High Throughput Escrow Algorithms for Replicated Databases

The traditional correctness criterion in replicated databases is one-copy serializability. However, this criterion is sometimes restrictive and degrades performance. Recent research has therefore focused on utilising application semantics to increase transaction throughput in certain high-performance applications. One such application involves resource allocation. To improve concurrency in such a system, the transaction escrow (TE) and site escrow (SE) algorithms have been proposed. In this paper, we present a generalized site escrow algorithm (GSE) that provides high site autonomy and throughput. GSE requires only a loose synchronization between sites, and employs the mechanisms of quorum locking and background gossip messages. We perform a comparison between GSE and TE, and outline regions in which GSE performs better. We also propose a family of hybrid algorithms that switch between GSE and TE under appropriate *This work was supported by NSF Grants No. CCR6701671 and No. CCR8901966 Permission to copy without fee all or part of this material ir granted provided that the copies are not made or distributed for direct commercial adoantage, the VLDB copyright notice and the title of the publication and itr date appear, and notice is given that copying is by permission of the Very Large Data Base Endowment. To copy otherwise, or to republish, requirea a fee and/or special permission from the Endowment. Proceedings of the 18th VLDB Conference Vancouver, British Columbia, Canada 1992 circumstances so that the benefits of both algorithms can be utilized. Finally, we present a variant of GSE that does not use locking.