SCIENTIFIC PROGRAMMING JOURNAL

Call for Papers

Special Issue on Dynamic Computational Workflows: Discovery, Optimization, and Scheduling

Guest editors:

David W. Walker, Cardiff University, UK

Ewa Deelman, USC Information Sciences Institute, USA

Scientific Programming invites the submission of papers for a special issue entitled Dynamic Computational Workflows: Discovery, Optimization, and Scheduling.

Workflows are widely viewed as offering a useful representation of computational applications that are to be run in a distributed environment. A workflow can be conveniently displayed as a directed graph in which nodes represent services and edges represent the flow of data and/or control between them. Here the term “services” is used in a generic sense to refer to any sub-unit of computation – often these are implemented as large grain-size Web services or other application components. A dynamic workflow can be described as a workflow that may be changed after it has been designed, either at runtime or prior to invocation. Various opportunities exist for workflow optimisation, and this process is related to both service discovery and workflow scheduling. A common example of workflow optimization arises when multiple equivalent services are available to fulfil a service request. In such cases semantic support to recognise equivalent services is required. In general, a workflow may be composed of abstract services that get bound to concrete service implementations either when the workflow is initially designed, or at any subsequent time prior to service invocation. Three different types of binding abstract to concrete services can be distinguished. In early binding the service binding is done statically at design time and does not subsequently change. In late binding the service binding is done at runtime immediately before a service is to be invoked. In intermediate binding the service binding is done after the workflow is designed but before it is submitted for execution. Although a workflow may be optimised at any of these three stages, delaying the binding of a service as late as possible allows more recent information to inform the optimisation process. It also makes it more likely that unavailable services will not be selected. Late binding performs optimisation on a service-by-service basis. However, intermediate binding allows the workflow as a whole to be optimised, taking into account factors such as the bandwidth between connected nodes in the workflow graph, and opportunities for algorithmic optimizations.

The special issue will focus on all aspects of dynamic workflows, including, but not necessarily be limited to, the following topics:

• Predictive scheduling

• Performance-based middleware

• Semantic support for service selection

• Web service discovery and optimisation

• Workflow optimisation

• Workflow restructuring

It is anticipated that most of the papers comprising the special issue will present original and unpublished research results. However, review and work-in-progress papers will also be considered. Review papers should include cogent analyses of the relevant issues relating to dynamic workflows, as well as discussion of the technologies and application requirements driving this development. Work-in-progress papers should place the work presented in the context of a coherent program of research, and in relation to more general service-oriented workflow issues.

Schedule and Deadlines

The schedule for the publication of the special issue is as follows:

16 April 2007	Deadline for submission of papers
31 May 2007	Notification of authors
27 July 2007	Final submission of accepted papers

It is hoped that the special issue will appear before the end of 2007.

Submission of Manuscripts

Papers should be prepared in accordance with the instructions given at the Scientific Programming web site at http://www.iospress.nl/html/10589244_ita.html, and submitted directly by email to one of the Guest Editors:

David W. Walker (david@cs.cf.ac.uk)

Ewa Deelman (deelman@isi.edu)

For any further information about this special issue please contact either of the Guest Editors.