The technology currently exists now to do some pretty cool stuff in the high availability DR Data Centre spaces.
I have recently been pondering some scenarios, and some things have come up that I have previously not considered.
Subject to bandwidth availability you can now easily build a virtual data centre (live-live site instead of more traditional prod-DR) across two or more sites using modern SAN and server virtualisation technologies.
We first played with geographically dispersed clustering in 2002, and back then it was painful, expensive and unreliable, now it is fairly simple.
So just because you can does that mean you should? This has been the pondering for me this week.
So my thoughts on this are that you had better have really good data centre facilities before heading down this route, many organisations have good primary facilities, but there DR sites are often not up to scratch, so would actually increase the chance of an outage due to a failure in the DR facility that would impact production. Straightforward enough, OK spend some money, get the DR facility up to scratch way you go, or should you?
So this is my second pondering for the week, what if one of the two data centres was in the organisations primary site with 80% of their staff? A couple of scenarios here, what if we loose the primary site, OK we go to DR with minimal impact or down time, but 80% of our staff can no longer work. A second scenario is we loose the server room but not the site, as comms often terminate in or near the server room, lets assume we loose that too, so now we have the DR site running but 80% of the staff cannot access it.
Given the above scenarios how much effort should be put into such a virtual data centre design? Not as hard a decision if both data centres are dedicated facilities with the majority of staff in other locations, in that case go for it.
Anyway there you go.