Strategies for distributed search

The problem of locating information among cooperating agents within a distributed computing environment is considered. Such collections of agents can be collectively thought of as a single distributed application providing a single set of services; the X.500 distributed directory is an example of such an application. A central problem in such applications is the cooperative search for certain information which might be located at one of the agents. One approach to this problem relies on flooding messages throughout the distributed environment, such as that which might be used in routing algorithms to locate entries. Although the use of such algorithms, in principle, takes advantage of reaching many different agents via parallel data communication paths, the additional number of messages may be a distinct disadvantage in “higher layer” applications. This paper presents a variant of such approaches which rely on a depth-first method of distributing messages. Results of simulations illustrating the savings in messages of this approach to one relying on traditional flooding are presented for different network configurations.