Browse Prior Art Database

A Scheduling Algorithm For Processing Mutually Exclusive Workloads in a multi-system Configuration

IP.com Disclosure Number: IPCOM000015826D
Original Publication Date: 2002-Aug-19
Included in the Prior Art Database: 2003-Jun-21
Document File: 2 page(s) / 93K

Publishing Venue

IBM

Abstract

Disclosed is a program algorithm that provides a mechanism for scheduling workloads in a multi-system configuration where: (A) only one workload may be active in the multi-system configuration at any point in time; (B) once a workload is started, it can be expanded to include new work; (C) no one workload can expand indefinitely and thereby prevent other pending workloads from executing; and (D) a reasonably predictable execution sequence of workloads occurs.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 52% of the total text.

Page 1 of 2

  A Scheduling Algorithm For Processing Mutually Exclusive Workloads in a multi-system Configuration

   Disclosed is a program algorithm that provides a mechanism for scheduling workloads in a multi-system configuration where: (A) only one workload may be active in the multi-system configuration at any point in time; (B) once a workload is started, it can be expanded to include new work; (C) no one workload can expand indefinitely and thereby prevent other pending workloads from executing; and (D) a reasonably predictable execution sequence of workloads occurs.

A workload can consist of one or more functions that may execute serially or concurrently in the multi-system configuration. A workload may consist of: (A) a single function that can only be performed by a single requester in the configuration; (B) a single function that can be performed concurrently by multiple requesters on multiple systems in the configuration; or (C) multiple functions that can be performed concurrently by multiple requesters on multiple systems in the configuration.

The algorithm provides a "distributed scheduler" where no one system in the configuration has sole responsibility for scheduling and dispatching workloads. Rather, all systems participate in concert to determine what workload should be dispatched when the current workload completes. This model simplifies recovery processing when a system exits the configuration.

The distributed scheduler uses a "scheduling queue" to represent active and pending workloads. The "scheduling queue" consists of two elements:

A common "Workload Scheduler Queue" that must reside on a media that each system in the configuration can access and update in a serialized manner. In order to limit (bound) the "common" resources associated with this queue, the queue element is the workload category (rather than individual requestors). Thus, the maximum depth of the queue of "pending workloads" is the number of workloads defined by the implementation. The workload queue consists of two queues: an active queue and a pending queue, defined as follows:

Active Queue - The active queue is a single-element queue that is used to determine what specific workload is currently being processed. Pending Queue - The pending queue is a multi-element queue of "n" queue elements, where "n" is the finite number of workloads that is defined by the implementation. The pending queue is used to keep track of what workloads are waiting to process. T...