JISC

Since the mid 1970s, computational scientists have seen an explosion in the power and functionality of the high-end systems to which they have had increasing access. Most developers of scientific software are now familiar with getting optimal performance by linking their bespoke applications to one or more specialised and highly tested libraries, for numerical algorithms, visualisation or data management. Several changes in paradigm for application development and execution have taken then from proprietary to open-source numerical libraries and to shared services. There is therefore a need for a lightweight client Grid toolkit to offer the scientists transparent access to the Grid.

Such a requirement was identified in late 2003 and resulted in the creation of GROWL (Grid Resources on Workstation Library) which utilises WS-I technology to provide a web service layer to grid middleware as a programmatic VRE interface.

Aims and Objectives

In this project we address the uptake of Grid-based computing and distributed data management by focussing on issues which may hinder or facilitate end-user application development. We refer to the difficulties identified as the ``client problem'' and suggest a solution to build upon the existing prototype GROWL library to produce a truly lightweight extensible toolkit which complements other solutions.

What is required is something that is easily installable but with extensible access mechanisms for Grid resources. It should obviate the problems associated with institutional firewalls for Grid protocols by implementing a client-side (polling based) strategy. It is an alternative to a Grid portal, but can share the same underlying services via an internal Web service interface layer and therefore is a possible interface to the facilities in a Virtual Research Environment.

Using C, it is possible to interface easily to ``heritage'' scientific applications written in C or Fortran. It is also possible, with a little additional work, to create wrappers to interface to Perl, R or other popular environments such as Python. It should be possible to install the GROWL library quickly on a variety of client workstations running Linux or a similar UNIX-like o/s with a minimum of additional software.

Evaluations will be done for applications from the research domains of bioinformatics, physics, chemistry and social science.

Project Methodology

The GROWL library will be structured in a modular was so that only the required parts need to be installed for a particular application. The software is "layered" with services located on named remote servers as specified in Web service WSDL XML description documents and these in turn accessing computational and data resources on the Grid. The client layer on the user's workstation comprises of the chosen C library modules and possible API interfaces to other languages such as Fortran, R and Python. This will enable remote resources to be accessed from existing "heritage" applications using familiar procedure of function calls to the Web service middleware.

Deliverables

The GROWL project will deliver a C library containing client functions to a variety of Grid-based resources using middleware such as Globus (GSI for authentication, GRAM for job submission and GridFTP for file transfer), NetSolve, SRB wrapped as Web services for ease of deployment and use. The work to develop services and client interface will be prioritised with the project partners to meet their requirements.

Stakeholders

Stakeholders in the project include:

• Participating institutions and projects - will be provided with the tools they need for Grid-enabling their applications;
• Developers - will learn new skills in Web and Grid service technologies and security;
• UK e-Research community - will be provided with a lightweight programming library for linking "heritage" applications into a distributed Virtual Research Environment;
• JISC - will be provided with feedback on using Web services as a framework for VRE development and delivery of remotely-hosted services;
• Service providers - will be encouraged to publish their services via Web service interfaces for remote programmatic access.

Project Partners

Partners will be included from research projects active in the following areas.

1. Bioinformatics applications (lead site, Cambridge) - this brings in a substantial community of life-science researchers, spanning a wide variety of fields including the study of cellular function, organism development and mechanisms of disease. 
2. Computational Chemistry, Molecular Dynamics and Materials Physics (Daresbury) - this brings in the Collaborative Computational Project communities via CCP1, CCP5 and CCP9. 
3. Statistical Computing (Lancaster) - this brings in the quantitative and qualitative Social Science and can extend to the Arts and Humanities communities.