Should I look to set the SRS reporting DB on either the Ops DB cluster or the DW cluster, or should this run on its own dedicated instance of SQL on the reporting server?
4) Can I use the Management Servers in my wintel resource pool for the Unix communication also, or should I have seperate pools? (Wintel pool, Unix pool, networks pool etc).
5) I am initially looking at having 2 management servers for wintel in my primary DC with a third management server in my DR facility (seperate with dark fibre, less than 1ms latency). Is this overkill? i.e. could I just have 2 management servers, 1 in the primary site and 1 in the DR site?
6) From a DR perspective, I am looking at having a standby SQL server in the DR site with the shipping of SQL logs occuring from primary to DR site. Would I need to seperate SQL servers in DR to match my primary site (1 x DB, 1 x DW) or could I combine the 2 for DR purposes?
7) If I only chose to fail the Ops DB in any DR situation, would this cause me any issues by not having the DW also running in DR?
I think that is all for now, I am hoping you guys can confirm what I am thinking!
Thanks, John.