WIRELESS COMMUNICATIONS RESEARCH TRENDS
Active In SP
Joined: Jun 2010
11-10-2010, 12:20 PM
This article is presented by:
Tong S Lee editor
Peer-to-Peer Systems and Mobile Ad-hoc Networks
As shown dramatically by YouTube, the power of decentralization is irresistible. In most areas, it is only a matter of time for the decentralization model to replace the prevalent client/server model. Although the decentralization model is powerful, not every technology of decentralization is successful. For instance, peer-to-peer (P2P) systems and mobile ad-hoc networks (MANETs), two leading technologies in decentralization, are divergent in their commercialization. In case of P2P systems, welcomed applications succeeded in the market, gained popularity, and attracted active research. Research in turn brought in better applications. This is a virtuous cycle. In the case of MANETs, theoretical research dominated the area for more than a decade, but very little has been transfer into commercial application, if we do not include sensor networks as MANETs. Virtually no application has been widely used in real world except Bluetooth-based MANETs. P2P systems and MANETs share fundamental homogeneity in many aspects. For example, both communicate by multi-hop messaging. Both are characterized by the absence of network infrastructure. Both are against the framework of central controller and instead rely upon self organization. These inherent similarities imply promising probability of successful synergy and transplantation. The term “P2P system” is used throughout this chapter. However, it is not semantically differentiated with “P2P network.”
Bootstrapping P2P Systems over MANETs
In the synergy of P2P systems and MANETs, a trend has been seen in transplanting achievement in P2P systems into MANETs. In this direction, majority of research has focused on issues of stable status, especially routing. Transplantation hence concentrated on layer substitution which match layer model in wired IP networks to layer model of MANETs. [HGRW2006, LLS2004, HPD2003, PDH2004] Very limited research has been done in exploring the bootstrapping. Bootstrapping has been largely circumvented using unrealistic assumptions. This has been a repeated characteristic in the research of P2P systems over wired networks. [RD2001, SMKKB2001, CCNOR2006] Bootstrapping includes two major tasks. The first task is automatic nodes address configuration. If we follow the traditional layer model of MANETs and keep the stiff separation between layers, we need two configurations: the lower in the networking layer and the higher in overlay or application layer. The second task is setting up overlay topology. This chapter focuses on the second task. In computer networks, topology is frequently used to define qualitative geographic relationships, such as “which node is directly connected to which node,” or “which node is neighbor of which node.” Certain type of structured P2P systems imposes particular topologies among nodes to form a specific global structure. For a structured P2P system, overlay topology is crucial to its functionality. It lays foundation for other functions like Topology Construction for Bootstrapping Peer-to-Peer Systems... 3 routing, resource sharing, advertising, looking up, retrieval, and data dissemination. It is one of dominant factors that affect primary performance parameters such as efficiency, robustness, scalability, and feasibility. In fact, topology has broader functionality. For example, Jelasity and Babaoglu [JB2005] have shown that problems such as clustering and sorting can be transformed into topology problems and be solved by specific topology construction.
There are two major approaches for bootstrapping a structured P2P system over wired networks. One is to jumpstart a network from one or a few predefined nodes, in which the common way to expand the network is node joining. In wired networks, many structured P2P systems require manual creation of a “seed” network in bootstrapping. Nodes have to be booted one by one in a slow, linear manner, which costs long time. In addition, the jumpstarted network often needs extra long time for stabilization before normal routing could work. In another approach all nodes cooperate concurrently to construct an overlay topology. This approach is distributed and decentralized. The concurrency makes it much faster than joining approach. Furthermore the P2P system could advance to normal working status immediately after bootstrapping. Remarkable advance has been recently made in topology construction in wired networks. Topology generators can construct topologies such as line, ring, mesh, star, and tree. Generic generator, which could construct any topology if given a mathematical expression, is already available. [JB2005] However, no such construction tool has been reported in MANETs. Many problems remain unsolved in this area. In wired networks, existing protocols for topology construction and maintenance are usually based upon unrealistic assumptions. Most of them assume the existence of a specific initial topology. Some protocols demand that the network remains in an ideal topology all the time as a necessary condition for normal operations. The second problem is the ignorance of network merger and partition. Some systems even require each node keep and monitor global state of the entire network. Another problem is: many schemes for topology construction still employ centralized strategy. Some require central coordinators; some follow a network-wide top-down view in protocol design. Some have deficiencies in fault-tolerance and recovery. For structured P2P systems over MANETs, some approaches can not keep up with the rate of change.
For more information about this article,please follow the link: