dangriffin11
Programmer
Does anyone have any idea what the "relative cost" of a checkpoint in an IMS/DB2 update program is? I suspect this may be one of those areas in which people have been coding programs the exact same way for 25 years without thinking about whether the old methods still make sense. The old rule was to take checkpoints relatively frequently in order to free up resources and lock in updates, but not too frequently because performance would suffer. My question is whether, with all of the improvements over the last 25 years, the performance issue still applies. If a checkpoint call "costs" one I/O, will more frequent checkpoints, say after each logical unit of work, end up creating a noticeable increase in program run-time?