A Systematic Review of Fault Tolerance in Mobile Agents

Abstract

Abstract:Mobile agents have engrossed substantial attention in recent years, especially in fault tolerance researches and several approaches have emerged. Fault tolerance design tends to put a stop to incomplete or complete loss of the agent in the face of failures. Despite these developments, reliability issues still remain a critical challenge. Moreover, there is no comprehensive detail bringing together, summaries of the existing efforts of researches in order to focus attention where it is needed most. Therefore, our objective in this systematic literature review (SR) is to explore and analyze the existing fault tolerance implementations in order to bring about the state-of-the-art and the challenges in mobile agent's fault tolerance approaches. We used studies from a number of relevant article sources, and our results showed the existence of twenty six articles. Our analysis indicates that the existing approaches are not generic and each focuses on a specific aspect of the problem, usually in one or two specific fault models which impacts on agent's reliability. The implication of the study is to give a clear direction to future researchers in this area for a better reliable and transparent fault tolerance in mobile agents.

Self-Healing Dilemmas in Distributed Systems: Fault Correction vs. Fault Tolerance

“…For each of the three validation fault profiles, the best fits observed in Figure 5b are selected to compare generalized regression to the other predictors. Generalized regression is performed with regularization 12 .…”

“…The latter further distinguishes between timeout and missing callback detection. Replication [12], [10] is a common approach that supports both fault tolerance and fault correction in terms of guarantying the availability of (backup) resources and repair modules for self-healing [13]. Replication can be active vs. passive based on whether replicas are used only when faults occur [14], adaptive based on criteria for replication [15], [16], dynamic by switching on-the-fly replication schemes [14], or homogeneous vs. heterogeneous based on whether replicas are identical copies or equivalent processes [13].…”

“…New replication strategies designed for the Internet of Things and cyber-physical systems are subject of recent work [21], [22]. Self-healing methods [23] can be preventive (proactive) [12] or reactive (resilient) [11]. The former methods require prediction based on probabilistic modeling and monitoring [24].…”

“…Empirical assessments conclude that multiple input traces covering a vast spectrum of failure characteristics are required to predict the performance of a self-healing system. Therefore, generalized models that predict the impact of faults and their correction on largescale decentralized systems are missing so far [12]. Predicting without knowledge about the computational/application scenario, executed application algorithms and application data is challenging [12].…”

“…Therefore, generalized models that predict the impact of faults and their correction on largescale decentralized systems are missing so far [12]. Predicting without knowledge about the computational/application scenario, executed application algorithms and application data is challenging [12]. Such models have the potential to fundamentally influence the understanding of how to design and deploy more cost-effective decentralized self-healing systems.…”