Fine-Grained Replication and Scheduling with Freshness and Correctness Guarantees

Lazy replication protocols provide good scalability properties by decoupling transaction execution from the propagation of new values to replica sites while guaranteeing a correct and more efficient transaction processing and replica maintenance. However, they impose several restrictions that are often not valid in practical database settings, e.g., they require that each transaction executes at its initiation site and/or are restricted to full replication schemes. Also, the protocols cannot guarantee that the transactions will always see the freshest available replicas. This paper presents a new lazy replication protocol called PDBREP that is free of these restrictions while ensuring one-copy-serializable executions. The protocol exploits the distinction between read-only and update transactions and works with arbitrary physical data organizations such as partitioning and striping as well as different replica granularities. It does not require that each read-only transaction executes entirely at its initiation site. Hence, each read-only site need not contain a fully replicated database. PDBREP moreover generalizes the notion of freshness to finer data granules than entire databases.

[1]  Kenneth Salem,et al.  Lazy database replication with ordering guarantees , 2004, Proceedings. 20th International Conference on Data Engineering.

[2]  Henry F. Korth,et al.  Replication and consistency: being lazy helps sometimes , 1997, PODS.

[3]  Alan L. Cox,et al.  Distributed Versioning: Consistent Replication for Scaling Back-End Databases of Dynamic Content Web Sites , 2003, Middleware.

[4]  Christos H. Papadimitriou,et al.  The Theory of Database Concurrency Control , 1986 .

[5]  Marc H. Scholl,et al.  Transactional information systems: theory, algorithms, and the practice of concurrency control and recovery , 2001, SGMD.

[6]  H SchollMarc,et al.  Transactional information systems , 2001 .

[7]  Michael Stonebraker,et al.  Concurrency Control and Consistency of Multiple Copies of Data in Distributed Ingres , 1979, IEEE Transactions on Software Engineering.

[8]  Esther Pacitti,et al.  Preventive Multi-master Replication in a Cluster of Autonomous Databases , 2003, Euro-Par.

[9]  Jonathan Goldstein,et al.  Relaxed currency and consistency: how to say "good enough" in SQL , 2004, SIGMOD '04.

[10]  Gustavo Alonso,et al.  Don't Be Lazy, Be Consistent: Postgres-R, A New Way to Implement Database Replication , 2000, VLDB.

[11]  Fuat Akal.  Fine-Grained Lazy Replication with Strict Freshness and Correctness Guarantees , 2004, VLDB 2005.

[12]  Dennis Shasha,et al.  The dangers of replication and a solution , 1996, SIGMOD '96.

[13]  Esther Pacitti,et al.  Replica Consistency in Lazy Master Replicated Databases , 2001, Distributed and Parallel Databases.

[14]  Rajeev Rastogi,et al.  Update propagation protocols for replicated databates , 1999, SIGMOD '99.

[15]  Gottfried Vossen,et al.  Transactional Information Systems: Theory, Algorithms, and the Practice of Concurrency Control and Recovery , 2002 .

[16]  S. S. Ravi,et al.  Deferred updates and data placement in distributed databases , 1996, Proceedings of the Twelfth International Conference on Data Engineering.

[17]  Heiko Schuldt,et al.  FAS - A Freshness-Sensitive Coordination Middleware for a Cluster of OLAP Components , 2002, VLDB.

[18]  Gustavo Alonso,et al.  Ganymed: Scalable Replication for Transactional Web Applications , 2004, Middleware.

[19]  Gustavo Alonso,et al.  Improving the scalability of fault-tolerant database clusters , 2002, Proceedings 22nd International Conference on Distributed Computing Systems.

[20]  Gustavo Alonso,et al.  Are quorums an alternative for data replication? , 2003, TODS.

[21]  Jonathan Goldstein,et al.  Support for relaxed currency and consistency constraints in MTCache , 2004, ACM SIGMOD Conference.