In this paper, a new search algorithm in heterogeneous hierarchical unstructured p2p networks is presented which takes into account heterogeneity of peers such as upload bandwidth, life time in the network and process capacity in its peer discovery policy as factors that influence download performance and user satisfaction. One advantage of this algorithm is that it can be performed in a completely unstructured network without imposing any change on network topology and file replication. However, it obviously decreases file download time as it is designed to discover broad-bandwidth nodes with a higher probability. This property results in a fair distribution of load over heterogeneous p2p network nodes since more capable peers are more likely to provide service. This algorithm is inspired by bacterial foraging strategy in which a bacterium climbs up the nutrient gradient.