Clue Tables: A Distributed, Dynamic-Binding Naming Mechanism

This paper presents a distributed, dynamic naming mechanism called clue tables for building highly scalable, highly available distributed file systems. The clue tables naming mechanism is distinctive in three aspects. First, it is designed to cope well with the hierarchical structure of the modern large-scale computer networks. Second, it implicitly carries out load balancing among servers to improve system scalability. Third, it supports file replication and dynamically designates a primary copy to resolve possible data inconsistency. This paper also reports a performance evaluation of the clue tables mechanism when compared with NFS, a popular distributed file system.