Failure-Tolerant Overlay Trees for Large-Scale Dynamic Networks

21st International Conference on Advanced Networking and Applications (AINA '07), 2007

Overlay management protocols have been introduced to guarantee overlay network connectivity in dynamic largescale peer-to-peer systems. Some of these protocols have been specifically designed to avoid the partitioning of the overlay in large clusters (network breakage) despite massive node failures and the continuous arrivals/departures of nodes (churn). In this paper we identify a second effect connected to churn, namely network erosion. We show how erosion affects overlay network connectivity and point out that even a strongly connected overlay network, when exposed to continuous churn, can be disgregated. More specifically the consequences of erosion are shown, through an experimental study, in the context of overlay management protocols based on the view-exchange technique. We finally propose a connection recovery mechanism to be endowed at each node which is able to collaboratively detect node isolation and the presence of small clusters. This mechanism is shown to be effective in reducing the erosion of an overlay network exposed to continuous churn and to quickly recover its connectivity during stability periods.

24th IEEE Symposium on Reliable Distributed Systems (SRDS'05)

We present an algorithm by which nodes arranged in a tree, with each node initially knowing only its parent and children, can construct a fault-tolerant communication structure (an expander graph) among themselves in a distributed and scalable way. The tree overlayed with this logical expander is a useful structure for distributed applications that require the intrinsic "treeness" from the topology but cannot afford any obstruction in communication due to failures. At the core of our construction is a novel distributed mechanism that samples nodes uniformly at random from the tree. In the event of node joins, node departures or node failures, the expander maintains its own fault tolerance and permits the reformation of the tree. We present simulation results to quantify the convergence of our algorithm to a fault tolerant network having both good vertex connectivity and expansion properties.

IEEE Transactions on Parallel and Distributed Systems, 2000

Large-scale distributed applications are subject to frequent disruptions due to resource contention and failure. Such disruptions are inherently unpredictable and, therefore, robustness is a desirable property for the distributed operating environment. In this work, we describe and evaluate a robust topology for applications that operate on a spanning tree overlay network. Unlike previous work that is adaptive or reactive in nature, we take a proactive approach to robustness. The topology itself is able to simultaneously withstand disturbances and exhibit good performance. We present both centralized and distributed algorithms to construct the topology, and then demonstrate its effectiveness through analysis and simulation of two classes of distributed applications: Data collection in sensor networks and data dissemination in divisible load scheduling. The results show that our robust spanning trees achieve a desirable trade-off for two opposing metrics where traditional forms of spanning trees do not. In particular, the trees generated by our algorithms exhibit both resilience to data loss and low power consumption for sensor networks. When used as the overlay network for divisible load scheduling, they display both robustness to link congestion and low values for the makespan of the schedule.

Proceedings of the 38th Annual Hawaii International Conference on System Sciences, 2005

Structured peer-to-peer systems have emerged as infrastructures for resource sharing in large-scale, distributed, and dynamic environments. One challenge in these systems is to efficiently maintain routing information in the presence of nodes joining, leaving, and failing. Many systems use costly periodic stabilization protocols to ensure that the routing information is up-to-date.

2010 Sixth International Conference on Networking and Services, 2010

Structured overlay networks are widely used as platforms for distributed systems. However, in a high churn situation, it is highly costly to maintain the structure and service availability of an overlay network because there is a limitation in churn handling. Several cooperation mechanisms among overlay networks are proposed to improve service availability of overlay networks in a high-churn situation. Several cooperation mechanisms are proposed among overlay networks. The cooperation mechanisms can group structured overlay networks to simplify routing functions. However, the existing mechanisms do not consider managing churn information among overlay networks. In this paper, we propose a new framework that can share states of the nodes and spread churn information to neighbor nodes among different overlay networks. Using our framework, we are able to maintain each overlay network effectively and improve the reachability of messages under churn. We evaluate our framework with Chord, Kademlia, and Pastry. Our framework improves the detection time of churn by about 30% and decreases the stabilization messages by about 30-40%. We discuss combinatorial effects of the sharing state of the nodes among different overlay networks.

We address the problem of building and maintaining distributed spanning trees in highly dynamic networks, in which topological events can occur at any time and any rate, and no stable periods can be assumed. In these harsh environments, we strive to preserve some properties such as cycle-freeness or the existence of a root in each tree, in order to make it possible to keep using the trees uninterruptedly (to a possible extent). Our algorithm operates at a coarse-grain level, using atomic pairwise interactions in a way akin to recent population protocol models. The algorithm relies on a perpetual alternation of \emph{topology-induced splittings} and \emph{computation-induced mergings} of a forest of spanning trees. Each tree in the forest hosts exactly one token (also called root) that performs a random walk {\em inside} the tree, switching parent-child relationships as it crosses edges. When two tokens are located on both sides of a same edge, their trees are merged upon this edge a...

IEICE Transactions on Information and Systems, 2011

Distributed systems desire to construct a random overlay graph for robustness, efficient information dissemination and load balancing. A random walk-based overlay construction is a promising alternative to generate an ideal random scale free overlay in distributed systems. However, a simple random walk-based overlay construction can be affected by node churn. Especially, the number of edges increases and the degree distribution is skewed. This inappropriate distortion can be exploited by malicious nodes. In this paper, we propose a modified random walk-based overlay construction supported by a logistic/trial based decision function to compensate the impact of node churn. Through event-driven simulations, we show that the decision function helps an overlay maintain the proper degree distribution, low diameter and low clustering coefficient with shorter random walks.

IICS, 2010

The lack of infrastructure and dynamic nature of mobile ad hoc networks demand newn etworking strategies to be implemented in order to provide efficient end-to-end communication. Some researches proposed to organize the network into groups called clusters and use different routing protocols for inter and intra cluster to propagate an information. But with these solutions, the network needs first to be organized into clusters and next, we need to construct each routing table. Other researchers proposed to build aspanning tree on the network to forward informations on atree but manysolutions need to knowthe global network topology.I nthis paper,wepropose aself-stabilizing algorithm both to construct cluster and simultaneously build aspanning tree on the network. Without anyg lobal knowledge, we use only one type of periodically exchanged messages of size Log(5n +3)bits, and we construct clusters and the spanning tree on the network with aconvergence time of at most D +6 rounds.

The Computer Journal

Highly dynamic networks are characterized by frequent changes in the availability of communication links. These networks are often partitioned into several components, which split and merge unpredictably. We present a distributed algorithm that maintains a forest of (as few as possible) spanning trees in such a network, with no restriction on the rate of change. Our algorithm is inspired by high-level graph transformations, which we adapt here in a (synchronous) message passing model for dynamic networks. The resulting algorithm has the following properties. First, every decision is purely local-in each round, a node only considers its role and that of its neighbors in the tree, with no further information propagation (in particular, no wave mechanisms). Second, whatever the rate and scale of the changes, the algorithm guarantees that, by the end of every round, the network is covered by a forest of spanning trees in which 1) no cycle occur, 2) every node belongs to exactly one tree, and 3) every tree contains exactly one root. We primarily focus on the correctness of this algorithm, which is established rigorously. While performance is not the main focus, we suggest new complexity metrics for such problems, and report on preliminary experimentation results validating our algorithm in a practical scenario.

Future Generation Computer Systems, 2017

Epidemic protocols are a bio-inspired communication and computation paradigm for large-scale network system based on randomised communication. These protocols rely on a membership service to build decentralised and random overlay topologies. In large-scale, dynamic network environments, node churn and failures may have a detrimental effect on the structure of the overlay topologies with negative impact on the efficiency and the accuracy of applications. Most importantly, there exists the risk of a permanent loss of global connectivity that would prevent the correct convergence of applications. This work investigates to what extent a dynamic network environment may negatively affect the performance of Epidemic membership protocols. A novel Enhanced Expander Membership Protocol (EMP+) based on the expansion properties of graphs is presented. The proposed protocol is evaluated against other membership protocols and the comparative analysis shows that EMP+ can support faster application convergence and is the first membership protocol to provide robustness against global network connectivity problems.