This paper presents a distributed binning scheme to infer network proximity information for use in overlay network construction and server selection. The scheme partitions nodes into bins based on their relative network latency to a set of well-known landmark nodes. Nodes independently determine their bin based on their latency measurements to these landmarks, resulting in a simple, scalable, and distributed approach. The binning strategy is evaluated using simulation and Internet measurement data, showing that even coarse-grained topological information can significantly improve application performance. The binning scheme is applied to two applications: overlay network construction and server selection. For overlay networks, the scheme helps construct topologically congruent overlays, reducing routing latency. For server selection, the scheme enables clients to choose servers that are closer in terms of network latency, improving retrieval performance. The paper evaluates the effectiveness of the binning scheme by comparing it to other binning techniques and server selection methods. Results show that the binning-based approach outperforms random selection and other heuristic methods, particularly in reducing latency stretch. The scheme is scalable, requiring minimal infrastructure support and offering practical benefits for large-scale distributed systems. The study also highlights the importance of scalability and practicality over accuracy in topology inference. While accurate topological information is not always necessary, the binning scheme provides sufficient information to improve application performance. The results suggest that the binning approach is effective for a wide range of Internet topologies, including Transit-Stub and Power-Law Random Graphs. The paper concludes that the proposed binning scheme is a practical and scalable solution for inferring network proximity information, which can be used to improve the performance of distributed applications such as overlay networks and content distribution systems. The scheme is simple to implement, requires minimal infrastructure support, and provides significant performance improvements for large-scale applications.This paper presents a distributed binning scheme to infer network proximity information for use in overlay network construction and server selection. The scheme partitions nodes into bins based on their relative network latency to a set of well-known landmark nodes. Nodes independently determine their bin based on their latency measurements to these landmarks, resulting in a simple, scalable, and distributed approach. The binning strategy is evaluated using simulation and Internet measurement data, showing that even coarse-grained topological information can significantly improve application performance. The binning scheme is applied to two applications: overlay network construction and server selection. For overlay networks, the scheme helps construct topologically congruent overlays, reducing routing latency. For server selection, the scheme enables clients to choose servers that are closer in terms of network latency, improving retrieval performance. The paper evaluates the effectiveness of the binning scheme by comparing it to other binning techniques and server selection methods. Results show that the binning-based approach outperforms random selection and other heuristic methods, particularly in reducing latency stretch. The scheme is scalable, requiring minimal infrastructure support and offering practical benefits for large-scale distributed systems. The study also highlights the importance of scalability and practicality over accuracy in topology inference. While accurate topological information is not always necessary, the binning scheme provides sufficient information to improve application performance. The results suggest that the binning approach is effective for a wide range of Internet topologies, including Transit-Stub and Power-Law Random Graphs. The paper concludes that the proposed binning scheme is a practical and scalable solution for inferring network proximity information, which can be used to improve the performance of distributed applications such as overlay networks and content distribution systems. The scheme is simple to implement, requires minimal infrastructure support, and provides significant performance improvements for large-scale applications.