Browse Prior Art Database

A mechanism for work allocation in a resource constrained environment using publish/subscribe messaging

IP.com Disclosure Number: IPCOM000031016D
Original Publication Date: 2004-Sep-07
Included in the Prior Art Database: 2004-Sep-07
Document File: 3 page(s) / 47K

Publishing Venue

IBM

Abstract

This article describes a workload distribution mechanism where every machine in a resource constrained environment behaves in an autonomous manner monitoring and advertising its state using publish/subscribe messaging. This mechanism removes the need for a central dispatcher type mechanism that periodically polls the network of servers to maintain a snapshot of their state. By employing the publish/subscribe model the appropriate workload is automatically directed to a server that can handle it without the need for a dispatcher to do so.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 41% of the total text.

Page 1 of 3

A mechanism for work allocation in a resource constrained environment using publish/subscribe messaging

This article describes a mechanism for work allocation in a resource constrained server network environment. Known solutions to this type of problem include using a network dispatcher type mechanism to fan out workload to servers. One drawback of this approach is that the dispatcher needs to constantly monitor the state and workload of each of the servers. The dispatcher also needs to be aware of each servers existence.

    An approach where the servers can themselves attract the appropriate working load depending on their resource availability without the need for a central dispatcher would be of interest. In addition certain types of workload may only be suitable for certain servers in the environment due to the application needed to process it. A mechanism that ensures that this load is distributed to the correct servers would also be desirable.

    This solution provides a workload distribution mechanism where every machine in a resource constrained environment behaves in an autonomous manner monitoring and advertises its state using publish/subscribe messaging. This technique removes the need for a central dispatcher type mechanism that periodically polls the network of servers to maintain a snapshot of their state. By employing the publish/subscribe model the appropriate workload is automatically directed to a server that can handle it without the need for a dispatcher to do so.

    The technique delegates the responsibility of the selection of a server for dispatching to the servers themselves. Workload automatically gets forwarded to the most appropriate server using the publish/subscribe model without the need for a dispatcher to make the decision. You can say that the message broker acts as the dispatcher in the case where the invention is implemented using the WebSphere* Business Integration Broker products.

    A workload distribution environment can be made up of a number of servers receiving data for processing via a dispatching mechanism. Examples of this type of environment include clusters of web server and the computing grid. The nature of data being processed in this type of environment is not restricted to any particular application types but the workload being handled is likely to have common characteristics like being CPU or memory intensive and requiring access to particular applications and systems.

    The network of servers in a workload environment does not have to constitute identical machines in terms of the physical hardware resource and the applications available on each server. In this type of an environment how does a particular data processing request get directed to a server that currently has the available hardware resources (like memory, CPU and disk space) and the necessary applications to process the workload?

    This technique allows the automatic location of a server resource that is capable of processing a req...