Olivier Marin

DimaX: a fault-tolerant multi-agent platform

Fault tolerance is an important property of large-scale multi-agent systems as the failure rate grows with both the number of the hosts and deployed agents, and the duration of computation. Several approaches have been introduced to deal with some aspects of the fault-tolerance problem. However, most existing solutions are ad hoc. Thus, no existing multi-agent architecture or platform provides a fault-tolerance service that can be used to facilitate the design and implementation of reliable multi-agent systems. So, we have developed a fault-tolerant multi-agent platform (named DimaX). DimaX deals with fail-stop failures like bugs and/or breakdown machines. It brings fault-tolerance for multi-agent applications by using replication techniques. It is based on a replication framework (named DARX).

Towards reliable multi-agent systems: An adaptive replication mechanism

Distributed cooperative applications are now increasingly being designed as a set of autonomous entities, named agents, which interact and coordinate (thus named a multi-agent system). Such applications are often very dynamic: new agents can join or leave, they can change roles, strategies, etc. This high dynamicity creates new challenges to the traditional approaches of fault-tolerance. In this paper, we will focus on crash failures, with usual preventive approaches by replication. But, as criticality of agents may evolve during the course of computation and problem solving, static design is not appropriate. Thus we need to dynamically and automatically identify the most critical agents and to adapt their replication strategies (e.g., active or passive, number of replicas), in order to maximize their reliability and their availability. In this paper, we describe a prototype architecture, supporting adaptive replication. We also discuss and compare various control strategies for replication, one using agent roles, and another using inter-agent dependences as types of information to infer and estimate criticality of agents. Experiments and measurements are also reported.

A fault-tolerant multi-agent framework

To make large-scale multi-agent systems reliable, we propose an adaptive application of replication strategies. Critical agents are replicated to avoid failures. As criticality of agents may evolve during the course of computation and problem solving, we need to dynamically and automatically adapt the number of replicas of agents, in order to maximize their reliability and availability based on available resources. We are studying an approach and mechanisms for evaluating the criticality of a given agent and for deciding what strategy to apply (e.g., active replication, passive) and how to parameterize it (e.g., number of replicas).

Matchmaking in multi-player on-line games: studying user traces to improve the user experience

Designing and implementing a quality matchmaking service for Multiplayer Online Games requires an extensive knowledge of the habits, behaviors and expectations of the players. Gathering and analyzing traces of real games offers insight on these matters, but game server providers are very protective of such data in order to deter possible reuse by the competition and to prevent cheating. We circumvented this issue by gathering public data from a League of Legends server (information over more than 28 million game sessions). In this paper, we present our database which is freely available online, and we detail the analysis and conclusions we draw from this data regarding the expected requirements for the matchmaking service.

Experience and prospects for various control strategies for self-replicating multi-agent systems

Distributed cooperative applications (e.g.,e-commerce) are now increasingly being designed as a set of autonomous entities, named agents, which interact and coordinate(thus named a multi-agent system). Such applications are often very dynamic: new agents can join or leave, they can change roles, strategies, etc. This high dynamicity creates new challenges to the traditional approaches of fault-tolerance. As relative importance of agents may evolve during the course of computation and problem solving,we need to dynamically and automatically identify the most critical agents and to adapt their replication strategies (e.g., active or passive, number of replicas), in order to maximize their reliability and their availability. One important issue is then: what kind of information could be used to estimate which agents are most critical agents? In this paper, we will first introduce our prototype architecture for adaptive replication. Then, we will discuss various kinds of information and strategies to estimate criticality of agents: static dependences, dynamic dependences, roles, norms, and plans. Some preliminary measurements and future directions will also be presented.

DRing: A Layered Scheme for Range Queries over DHTs

Traditional DHT structures provide very poor support for range queries, since uniform hashing destroys data locality. Several schemes have been proposed to overcome this issue, but they fail to combine load balancing, low message overhead, and low latency in search operations. In this article we present DRing, an efficient layered solution that directly supports range queries over a ring-like DHT structure. We improve load balancing by using only the nodes that store data, and by updating neighbour information through an optimistic approach. DRing produces low overhead and low latency in environments where queries significantly outnumber data insertion operations. We analyze DRing through simulation and show that our solution does not rely on data distribution.

RepFD - Using Reputation Systems to Detect Failures in Large Dynamic Networks

Failure detection is a crucial service for dependable distributed systems. Traditional failure detector implementations usually target homogeneous and static configurations, as their performance relies heavily on the connectivity of each network node. In this paper we propose a new approach towards the implementation of failure detectors for large and dynamic networks: we study reputation systems as a means to detect failures. The reputation mechanism allows efficient node cooperation via the sharing of views about other nodes. Our experimental results show that a simple prototype of a reputation-based detection service performs better than other known adaptive failure detectors, with improved flexibility. It can thus be used in a dynamic environment with a large and variable number of nodes.

Towards distributed geolocation for large scale disaster management

In the aftermath of major disasters such as earthquakes, locating individuals is crucial for passing on vital information, for example warnings and safety announcements. However, large scale disasters cause extensive damage to the network infrastructures and a generalized loss of communications in the chaos that ensues. This position paper presents a preliminary study for a geolocation service that relies on inter-device connections: mobile devices exchange positions of previously encountered devices when they come into contact. Every device thus builds a partial map of device locations and can use it to enforce geographic routing protocols that are resilient to large scale disasters.

Vers une plate-forme MapReduce tolérant les fautes byzantines

Les pannes arbitraires sont inherentes aux calculs massivement paralleles tels que ceux vises par le modele MapReduce ; or les implementations courantes du MapReduce ne fournissent pas d’outils permettant de tolerer les fautes byzantines. Il est donc impossible de certifier l’exactitude des resultats obtenus au terme des traitements longs et couteux. Nous presentons dans cet article une architecture permettant de repliquer les tâches dans le modele MapReduce afin de garantir l’integrite des traitements et d’isoler les tâches defaillantes. Dans une premiere etude de performances nous avons evalue certains mecanismes lies a la replication. Une seconde etude, effectuee avec un prototype implementant l’ensemble de l’architecture, a permis de valider certains choix en montrant qu’il est possible de minimiser le surcout de la tolerance aux fautes byzantines.

Probabilistic Byzantine Tolerance Scheduling in Hybrid Cloud Environments

This work explores scheduling challenges in providing probabilistic Byzantine fault tolerance in a hybrid cloud environment, consisting of nodes with varying reliability levels, compute power, and monetary cost. In this context, the probabilistic Byzantine fault tolerance guarantee refers to the confidence level that the result of a given computation is correct despite potential Byzantine failures. We formally define a family of such scheduling problems distinguished by whether they insist on meeting a given latency limit and trying to optimize the monetary budget or vice versa. For the case where the latency bound is a restriction and the budget should be optimized, we present several heuristic protocols and compare between them using extensive simulations.