Should the grid middleware look to self-managing capabilities?

Abstract

Grid technologies have enabled the clustering of a wide variety of geographically distributed resources and services. While providing this, the grid middleware layers that make up the supporting platform for grid applications have taken an increased level of importance in the overall performance of such distributed applications. In this paper we discuss the necessity of introducing self-managing capabilities to the core functionalities of the grid middleware. Our opinions are based on a simple example that shows how Globus Toolkit 4 (GT4) can be driven to a state of unavailability under certain overloading conditions, and how a simple but effective self-managing policy applied to its resource management mechanisms could overcome such an unfavourable scenario. We present an approach showing the benefits that a resource management introduced in the middleware can provide to the user. This approach is the base of a self-managing layer that is being developed for use under more generic conditions