Efficient optimistic concurrency control using loosely synchronized clocks
SIGMOD '95 Proceedings of the 1995 ACM SIGMOD international conference on Management of data
A transactional API for the EAN X.500 directory service
CASCON '92 Proceedings of the 1992 conference of the Centre for Advanced Studies on Collaborative research - Volume 2
Weakly consistent transactions in ROSS
CASCON '93 Proceedings of the 1993 conference of the Centre for Advanced Studies on Collaborative research: distributed computing - Volume 2
| Hi-index | 0.00 |
Optimistic concurrency control techniques allow atomic transactions (or actions for short) to execute without synchronization, relying on commit-time validation to ensure serializability. Previous work in this area has focussed on single-level actions. This thesis extends previous work on optimistic concurrency control to distributed systems with nested actions. The thesis presents two contrasting models for managing nested actions, which we call the fixed action model and the fixed object model. In the fixed action model, an action executes at only a single network node; if an action accesses an object whose storage is provided by another node, it brings a copy of the object to its node. In this model, caching copies of non-local objects can be used to improve performance by reducing the number of non-local object requests. We show that optimistic concurrency control has an advantage over pessimistic concurrency control with respect to this object caching. In the fixed object model, actions can span network nodes: an action at one node can start a nested action at another node. However, an object''s state is never moved from the node providing its storage. While there is a clear reason for using optimistic concurrency control for systems following the fixed action model, there is no clear reason for choosing either optimism or pessimism for systems following the fixed object model; however, the development of this model is an important step in the study of hybrid models.