Dynamic Load Sharing in Peer-to-Peer Systems: When Some Peers Are More Equal than Others

Abstract

Abstract-In the past few years, several DHT-based abstractions for peer-to-peer systems have been proposed. The main characteristic is to associate nodes (peers) with keys (objects) and to construct distributed routing structures to support an efficient location. These approaches address the load problem, and load balancing is achieved by moving the keys. However, the problem is still not properly covered. In this paper we present an analysis of structured peer-to-peer systems taking into consideration Zipf-like requests distribution. Based on our analysis, we propose a novel approach for load balancing relying on object popularity. Our approach is based on routing table reorganization in order to balance the lookup traffic load. We have implemented this approach in a Pastry-like system. The obtained results demonstrate a better balance of load, which can lead to improved scalability and performance.

Energy-efficient peer-to-peer networking for constrained-capacity mobile environments

“…This type of load imbalance is independent of whether the used P2P system is structured or unstructured. The actual request distribution in P2P applications roughly follows Zipf-like distribution, in which relatively few popular objects account for most of the requests (Hefeeda & Saleh 2008, Serbu et al 2007). Consequently, popular data objects inflict higher routing load for the nodes near the object's location (Zhou & Koyanagi 2013).…”

“…If a node is removed, its address interval is taken over by another node. In the worst case, consistent hashing may result in O (log N) imbalance of object distribution between the peer nodes (Rao et al 2003, Godfrey & Stoica 2005, Serbu et al 2007. Since a node is responsible for all requests targeted at identifiers belonging to its address space, the average number of incoming requests messages depends (along with node popularity) on the address space size (Rao et al 2003).…”

“…The key characteristic of the mechanism is that the IDs of the other nodes remain unchanged, reducing the overhead in the overlay. However, in the worst case, consistent hashing may result in O (log N) imbalance of object distribution between the peer nodes (Rao et al 2003, Godfrey & Stoica 2005, Serbu et al 2007.…”

Managing collaborative feedback information for distributed retrieval

“…Routes from all nodes to some key in the network collide in the last hops, as illustrated below. The path convergence property is particularly useful for the design of load balancing and fault tolerance mechanisms [16,20]. A "standard" DHT provides a raw put/get interface to the application [6].…”

“…When some peer P (Q) gets overloaded by requests to a popular query Q, it replicates its responsibility for managing information and answering requests related to Q. A wide range of techniques has been proposed for balancing load in structured overlays (e.g., [12,16,18,20]). All these proposals however target scenarios where the number of accesses is much greater than the number of updates to the data.…”