With the aid of the SERVICE-QUOTA (S-Q) parameters, the performance capacity of the system is distributed among the categories. The parameters SERVICE-QUOTA-MIN and SERVICE-QUOTA-MAX define a framework within which PCS plans service requirements for the specified category (S-Q-PLN). If a category does not make full use of its share of performance, the remainder is then available to other categories.
By means of the dilation factor (REQUEST-DELAY parameter), PCS recognizes how extensively the tasks of this category are delayed by competing tasks. Dilation thus serves to indicate workload levels and specifies the factor by which the runtime of a job will be delayed if the job is forced to wait for the allocation of a resource owing to multiprogramming operation.
Dilation = Runtime in multiprogramming operation / Single runtime
In the case of the simultaneous use of SM2 and the accompanying SM2 monitoring program SYSTEM for all disks, PCS periodically measures the current I/O time per category, which increases the accuracy of the dilation calculation accordingly.
Within a given dilation range (REQUEST-DELAY-MIN, REQUEST-DELAY-MAX), PCS dynamically adjusts the S-Q-PLN value of the category in accordance with the current load (REQUEST-DELAY-ACTUAL).
SERVICE is distributed over the various individual categories according to the following hierarchy:
Categories with a dilation range and max. share of performance of 100%
These categories are allocated the performance capacity to which the current dilation entitles them before all other categories (“planned values”).
Insufficient capacity:
If several such categories exist and if the sum of their planned values exceeds 100%, the planned values are reduced proportionally so that they total 100%.Residual capacity:
If the tasks of the categories with a dilation range and max. share of performance of 100% are not in a position to use the planned capacity, the (current) residual capacity is made available to the other categories with and without dilation range. (Exception: full global response-time orientation).Categories with a dilation range and max. share of performance < 100%
These categories are allocated the performance capacity which is left over by categories with a dilation range and max. share of performance of 100% and to which their current dilation entitles them, before categories without a dilation range.
Insufficient capacity:
If several categories with a dilation range exist and if the sum of their planned values exceeds the capacity still available, the planned values are reduced proportionally so that their sum is equal to the available capacity.Residual capacity:
If the sum of the planned values is less than the available capacity, the residual capacity is planned for use by categories without dilation range.
If the tasks of the categories with a dilation range are not in a position to use the planned capacity, the (current) residual capacity is made available to the categories without dilation range. (Exception: full global response-time orientation).Categories without dilation range
Categories without dilation range are allocated the performance capacity which is left over by the categories with a dilation range. This capacity is distributed proportionally among the categories without dilation range according to their defined maximum shares of performance.
In order to take full advantage of the regulation functions of PCS, at least one category without a dilation range should always be provided.
The function “automatic category changeover” serves to improve differentiation between load requests with high resource requirements (long-running jobs) and those with low resource requirements (short-running jobs). In this case the user employs the DURATION variable to specify how many SERVICE UNITs may be used per processing step before an automatic changeover to the following category (NEXT CATEGORY) with a “weaker” performance capacity is to take place.
SERVICE UNITs are the weighted sum of the work carried out by the resources CPU, IO and MEMORY.
By means of the THROUGHPUT-QUOTA parameter, a percentage indicating the ratio between response-time and throughput optimization within the category is defined. The effects of this are as follows:
| THROUGHPUT-QUOTA | response time optimization | throughput optimization | ||
|---|---|---|---|---|
| full | full | |||
| 0% | <50% | >50% | 100% | |
For systems with TP or interactive mode, response-time optimization is of more importance; THROUGHPUT-QUOTA=0 is therefore recommended in the corresponding TP categories as well as in the DIALOG start category (short-running jobs).
For systems operating predominantly in batch mode, the most important performance goal is the greatest possible throughput together with high utilization of the resource capacity. For throughput-oriented categories, THROUGHPUT-QUOTA values greater than 50% should be provided.