Browse Prior Art Database

Priority Control Method for a Resource Partitioning System

IP.com Disclosure Number: IPCOM000084511D
Original Publication Date: 1975-Nov-01
Included in the Prior Art Database: 2005-Mar-02
Document File: 5 page(s) / 38K

Publishing Venue

IBM

Related People

Gilbert, DC: AUTHOR [+2]

Abstract

I. Definitions. The following definitions will be helpful in understanding this description.

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 30% of the total text.

Page 1 of 5

Priority Control Method for a Resource Partitioning System

I. Definitions.

The following definitions will be helpful in understanding this description.

Application: An identifiable kind of work (i.e., set of programs, sets of data) which some number of users can utilize.

Process: The unit of parallel execution to which resources are allocated.

Resource Environment: A subset of the system machine resources (processing rate, storage, etc.) and serially usable resources (data, devices) necessary to perform some work. Machine resources can be partitioned into pools. Processing rate is distributed to processes under a time constraint called granularity, which is a reflection of the responsiveness of the process. A process is allocated processing rate (instruction execution rate) by being given an amount of processing rate and a granularity, or time, within which the process should get the specified amount of processing rate. User Profile: A description of user characteristics. II. Philosophy.

The following discussion assumes an architecture which allows the partitioning and distribution of machine resources to active work units, and above this, a set of facilities for scheduling the availability of such resources to work units. With these facilities, an administrator can achieve work deadlines and performance objectives through appropriate machine resource distributions given these conditions:.
1. There are sufficient resources to satisfy work units

resource requirements.
2. Work units can consume the resources given to them.

Ideally, work unit resource demand specifications are performance objective specifications. In actuality, resource distribution specifications (amounts and schedules) plus resource contention situations constrain the achievement of the performance objectives of work units. The system will sometimes be unable to meet the performance objectives of all work units (as specified by their cumulative resource demands). The system will have to distinguish among work units and, where possible, attempt to honor the performance objectives of some work units with a possible impact on other work units.

The intent is to consider the machine resource distribution specifications as specifications of performance objectives, and to attempt to make the above two conditions true for "important" work units. Priorities assigned to work units will be used to:
1. Resolve contention for allocations of pool resources in

favor of important work units.
2. Resolve serially reusable resource (SRR) contention in favor

of important work units.
3. Cause serialized (queued) services to favor important

requests.

1

Page 2 of 5

Thus, priority becomes the indication to the system of how important it is to meet the performance objectives of a work unit. These objectives, as stated before, are specified in terms of machine resources requirements for the work unit; they are not influenced by priority value, nor is priority influenced by the specified req...