Jon, see my post here: http://communities.vmware.com/message/2191541#2191541
I don't want to use multisite because I don't see a need for its complexity relative to what we 'get' out of it in return. Multisite will involve multiple SQL databases for something like SSO which is really such a tiny/non-intensive/non-chatty database app to begin with, and the multiple servers that inherently go with them. Then they start talking about custom sync jobs being setup between each SSO database and it's like, 'What for? Why go through all that?'. I'm only speaking for US/my shop, because in some environments it might make a lot of sense. But for us personally? No... not even SSO in HA mode makes sense where 'return on effort' is concerned. We have the tools/infrastructure to afford ourselves a fairly high level of both site and local highly available SSO for our vCenters to point to over fairly high speed, redundant links between the SSO server and the Infrastructure/Web/vCenter servers that will point to it remotely (well, 3 remotely and 1 within same site as SSO box).
So we're gonna keep it nice and simple---going back to the whole 'KISS' acronym.
Yes, the SSO database will reside on a server other than the SSO server itself (in a MSCS/SQL cluster). The whole thing about creating the db users and options therein is reasonably well explained in the guides, but in the end despite its being somewhat convoluted, it seems quite workable on paper and so you take that and test it in a lab. After all, all you need is vCenter 5.1.0b installer which incl all components, and a SQL server you have sa/dbadmin rights to---no need for vCenter itself at all... then you get your steps wired tight where db rights/tables/creation etc are concerned.
Yes, do yourself a favor and don't trust the installer/upgrade to remove Linked Mode itself (which it claims it will do) but I'm not even going to test that in the lab and our steps will assume it CAN'T. We will re-link things ourselves, thank you very much. Linked mode can be goofy/weird enough to add that little factor in to what is far from a trivial upgrade. Yeah, we'll get that one ourselves.
So... test test test in a lab and figure out what might work best for you. I started that new thread after having yet more doubts about setting up SSO in anything more 'elaborate' than a single server (setup as 'primary node' of course so we can always opt to 'scale' at a later date should it become necessary).
