Several routing algorithms for mobile ad hoc networks have been proposed in the recent past [Broch et al., The Dynamic Source Routing Protocol for Mobile Ad Hoc Networks, Internet Draft draft-ietf-manet-dsr-03.txt, October 1999; Perkins et al., Ad Hoc On Demand Distance Vector (AODV) Routing, Internet Draft draft-ietf-manet-aodv-04.txt, October 1999; Haas and Pearlman, The Zone Routing Protocol (ZRP) for Ad Hoc Networks, Internet Draft draft-zone-routing-protocol-01.txt, August 1998; IEEE J. Select. Areas Commun. 17 (8) (1999) 1454]. With the exception of a few, these protocols (i) involve all nodes in the route management process, (ii) rely on the use of broadcast relays for route computation, and (iii) are primarily reactive in nature. Related work [Broch et al., Performance Comparison of Multi-Hop Wireless Ad Hoc Network Routing Protocols, Proceedings of IEEE MOBICOM, Dallas, TX, October 1998; Johansson et al., Scenario-based Performance Analysis of Routing Protocols for Mobile Ad-hoc Networks, Proceedings of IEEE MOBICOM, Seattle, August 1999] has shown that the capacity utilization in ad hoc networks decreases significantly when broadcast relays or “broadcast storms” are performed frequently. This effect is compounded when all nodes in the network take part in the route computation. We propose and study an approach based on overlaying a virtual infrastructure (adaptation of the core, proposed in [IEEE J. Select. Areas Commun. 17 (8) (1999) 1454]) on an ad hoc network and operating routing protocols over the infrastructure. The core enables routing protocols to use only a subset of nodes in the network for route management and avoid the use of broadcast relays. We evaluate the performance of dynamic source routing (DSR) [Broch et al., The Dynamic Source Routing Protocol for Mobile Ad Hoc Networks, Internet Draft draft-ietf-manet-dsr-03.txt, October 1999] and AODV [Perkins et al., Ad Hoc On Demand Distance Vector (AODV) Routing, Internet Draft draft-ietf-manet-aodv-04.txt, October 1999], when they are operated over the core and compare their performances against those of their basic versions. Through extensive simulations using ns-2 [Fall and Vardhan, ns notes and documentation, available from http://www-mash.cs.berkeley.edu/ns/, 1999], we show that using a virtual infrastructure significantly improves the performance of both DSR and AODV, in terms of data delivery and routing overhead, under varied network characteristics.