Recently I encountered a situation where we had a SaveDataAll process running every half hour that had started impacting the system performance. The environment is largely for enterprise planning and we wanted the protection of having TM1 saved to disk regularly through the day to protect against any unforeseen outages. This was largely able to be executed in a couple of seconds and thus had no real impact.
Then we added a large analytical cube that took the TM1 system up by about 15Gb and thus blew out the SaveDataAll to a few minutes which in turn had an impact on users ability to interact with TM1. The solution to this little problem was to implement a CubeSaveData TI for the planning cubes only, rather then use SaveDataAll for the entire environment. This is now scheduled to run every 30 minutes and the CubeSaveData is set for the end of each day and after the morning load chore.
It is also worth noting that IBM has changed the behaviour of the locks in both SaveDataAll and CubeSaveData to limit the impact on users. More information on this change and on the impact of locking can be found on IBM’s website here.