Evaluating fault tolerance approaches in multi-agent systems

Self-Corrective Dynamic Networks via Decentralized Reverse Computations

“…The goal of the experimental evaluation is to show how the self-corrective model performs in DIAS under a lightweight 6 Available at https://github.com/epournaras/PeerSamplingService (last accessed: January 2017 7 Available at http://brutuswiki.ethz.ch (last accessed: January 2017). 8 Available at http://www.ucd.ie/issda/data/ commissionforenergyregulationcer/ (last accessed: February 2017) 9 Available at https://weka.wikispaces.com (last accessed: February 2017) and heavyweight scenario of network adjustments focusing 10 on leaves and rejoins of nodes. In both scenarios, the specific nodes leaving the network are fixed between the different experiments to compare the results.…”

“…A majority of fault-detection, replication and fault-tolerance mechanisms for multi-agent systems are managed in a centralized fashion by dedicated, for this purpose, replication servers, proxies and storage of checkpoints [18], [19], [8]. In contrast, the agents of the proposed self-corrective model rely on the peer sampling service for a fully decentralized detection and orchestration of reverse computations.…”

“…The pervasiveness and increasing computational capacity of smart Internet of Things devices equipped with networking capabilities allow complex distributed computations to be performed over networks, for instance, sensor networks computing the spread of oil spills [1], smart grids measuring power peaks in energy demand [2] or monitoring of automotive traffic [3]. Decentralized computations over dynamic networks are highly challenging to perform accurately and fast under changing input data, nodes temporarily leaving, failing or rejoining the network [4], [5], [6], [7], [8]. However, algorithms for computations in decentralized networks are by design more privacy-preserving, scalable, respect users' autonomy and do not require significant investments in expensive big data computational resources [9], [10], [11], [12].…”

“…This paper introduces a new agent-based self-corrective model for dynamic computational networks, which, in contrast to related work [18], [19], [8], [12], does not rely on replication servers, proxies, storage of checkpoints or big data analytics for fault analysis. Each node in the network creates two software agents, the status and the corrective agent.…”

Adaptive byzantine fault tolerance support for agent oriented systems: The BDARX

“…These systems need to run smoothly without any failure even in the presence of faulty behavior of different agents (Briot et al, 2007). In a large scale MAS, failures ratio grows with the number of agents, hosts and duration required for agents' task execution (Stanković et al, 2017). It is very difficult to recognize faulty agents in advance to avoid their crashes (Ductor et al, 2011), so there is strong need for fault tolerance schemes in such systems.…”