Abstract
The efficient execution of irregular data parallel applications, on dynamically shared computing clusters, requires novel approaches to manage the runtime load distribution. Stich environments have an unpredictable dynamic behaviour both due to the application requirements and to the available system's resources. This uncertainty was the main motivation to propose and evaluate an application level scheduler, where decisions are efficiently taken with improved accurate predictions on the environment's current and near future state, based on available incomplete and aged measured data. Bayesian decision networks are used as the scheduler's decision making mechanism: its effectiveness to manage the load distribution of a parallel ray tracer is assessed and compared with alternative strategies. The evaluation results, with complex scenes on a 7 shared nodes cluster with dynamically variable workloads, show considerable performance improvements over blind strategies, and stress the benefits over a sensor based deterministic approach of identical complexity.

Abstract
In this paper we describe a transaction management system designed to face the inherent characteristics of mobile environments. Mobile clients cache subsets of the database state and allow disconnected users to perform transactions independently. Transactions are specified as mobile transactional programs that are propagated and executed in the server, thus allowing the validation of transactions based on application-specific semantics. In the proposed model (as in others previously presented in literature) the final result of a transaction is only determined when the transaction is processed in the central server. Users may be notified of the results of their transactions using system support (even when they are no longer using the same application or even the same computer). Additionally, the system implements a reservation mechanism in order to guarantee the results of transactions performed in disconnected computers. © Springer-Verlag Berlin Heidelberg 2000.

Abstract
Reliable multicast protocols can strongly simplify the design of distributed applications. However it is hard to sustain a high multicast throughput when groups are large and heterogeneous. In an attempt to overcome this limitation, previous work has focused on weakening reliability properties. In this paper we introduce a novel reliability model that exploits semantic knowledge to decide in which specific conditions messages can be purged without compromising application correctness. This model is based on the concept of message obsolescence: A message becomes obsolete when its content or purpose is overwritten by a subsequent message. We show that message obsolescence can be expressed in a generic way and can be used to configure the system to achieve higher multicast throughput.

Abstract
This paper is a formal study of how to implement interaction nets, filling an important gap in the work on this graphical rewriting formalism, very promising for the implementation of languages based on the gimel -calculus. We propose the first abstract machine for interaction net reduction, based on a decomposition of interaction rules into more atomic steps, which tackles all the implementation details hidden in the graphical presentation. As a natural extension of this, we then give a concurrent shared-memory abstract machine, and show how to implement it, resulting in the first parallel implementation of interaction nets.

Abstract

Abstract
This paper presents a new strict, purely functional implementation of attribute grammars. Incremental evaluation is obtained via standard function memoization. Our new implementation of attribute grammars increases the incremental behaviour of the evaluators by both reducing the memoization overhead and increasing their potential incrementallity. We present also an attribute grammar transformation, which increases the incremental performance of the attribute evaluators after a change that propagates its effects to all parts of the syntax tree. These techniques have been implemented in a purely functional attribute grammar system and the first experimental results are presented.