The paper presents a binning scheme that allows nodes to partition themselves into bins based on their relative network latency to a set of well-known landmarks. This scheme is designed to be simple, scalable, and distributed, requiring minimal infrastructure support. The authors apply this binning scheme to two applications: overlay network construction and server selection. For overlay network construction, the binning scheme is used to improve the structure of the overlay network by ensuring that nodes within the same bin are relatively close to each other, leading to shorter paths and improved routing performance. For server selection, the binning scheme helps clients choose servers that are close to them in terms of latency, reducing the average latency of requests. The performance of these applications is evaluated using simulations and Internet measurement traces, showing that even coarse-grained topological information can significantly improve performance. The results also suggest that the Internet's topology is well-modeled by a power-law random graph.The paper presents a binning scheme that allows nodes to partition themselves into bins based on their relative network latency to a set of well-known landmarks. This scheme is designed to be simple, scalable, and distributed, requiring minimal infrastructure support. The authors apply this binning scheme to two applications: overlay network construction and server selection. For overlay network construction, the binning scheme is used to improve the structure of the overlay network by ensuring that nodes within the same bin are relatively close to each other, leading to shorter paths and improved routing performance. For server selection, the binning scheme helps clients choose servers that are close to them in terms of latency, reducing the average latency of requests. The performance of these applications is evaluated using simulations and Internet measurement traces, showing that even coarse-grained topological information can significantly improve performance. The results also suggest that the Internet's topology is well-modeled by a power-law random graph.