Abstract

In this paper we consider the problem of searching for a node or an object (i.e., piece of data, file, etc.) in a large network. Applications of this problem include searching for a destination node in a mobile ad hoc network, querying for a piece of desired data in a wireless sensor network, and searching for a shared file in an unstructured peer-to-peer network. We limit our attention in this study to the class of controlled flooding search strategies where query/search packets are broadcast and propagated in the network until a preset TTL (time-to-live) value carried in the packet expires. Every unsuccessful search attempt results in an increased TTL value (i.e., larger search area) and the same process is repeated. The primary goal of this study is to derive search strategies (i.e., sequences of TTL values) that will minimize the cost of such searches associated with packet transmissions. The main results of this paper are as follows. When the probability distribution of the location of the object is known a priori, we present a dynamic programming formulation with which optimal search strategies can be derived that minimize the expected search cost. We also derive the necessary and sufficient conditions for two very commonly used search strategies to be optimal. When the probability distribution of the location of the object is not known a priori and the object is to minimize the worst-case search cost, we show that the best strategies are randomized strategies, i.e., successive TTL values are chosen from certain probability distributions rather than deterministic values. We show that given any deterministic TTL sequence, there exists a randomized version that has a lower worst-case expected search cost. We also derive an asymptotically (as the network size increases) optimal strategy within a class of randomized strategies.

Citations