Checkpoint time and periodicity
Posted on 2004-04-21
Checkpoints on one of the DBs in our system happen about every 150 minutes and take up to 30 sec and plenty resources with them. Also the checkpoint time seem to grow with time. What can I do from the software perspective to make checkpoints faster and less resource-intensive (please - no hardware solutions, this is beyond my reach)? Any maintenance procedures like defragmenting DB or drives?
This is the transactional database where only "recent" (newer then several days) records are kept, the older records are deleted daily so I can imagine that the database itself may become pretty fragmented.
Generally - are there any benchmarks on how often the checkpoints should happen and how long they should take depending on the system load and hardware configuration?