This paper introduces the concept of a Content-Addressable Network (CAN), a distributed infrastructure that provides hash table-like functionality on Internet-scale. CANs are designed to be scalable, fault-tolerant, and self-organizing, offering low latency and robustness through simulation. The authors argue that hash table functionality is valuable for large-scale distributed systems, particularly in peer-to-peer file sharing systems like Napster and Gnutella, which currently lack scalability and fault tolerance. The paper details the design of CANs, including their routing mechanisms, construction, and maintenance, emphasizing the use of a virtual coordinate space and multi-dimensional partitioning. Additional design improvements, such as multi-reality coordination, RTT-weighted routing, zone overloading, and load balancing, are discussed to enhance performance and robustness. The authors also present simulation results demonstrating the effectiveness of these improvements, showing that CANs can achieve latency within a factor of two of the underlying network latency, even in systems with over 260,000 nodes. The paper concludes by discussing related work and future directions, highlighting the potential of CANs in various applications beyond peer-to-peer systems.This paper introduces the concept of a Content-Addressable Network (CAN), a distributed infrastructure that provides hash table-like functionality on Internet-scale. CANs are designed to be scalable, fault-tolerant, and self-organizing, offering low latency and robustness through simulation. The authors argue that hash table functionality is valuable for large-scale distributed systems, particularly in peer-to-peer file sharing systems like Napster and Gnutella, which currently lack scalability and fault tolerance. The paper details the design of CANs, including their routing mechanisms, construction, and maintenance, emphasizing the use of a virtual coordinate space and multi-dimensional partitioning. Additional design improvements, such as multi-reality coordination, RTT-weighted routing, zone overloading, and load balancing, are discussed to enhance performance and robustness. The authors also present simulation results demonstrating the effectiveness of these improvements, showing that CANs can achieve latency within a factor of two of the underlying network latency, even in systems with over 260,000 nodes. The paper concludes by discussing related work and future directions, highlighting the potential of CANs in various applications beyond peer-to-peer systems.