Cyclic Distributed Garbage Collection

With the continued growth of distributed systems as a means to provide shared data, designers are turning their attention to garbage collection, prompted by the complexity of memory management and the desire for transparent object management. Garbage collection in very large address spaces is a difficult and unsolved problem, due to problems of efficiency, fault-tolerance, scalability and completeness. The collection of distributed garbage cycles is especially problematic. This thesis presents a new algorithm for distributed garbage collection and describes its implementation in the Network Objects system. The algorithm is based on a reference listing scheme, which is augmented by partial tracing in order to collect distributed garbage cycles. Our collector is designed to be flexible, allowing efficiency, promptness and fault-tolerance to be traded against completeness, albeit it can be also complete. Processes may be dynamically organised into groups, according to appropriate heuristics, in order to reclaim distributed garbage cycles. Multiple concurrent distributed garbage collections that span groups are supported: when two collections meet they may either merge, overlap or retreat. This choice may be done at the level of different partial tracings, of processes or of individual objects. The algorithm places no overhead on local collectors and does not disrupt the collection of acyclic distributed garbage. Partial tracing of the distributed graph involves only objects thought to be part of a garbage cycle: no collaboration with other processes is required.