3dls

Abstract

Finding data items is one of the most basic services of any distributed system. It is particular challenging in ad-hoc networks, due to their inherent decentralized nature and lack of infrastructure. A data location service (DLS) provides this capability. This paper presents 3DLS, a novel densitydriven data location service. 3DLS is based on performing biased walks over a density based virtual topography. 3DLS also includes an autonomic dynamic configuration mechanism for adapting the lengths of the walks, in order to ensure good performance in varying circumstances and loads. This is without any explicit knowledge of the network characteristics, such as size, mobility speed, etc. Moreover, 3DLS does not rely on geographical knowledge, its decisions are based only on local information, it does not invoke multi-hop routing, and it avoids flooding the network. The paper includes a detailed performance study of 3DLS, carried by simulations, which compares 3DLS to other known approaches. The simulations results validate the viability of 3DLS.Data Location Services (DLSs) provide this capability, and are the focus of this paper. A DLS typically has two methods, one for publishing data items and the other for looking up the data. The number of advertisement and lookup messages generated by a DLS implementation impacts the efficiency of the implementation.By examining existing DLS implementations [1,13,15,16,27], including utilizing Distributed Hash Tables (DHTs) over MANETs [3,7,21,29], it is evident that they exhibit at least one of the following shortcomings:• The occasional need to flood the network with data advertisement/lookup or other types of messages. Flooding the network is very resource consuming.• The use of a multi-hop routing protocol, which is very expensive, since route discovery and route maintenance are costly in MANETs [20].• The reliance on geographical knowledge, which demands additional hardware (e.g., GPS).