Recent proposals in multicast overlay networks have demonstrated the importance of exploiting underlying network topology data to construct efficient overlays. While they avoid virtual coordinates embedding and fixed landmarks measurements, these topology-aware proposals often rely on incremental and periodic refinements to improve each node's position in the delivery tree. However, these approaches are neither scalable, as they induce high communication cost due to refinement overhead, nor efficient because long convergence time is necessary to obtain a stabilized structure. In this paper, we propose a novel highly scalable locating algorithm to initially direct newcomers to the closest set of existing nodes. Each newcomer sends request to a few nodes to build its neighborhood information. On the basis of the locating process, we build a two-level topology-aware scheme, namely LCC. We evaluated the LCC scheme using real WAN deployment and by extensive simulation. WAN experimentation was carried over the PlanetLab wide area testbed using more than 200 machines, and simulations were performed using the BRITE topology generator with over 10000 nodes. We compare the scalability and efficiency of LCC with that of initially-randomly connected overlays. Results demonstrate promising performance of LCC, and show that locating-based overlays achieve 70\% less link adjustments than initially randomly-connected structures, with three times faster convergence. Moreover, while being accurate, the locating process entails modest resources and incurs low overhead during new nodes arrivals.