Proposed Research

 

General Objectives

This project is concerned with fundamental research relating to advanced middleware and advanced networking—and, crucially, an approach that integrates these two key areas. We are particularly interested in supporting classes of advanced Grid application that exhibit the following properties:

  • high levels of heterogeneity in terms of both networking and end-systems;
  • real-time interactive collaboration employing multiple media-types;
  • large scale, complexity and dynamic (re-)configuration;
  • QoS-sensitivity, and adaptability to changes in underlying environmental conditions.

A prime illustrative example of such an application is a world-wide collaborative visualization session involving large numbers of scientists who join and leave the session dynamically and are connected by a variety of access networks and end-systems (including wireless networks/PDAs), and involving multiple media such as visualization data, live sensor output, vector graphics and video [VESC,03]. We contend that such scenarios fundamentally over-stretch the state-of-the-art in existing Grid support. In particular, while OGSA [OGSA,03] is an advance on the initial Globus releases (e.g. in terms of service abstraction), significantly more is needed to address the above needs.

Our analysis is that current support has three major deficiencies in terms of supporting such advanced applications:

  1. Integration with advanced network services. One of the attractions of OGSA is its simple SOAP-based model of interaction. However, advanced applications often require more sophisticated communications services in terms of, for example, QoS management, different modes of interaction (e.g. multicast, streaming, publish-subscribe, peer-to-peer), and application-specific semantics (much of this research is now subsumed as overlay networks [El-Sayed,03]).

  2. Architectural framework. OGSA focuses on interoperability through the use of ubiquitous web protocols and associated abstractions (e.g. WSDL). To enable greater exploitation, this focus on interoperability needs to be complemented with a strong overall architecture that supports the integration of diverse system elements in terms of both breadth (e.g. generic, ‘horizontal’ distributed services such as persistence, visualization, conferencing) and depth (e.g. underlying ‘vertical’ communication services in intimate contact with the network; and with other, end-system based, resources).

  3. Complexity management. As Grid applications become increasingly large, complex, and long-lived, there emerges a strong need for their sophisticated management. It is becoming recognised that the scale and complexity of such systems demand a self-managing (cf. autonomic) approach. Linking back to the previous two points, it is crucial that such self-management is applicable to the architecture of the whole system including communication services. We argue that this implies an open and programmable approach to system construction.

We propose to address these problems through a pervasive component-based approach leading to platform support for next-generation Grid applications that integrates middleware and networking functionality. Component technologies have already been adopted successfully in Grid research to promote structure and re-use at the application level [Furmento,02]. But here we propose the use of component technology not only for applications but throughout the platform architecture in terms of both breadth and depth as outlined above. It is important to stress that we do not propose the use of traditional heavyweight component technologies like Enterprise JavaBeans (EJB). Rather, a major aim of the research is to develop and apply a lightweight component model that imposes minimal overhead, and can be used to build even low-level, system-oriented, functionality. The component model will also be system and language independent, and API-neutral, so that it can be used to construct arbitrary application-level programming environments as required (e.g. OGSA, Web Services, or EJB).

Our central proposal is to apply lightweight component-based technology to construct an extensible family of open and programmable overlay networks, thus providing an approach that is network-centric, offers a strong architecture for the system infrastructure, and facilitates self-management through the inherent openness of component-based structures [Blair,02]. This approach also promises other important benefits: i) a range of interaction styles can be offered depending on the application domain and/or context (e.g. lightweight real-time event channels; QoS-managed data streams; peer-to-peer protocols; data-sharing services); and ii) it facilitates dynamic re-configuration of communications (and other services) as context changes (e.g. to maintain a visualization session when an end user roams to a wireless network).

To validate our approach and provide context for practical experimentation we will call upon selected collaborative visualization-based applications and scenarios. Collaborative visualization is highly appropriate for this purpose because of its inherent properties as outlined above. It is also data and compute intensive which makes it an ideal case study for an infrastructure that aspires to manage both network and end-system resources in an integrated manner.

 

Programme of Research

Aims, Objectives, and Approach

The central aim of the project is to investigate the role of component technology in addressing the challenges in middleware support for Grid applications identified in Section A. More specifically, the project has the following objectives:

  • to develop and evaluate the novel concept of Open and Programmable Overlays for constructing Grid applications;
  • to develop an extensible family of Open Overlays that leverage Lancaster’s reflective component-based approach to systems building, and are tailored for prototypical Grid application scenarios based on collaborative visualization;
  • to develop and evaluate techniques and strategies for self-management in Open Overlays;
  • to integrate Open Overlays into advanced middleware structures that are explicitly extensible and (re)configurable in terms of both breadth and depth as defined in Section A;
  • to validate the Open Overlays concept in the context of selected collaborative visualization scenarios;
  • to disseminate our results to the Grid community with a view to influencing post-OGSA Grid middleware solutions.

The approach to meeting these objectives will be hands-on and practical in nature, and will build on Lancaster’s existing library of components that comprise the OpenORB and NETKIT platforms. These include a Component Framework for a programmable router, Web Services components such as SOAP, and a range of other middleware-based communications services such as publish-subscribe, application-level multicast etc. We also have an initial set of visualization components from the Visual Beans project.

 

References

[Blair,02] Blair, G.S., Coulson, G., Blair, L., Duran-Limon, H., Grace, P., Moreira R., Parlavantzas, N., “Reflection, Self-Awareness and Self-Healing in OpenORB”, Proc. ACM Sigsoft Workshop on Self-Healing Systems (WOSS’02), Nov 02.

[El-Sayed,03] El-Sayed, A., Roca, V., Mathy, L., “A Survey of Proposals for an Alternative Group Communication Service”, IEEE Network, Vol 17, No 1, pp46-51, Jan 03.

[Furmento,02] Furmento, N., Mayer, A., McGough, S., Newhouse, S., Field, T., Darlington, J., “ICENI: Optimisation of Component Applications within a Grid Environment”, Parallel Computing, Vol 28, No 12, pp1753-1772, 02.

[OGSA,03] Tuecke, S., Grid Service Specification. http://www.gridforum.org/ogsi-wg/drafts/draft-ggf-ogsi-gridservice-29_2003-04-05.pdf

[VESC,03] Visualization for e-Science, Report on a workshop held at the National e-Science Centre, Edinburgh, Jan 03 http://umbriel.dcs.gla.ac.uk/NeSC/general/esi/events/130/workshop_report.pdf