Handling Network Partitions and Mergers in Structured Overlay Networks

Abstract

Structured overlay networks form a major class of peerto-peer systems, which are touted for their abilities to scale, tolerate failures, and self-manage. Any long-lived Internet-scale distributed system is destined to face network partitions. Although the problem of network partitions and mergers is highly related to fault-tolerance and self-management in large-scale systems, it has hardly been studied in the context of structured peer-to-peer systems. These systems have mainly been studied under churn (frequent joins/failures), which as a side effect solves the problem of network partitions, as it is similar to massive node failures. Yet, the crucial aspect of network mergers has been ignored. In fact, it has been claimed that ring-based structured overlay networks, which constitute the majority of the structured overlays, are intrinsically ill-suited for merging rings. In this paper, we present an algorithm for merging multiple similar ring-based overlays when the underlying network merges. We examine the solution in dynamic conditions, showing how our solution is resilient to churn during the merger, something widely believed to be difficult or impossible. We evaluate the algorithm for various scenarios and show that even when falsely detecting a merger, the algorithm quickly terminates and does not clutter the network with many messages. The algorithm is flexible as the tradeoff between message complexity and time complexity can be adjusted by a parameter. *

Synapse: A Scalable Protocol for Interconnecting Heterogeneous Overlay Networks

“…The first approach was based on hierarchical systems [6,18], with some elected super-peers. In a more general view, merging several co-existing structured overlay networks has been shown to be a very costly operation [5,16]. In the context of mobile ad hoc networks, Ariwheels [1,12] was designed to provide a large variety of services through a multi-layer overlay network.…”

Self-stabilizing and Byzantine-Tolerant Overlay Network

“…Conversely, m 2 / ∈ m 1 .view means that m 1 believes that m 2 is stopped or faulty. 3 The subset m.neighbors of a member's view represents its neighbors.…”

“…The literature studies various efficiency measures and availability issues. Recent papers address the issues of stabilization [1,2,3] and overcoming Byzantine faults [4,5,6]. In the current paper we extend the Fireflies construction [6], making it self-stabilizing.…”

“…In [1] the authors assume the existence of some failure detection subsystem, and when instability is identified, members flood the network to form a new stable membership. In [3] a gossiping style is used, with members occasionally probing potential neighbors to identify whether they are active or not. Our technique resembles this later one, though we reduce the ability of Byzantine members to probe all members all the time.…”

Babelchord: a social tower of DHT-based overlay networks

“…To achieve exhaustiveness, hierarchical approaches require mergers. Authors in [7] focused on merging several similar overlays together. However, as argued in [2], such mechanisms generate a significant communication overhead, not to mention the time required before converging towards a usable single overlay.…”