I don't remember a whole lot from the class on Systems Administration that I needed for my degree. Mainly because it was a matter of rote memorization, and that semester was pretty programming-intensive. But one of the few things that pops into mind from a decade ago is the set of options and trade-offs involved for cutting over from one software system to its replacement.
Option 1: Parallel systems transition Both systems are fully operational during the cutover time. The good: Output from the new system can be validated against the old one. Plus, if the new system isn't working as planned, it's simply switched off until the bugs can be worked out. The bad: Something like double the work (i.e. expense) will be involved, not to mention a certain amount of organizational disruption while things are sort of in wait-and-see limbo.
Option 2: Cold turkey cut-over You switch off the old system and switch on the new. The good: No double-expenditures, and the sheer act of cutting over is its own incentive to make things work. The bad: Not for the faint of heart, because if the new system is not working as designed, firing up the old one and bringing it back up to speed with the new data is pretty much guaranteed pain.
Seem pretty simple, don't they: The trade-offs between cost and peace-of-mind? But it boggles my mind to see how many people--even those who should know better--think they can have Option 1's safety-net for the price of Option 2. It just doesn't work that way, and I can confidently tell you that anyone who says otherwise is either a naif or a liar. Don't fall for it. Whichever cut-over option you choose, decide up front how much pain you and the organization can handle. Budget (time as well as money) for hiccups and for pete's sake make sure folks know the backup plan (you do have a backup plan?) well ahead of time.
This is probably old news to the veteran of even a single rollout. (Although I've noticed that cut-overs are like child-birth for some people in that you forget how gut-busting it can be until the next time around.) But for anyone who hasn't ridden that roller-coaster...don't say you weren't warned.
Thoughts on computers, companies, and the equally puzzling humans who interact with them