A Simulation Framework to Analyze Knowledge Exchange Strategies in Distributed Self-adaptive Systems

Abstract

Distributed self-adaptive systems are on the verge of becoming an essential part of personal life. They consist of connected subsystems, which work together to serve a higher goal. The highly distributed and self-organizing nature of the resulting system poses the need for runtime management. Here, a particular problem of interest is to determine an optimal approach for knowledge exchange between the constituent systems. In the context of multi-agent systems, a lot of theoretical work investigating this problem has been conducted over the past decades, showing that different approaches are optimal in different situations. Thus, to actually build such systems, the insights from existing theoretical approaches need to be validated against concrete situations. For this purpose, we present a simulation platform to test different knowledge exchange strategies in a test scenario. We used the open source context simulator Siafu as a basis for our simulation. The described platform enables the user to easily specify new types of constituent systems and their communication mechanisms. Moreover, the platform offers several integrated metrics, which are easily extensible. We evaluate the applicability of the platform using three different collaboration scenarios.