question

I am doing some preliminary research on providing a mirroring solution, as part of a new project. Management don't want to make a great deal of budget available for this, so I'm being asked to shoehorn it into existing kit (of course, the *project* is allowed new kit!). Essentially, the proposed server for witnessing is a lives in a different VLAN from the new home of the project servers and runs a different version of SQL Server. A firewall enforces separation between the VLANs such that the witness is on the outside, while the project servers are on the inside. I need to know whether I can rely solely on the implicit allow-inside-to-outside rule, or whether I need to allow explicit outside-to-inside traffic (for the witness endpoint port); and whether or not I need to add special rules to prevent the firewall dropping the connection after a certain amount of time. So, I have a couple of questions that I can't seem to find any *definitive* answer on. I'm hoping that they are actually faily straightforward to someone who has actually done some or all of this: 1. Does the witness server instance have to be of the same version (not edition) as the mirrored pair (i.e. can a mirrored pair running on 2008R2 be witnessed by a 2005 or 2008 instance)? 2. How does the TCP network traffic flow between the pair and the witness? 1. Is it a continuous open/half-open connection, with regular traffic flow; or is it more like a 'ping' that does the full open/close each time? 2. Which server 'initiates' the connection — do the paired servers connect to/ping the witness, or does the witness connect to/ping the paired servers? Or both, depending on circumstance?