"Dealing with Extreme Heterogeneity and Dynamicity in Next Generattion Middleware"

This research project is jointly carried out by University College London (within the Dept. of Computer Science's Software Systems Engineering Group), and by Lancaster University's Computing Department. This project aims at dealing with extreme heterogeineity and dynamicity in next generation middleware.

With the success of the Grid, there is an accelerating trend towards diversity in terms of both application domains and also, crucially, the underlying networked infrastructures in use. For example, with the emergence of the pervasive Grid, we can see a spectrum ranging from very large cluster computers interconnected with high-speed networks through to embedded devices interconnected by often intermittent and low bandwidth wireless networks.

This infrastructure heterogeneity is especially exhibited at the network level and at the middleware level.

• At the network level - high-performance local and lower-performance wide-area networks; infrastructure-based wireless networks; ad-hoc wireless networks (themselves ranging from relatively static to highly dynamic configurations); and specialised sensor networks.
• At the middleware level - basic point-to-point communication interaction patterns (e.g. SOAP interactions), reliable and unreliable multicast; workflow; media streaming; publish-subscribe based content dissemination; specialised paradigms such as tuple-space/ generative communication; and peer-to-peer based resource location or file sharing.

The position is becoming even more complex with the emergence of virtual or overlay networks (often designed to support the list of advanced middleware-level services listed above). This has resulted in what we refer to as the “Divergent Grid” in which hybrid infrastructures (mixtures of network styles and interaction patterns) are increasingly becoming the norm.

Dealing with such extreme heterogeneity is a fundamental challenge for future Grid Computing and computer science generally. Iit is also important to consider the extreme dynamicity of future Grid environments in terms of the changing underlying networked environment (due primarily to changing network conditions, mobility of participants or devices, and opportunistic introduction of new elements into the Grid), and also the evolving needs of the application base

To deal with such heterogeneity and dynamicity in the grid, the Divergent Grid project will investigate autonomic solutions to the Divergent Grid, building on the contemporary trends in open and reflective middleware architecture [Kon,02]. More specifically, the project seeks to investigate:

• Techniques and tools to support the automatic (self-)configuration of middleware to a potentially highly-heterogeneous underlying networked environment and set of application requirements (specifically by extensions to the concept of model-driven architecture);
• Intrinsic support for self-adaptation of middleware platforms to reflect the current context in a manner consistent with constraints that are imposed by the MDA model;
• Support for the self-configuration and adaptation of key non-functional properties related to security and dependability;
• The evaluation of the above concepts in computational chemistry, a demanding e-Science application domain.