Exploiting Atomic Broadcast in Replicated Databases.
||Exploiting Atomic Broadcast in Replicated Databases.
||D. Agrawal, G. Alonso, A. El Abbadi, I. Stanoi
||Proceedings of Euro-Par'97Passau, Germany
||Institute for Information Systems, ETH Zürich
In this paper, we propose a series of protocols that
eliminate the mismatch between database transactions and broadcast
communication in the context of replicated databases. Our goal is
twofold. From a database perspective, by using a powerful broadcast
communication primitive, we hope to simplify the management of
replicated databases and perhaps obtain some benefits. In particular,
broadcasts can be used to guarantee the consistency of multiple copies
and at the same time avoid some of the pitfalls of previous proposals
based on point-to-point communication, namely reducing the probability
of deadlocks. In fact, in this paper we show that broadcast
primitives hold the promise of eliminating single object deadlocks,
and either localizing deadlock resolution or even completely
eliminating deadlocks. Our second goal is to provide a realistic
platform for evaluating some of the broadcast primitives. We show that
if broadcast are powerful enough to provide certain ``atomicity''
guarantees, the database application can execute transactions very
efficiently while almost completely eliminating deadlocks. However,
if such atomic properties are weakened due to various system
assumptions, a database application that requires strict database
consistency must perform its own atomic commitment operations in spite
of the properties of broadcast primitives.
!!! Dieses Dokument stammt aus dem ETH Web-Archiv
und wird nicht mehr gepflegt !!!
!!! This document is stored in the ETH Web archive
and is no longer maintained !!!