Your Browser is not longer supported

Please use Google Chrome, Mozilla Firefox or Microsoft Edge to view the page correctly
Loading...

{{viewport.spaceProperty.prod}}

Setting priorities

If you distribute an IT workload across several different guest systems, the degree of concurrency of the individual components of the load changes. This means that priorities sometimes need to be changed. The following example illustrates this.

The following load components are running on a server in native operation:

  • one UDS/UTM application

  • multiple batch database applications

  • a number of dialog tasks involving program development with no access to the databases

Up to now priorities and CPU utilization have been as follows:


UDS tasks

UTM tasks

Batch

Dialog1

Dialog2

Task number

1

8

2 - 6

3

10

Priority

150

150

255

180

210

Workload (%)

20

30

10

5

10

The IT load is now to be distributed over two guest systems as follows: The developer activities are to be outsourced to a separate guest system. All other components should run on another guest system on account of their shared databases

VM

Name

CPU quota

Load

VM1

TP system

80

UDS/UTM + batch

VM2

Dialog system

20

Dialog1 + Dialog2

The fact that the dialog load has been relocated means that the situation for batch tasks improves, since these can now always run when the TP load is waiting for inputs/outputs. The throughput of CPU-intensive batch tasks improves, since they tie the real CPU to the guest system.

Systems used purely for dialog tasks generally have a high proportion of idle time. The low CPU quota means that the dialog system will have to wait for a relatively long time after each idle state. This means that the response times for the dialog tasks may increase.

The preference given to dialogs with a high priority compared with those with a low priority also decreases. The following measures are necessary if the same results are to be maintained after the load has been split:

  • If the dialog response times increase dramatically, the CPU quota of the dialog system must be raised. The TP response times and the batch throughput must be constantly monitored.

  • The difference in priorities between Dialog1 and Dialog2 must be increased.

  • The priorities of the UDS and UTM tasks must be increased in order to restore the original difference to the batch tasks, which now have a greater priority.