Every so often, a client of mine will ask me about parallel processing. Generally they’ll ask something like “How long should we be running both our current AMS and our new AMS in parallel?” My answer is always “Don’t bother with parallel processing.” Here are three reasons:
- Staff never actually does it. Think about it; you’re asking data processors to double work. Are you giving them twice as much time to do it, or taking half of their other work away from them? Probably not. The end-result is that in real-life, parallel processing rarely happens.
- You can develop test scenarios that mimic real-life processing, which will allow you to test the new AMS without having to do parallel processing. This is easier, less time consuming, more focused, and thus more likely to actually get done.
- Parallel processing has a tendency to cause the users to focus on how things are done now, vs. how they SHOULD be done in the new AMS (i.e. business process changes). By changing the focus to testing on the new AMS, you’ll also give the users a chance to really think about current business processes and how they should change with a new AMS.
Back in the day, one almost had to do parallel processing, due to technology restrictions (i.e., it was too expensive to run test environments). That restriction no longer applies, and in my experience, parallel processing is rarely effective.