Abstract:         In this paper, we propose a searching service optimized for highly dynamic mobile ad-hoc networks based on a flooding approach. MANETs unreliability and routing costs prevent the use of central servers or global infra-structured services on top of a priori defined virtual overlay networks. A flooding approach over a virtual overlay network created on-demand performs better. Flooding is supported by a light-weight clustering algorithm. The paper compares the relative efficiency of two clustering approaches using 1.5-hop and 2.5-hop neighborhood information, and of a non-clustered approach. It presents a set of simulation results on the clustering efficiency and on searching efficiency for low and high mobility patterns, showing that the 1.5-hop algorithm is more resilient to load and to node movement than the 2.5-hop algorithm.