A Mobile Computing Architecture for Numerical Simulation

TL;DR

This paper proposes a new mobile computing architecture utilizing mobile agents and JavaSpace to enable adaptive numerical simulations that respond dynamically to changing resource availability.

Contribution

It introduces an innovative architecture for numerical simulation that enhances adaptability through mobile agents and JavaSpace, addressing resource variability.

Findings

Successful application to case studies demonstrating adaptability

Initial results show improved resource utilization

Architecture supports dynamic reconfiguration of simulations

Abstract

The domain of numerical simulation is a place where the parallelization of numerical code is common. The definition of a numerical context means the configuration of resources such as memory, processor load and communication graph, with an evolving feature: the resources availability. A feature is often missing: the adaptability. It is not predictable and the adaptable aspect is essential. Without calling into question these implementations of these codes, we create an adaptive use of these implementations. Because the execution has to be driven by the availability of main resources, the components of a numeric computation have to react when their context changes. This paper offers a new architecture, a mobile computing architecture, based on mobile agents and JavaSpace. At the end of this paper, we apply our architecture to several case studies and obtain our first results.