Snapshot isolation and integrity constraints in replicated databases

Abstract

Database replication is widely used for fault tolerance and performance. However, it requires replica control to keep data copies consistent despite updates. The traditional correctness criterion for the concurrent execution of transactions in a replicated database is 1-copy-serializability. It is based on serializability, the strongest isolation level in a nonreplicated system. In recent years, however, Snapshot Isolation (SI), a slightly weaker isolation level, has become popular in commercial database systems. There exist already several replica control protocols that provide SI in a replicated system. However, most of the correctness reasoning for these protocols has been rather informal. Additionally, most of the work so far ignores the issue of integrity constraints. In this article, we provide a formal definition of 1-copy-SI using and extending a well-established definition of SI in a nonreplicated system. Our definition considers integrity constraints in a way that conforms to the way integrity constraints are handled in commercial systems. We discuss a set of necessary and sufficient conditions for a replicated history to be producible under 1-copy-SI. This makes our formalism a convenient tool to prove the correctness of replica control algorithms.

A formal characterization of SI-based ROWA replication protocols

“…This issue was solved in [1], where serialization graphs were extended to all currently specified isolation levels and a set of validation rules were given for all of them. Inspired in such work, Lin et al [26] have provided a graph-inspired specification of the correctness conditions of SI replication protocols, including also support for integrity constraints. So, that work perfectly complements ours, using a widely accepted approach.…”

Scalable Transactions in the Cloud: Partitioning Revisited

“…Management of distributed transactions [17] has been thoroughly studied in the literature ranging from distributed databases to the particular case of replicated databases [18,19]. The latter one aims to hide all aspects of data replication from the user.…”

“…It ensures that the interleaved execution of all transactions in the replicated system is equivalent to a serial execution of those transactions on a single database. With the advent of Snapshot Isolation (SI) databases [20] a similar correctness criterion was formalized as 1CSI [19]. It is clear that some kind of coordination is needed to the decide the outcome of a transaction; this is done by a replication protocol.…”

On the Consistency, Characterization, Adaptability and Integrity of Database Replication Systems

“…These DBMSs are actually providing an enhanced integrity management in their systems, equivalent to the one provided in a serializable level. The resulting isolation level cannot be tagged as a pure SI one (as it was first defined [9]) but as an extended SI level able to support integrity constraints in a serializable way [80]. As a result of this, pure SI certificationbased protocols [40,69,79,93] should not be criticized for omitting the extensions shown above.…”

“…As originally described by Lin et al [79], SRCA was a simple replica control algorithm for SI and its pseudocode did not include any special treatment for possible errors detected during the commit operation. Nevertheless, Lin et al [80] assert that the middleware is able to track the failure of a commit operation accordingly (unfortunately, they do not provide details about what actions are made by this tracking). From their description of the protocol and following our classification of integrity-related problems, we can say that SRCA does not suffer from infinite reattempting nor it does a premature client notification, thanks to such failure tracking.…”