Tapestry is an overlay infrastructure designed to enable fault-tolerant and scalable wide-area location and routing. It addresses the challenges of mobile and replicated data and services by providing location-independent routing using only point-to-point links without centralized resources. The routing and directory information in Tapestry is soft state, making it easily repairable and allowing for self-administration and fault tolerance. Tapestry employs randomness to achieve load distribution and routing locality, and it is resilient under heavy loads and network node faults. The paper presents the architecture, algorithms, and experimental results, highlighting Tapestry's advantages in stability, redundancy, and adaptability. Key features include fault handling, surrogate routing, dynamic node insertion, and introspective optimizations to improve performance and adapt to environmental changes.Tapestry is an overlay infrastructure designed to enable fault-tolerant and scalable wide-area location and routing. It addresses the challenges of mobile and replicated data and services by providing location-independent routing using only point-to-point links without centralized resources. The routing and directory information in Tapestry is soft state, making it easily repairable and allowing for self-administration and fault tolerance. Tapestry employs randomness to achieve load distribution and routing locality, and it is resilient under heavy loads and network node faults. The paper presents the architecture, algorithms, and experimental results, highlighting Tapestry's advantages in stability, redundancy, and adaptability. Key features include fault handling, surrogate routing, dynamic node insertion, and introspective optimizations to improve performance and adapt to environmental changes.