In this report, we present the results of a two-week research visit by the first author to the University of Pisa. We argue that problems like load-balancing or fault-tolerance can be addressed by integrating autonomic behavior in existing Grid applications. This can be achieved either by investigating novel approaches, such as multi-objective strategies, or by enhancing existing, specific solutions like cluster-aware random stealing [11] or empirically determined threshold of compute nodes [15]. We investigate both the JavaGAT [12] and Satin [14] systems to be augmented by autonomic behavior, based on autonomic managers from Muskel [7]. The main outcome of this report are items of future research work.

In this paper we present the Grid environment that supports application building basing on a High Level Architecture (HLA) component model. The proposed model is particularly suitable for distributed multiscale simulations. Original HLA partly supports interoperability and composability of simulation models, where connections between modules (federates) in a simulation system (federation) are defined and set by federates themselves. On the contrary, in the proposed component model the particular behavior of component and it’s connections are defined and set by an external module (e.g. builder) on the user request which is more flexible and increases reusability of components. We are also proposing to integrate our HLA component solution with the Grid which will allow users working on distributed simulations to more easily exchange the models already created. The focus of this paper is on design of the HLA component. We show how to insert simulation logic into a component and make possible to steer from outside its connections with other components. Its functionality is shown on the example of multiscale simulation of a stellar system.