> home > services > policies > capacity
Capacity Disincentives

The major difference between capacity and capability jobs has been defined as the number of processors used. Similarly, the major difference between capacity jobs and development jobs is the amount of time they run for, ie, development jobs are usually much shorter. For this scheme, we define capacity jobs as jobs which use small numbers of processors for large amounts of time. For the start of service, we will set the levels to define capacity jobs as those which used fewer than 32 CPUs for more than 1 hour. These levels could be varied over the course of the project to encourage users to use more scaleable codes and algorithms.

Having defined capacity jobs, there are a number of disincentives we could apply to them:

1. Set up a queue structure which does not allow capacity jobs or which only runs them when no other jobs are available.
2. Charge a premium rate for capacity jobs, eg, 10% above baseline.
3. Charge capacity jobs for a minimum number of CPUs irrespective of the number actually used, eg, capacity jobs using less than 32 CPUs would be charged to users as if they used 32 CPUs.

Based on our experiences of running other capability services, we believe that a combination of options 1 and 3 will be effective in discouraging capacity usage without alienating users.