Maintaining consistency in a replicated software archive

Abstract

SUMMARYArchives of software packages made available on the Internet have become an increasingly common and important way of distributing these resources. To improve local access speeds, it is common for these archives to be mirrored, i.e. replicated at regional sites throughout the world. When these sites are also active participants in the augmentation and maintenance of the archive, it becomes necessary to impose a regime which will ensure that errors and inconsistencies do not arise as a result of conflicting activities at different centres. We describe here procedures which have been developed for the organisation and management of a multi-site software archive in which items of software may be introduced or updated at any of the participating sites. A simple algorithm is outlined to propagate changes made to all sites, protecting against conflicting changes and ensuring consistency of the archive is maintained. Similar methods are applicable to the management of other kinds of distributed system, especially internet-based information services, including World Wide Web sites which allow regional updates.

Adaptive trade‐off between consistency and performance in data replication

“…Data consistency is a prominent problem for replication systems and has been receiving a lot of attention for many years. Early replication systems pursue strong consistency through ad hoc methods or widely‐used protocols. In recent years, a lot of research efforts have turned to understand the associated factors that affect consistency and varying consistency requirements in practical applications.…”