- In the case of applications which are not CPU-intensive (e.g. OLTP mode or batch mode with a large number of I/O operations of other interruptions), there is a need for a large number of generally only very brief intervals in which CPU performance is required ("CPU time slices").
When systems are utilized significantly less than 100%, normally only a few of these short CPU time slices are used immediately one after the other; the CPU then switches to the wait state (idle).
In the case of TP loads severe fragmentation of the CPU requirement into small CPU time slices must therefore be reckoned with, as well as frequent, brief wait states of the CPU.
The length of these CPU time slices decreases as the CPU performance increases (faster processing) and as the multiprocessor level increases (shorter queues).
As a result of this, the number of VM2000 scheduling operations rises as the CPU performance increases and as the multiprocessor level of the VMs increases. The increased throughput and reduced response times thus lead to a slight increase in the VM2000 overhead.
For OLTP applications it is therefore recommended that the priority controls between the DB server and DB client tasks be set in such a way that the job log of the DB server tasks is always full. This avoids overhead through additional communication.