Database Replication and Clustering

Tutorial Proposal for ACM SAC 2008

Rui Oliveira
rco@di.uminho.pt Dep. Informtica/CCTC, Universidade do Minho Campus de Gualtar, 4710-057 Braga, Portugal Phone: +351 253 604 470, Fax: +351 253 604 471

Lus Rodrigues
ler@ist.utl.pt IST/INESC-ID Rua Alves Redol 9, 1000-029 Lisboa Portugal Phone: +351 213 100 346, Fax: +351 213 145 843

Jos Pereira
jop@di.uminho.pt Dep. Informtica/CCTC, Universidade do Minho Campus de Gualtar, 4710-057 Braga, Portugal Phone: +351 253 604 470, Fax: +351 253 604 471

Database Replication and Clustering

Abstract Database replication and clustering are increasingly widespread solutions to scalability and high availability challenges in current information systems. The tutorial offers a complete overview of these techniques, from theory to practice. It starts by making an overview of the main challenges involved and the algorithms and engineering techniques to address them. Then, it provides an in depth description of the architecture of a modern open-source database replication system and a hands-on guide to its deployment. Therefore, the tutorial aims both at database management system and middleware developers, as well as those making use of database technology in planning and deploying large scale applications.

1 Motivation
Replicated database management systems are at the core of modern information systems, given the importance of achieving high availability, fault tolerance and prompt disaster recovery. Consistent database replication technology is key to shared-nothing database clusters, that are an increasingly popular solution to the scalability of multi-tier applications. Early approaches to database replication suffered from performance limitations that constrained the full use of deployed replicas. However, recent advances in consistency algorithms, communication protocols, and replication-friendly extensions to database kernels paved the way for new replication algorithms that are able to offer highly efcient, strongly consistent and fully replicated systems, even in wide-area networks. Such advances are now making the transition from the research labs to concrete products and applications, allowing novel approaches to hard scalability and availability challenges. Familiarity with emerging database replication technology is thus of interest to database management systems and middleware developers, which look forward integrating them in their products, but also to enterprise IT planners involved in the design of next generation applications. The tutorial is offered by researchers involved in the GORDA Open Replication of Databases project,1 which has co-organized the Dependable and Adaptive Distributed Systems track at the ACM SAC conference since 2006 with noticeable success.

2 Outline
The tutorial follows an incremental approach. It introduces basic elements of database replication in an intuitive manner and leverages on these elements to show how full edged solutions can be built. It is organized as a full-day tutorial, where the morning is devoted to the conceptual foundations and the afternoon is concerned with an analysis of practical case-studies. The tutorial starts by making an overview of the challenges of database replication and of the main algorithms and engineering techniques to address these challenges, covering both replica consistency algorithms and the communication and coordination algorithms required to implement database replication. Then, the
1

http://gorda.di.uminho.pt

tutorial provides an in depth description of the architecture of a concrete modern open-source database replication system. Finally, the participants will be provided a guide tour on the conguration steps required to deploy an open source database replication system. In detail, the tutorial includes the following four modules: Database replication fundamentals This module introduces database replication. It starts by introducing the main goals and challenges of database replication. Then it provides a comprehensive overview of the main database replication techniques, including a historical perspective on how these approaches have been adopted in the industry. Communication support Modern database replication algorithms are based on advanced communication and coordination paradigms. This module provides a introduction to the most relevant services that are at the core of fault-tolerance distributed computing such as: reliable group communication, uniformity, and order algorithms. State-of-the art system This module provides a detailed description of a concrete case study, based on existing open-source database solutions. It highlights the main components of a practical database replication system, including the most recent advances on database kernel extensions, optimistic communication, and database replication algorithms. System conguration and deployment Using a concrete open-source system, this module provides an walkthrough to the main conguration and deployment issues when applying an existing database replication solution.

3 Goals
This tutorial aims at providing a thorough overview of the state of the art database replication techniques and systems, both from the fundamental and the applied points-of-view. After the tutorial, attendants should be able to: Relate database replication technology with scalability and availability challenges in current information systems. Compare and combine key communication and replication protocols. Apply innovative database replication technology in the design and development of large scale applications.

4 Expected Background
The tutorial will be given in a self-contained manner. Basic knowledge of database management and distributed systems topics are recommended.

5 Presenters
Rui Oliveira graduated (1991) in Electrotechnic and Computers Engineering by Universidade do Porto, has a Master (1994) in Computer Science from the Universidade do Minho and a PhD (2000) from the Swiss Federal Institute of Technology in Lausanne. He is associate Professor in Computer Science at Universidade do Minho, teaching dependable distributed systems to Masters and Doctoral Programmes. He currently leads the Computer Science and Technology Center of Universidade do Minho.

Rui Oliveira is the Project Manager of the GORDA project an european research project devoted to consistent database replication systems. His current research interests include fault-tolerant and large scale distributed systems, distributed data management and peer-to-peer computing. He is a member of ACM and IEEE.

Lus Rodrigues graduated (1986), has a Master (1991) and a PhD (1996) in Electrotechnic and Computers Engineering, by the Instituto Superior Tcnico da Universidade Tcnica Lisboa (IST). He obtained the "Agregao" in Informatics (2003) by the Universidade de Lisboa. He is full Professor at Departmento de Engenharia Informtica, of Instituto Superior Tcnico da Universidade Tcnica Lisboa (IST) and a senior researcher and Distributed Systems Group of INESC-ID. Previously he was at the Faculdade de Cincias da Universidade de Lisboa where he was a founding member of the LASIGE laboratory and leader of the Distributed Algorithms and Network Protocols group. His current research interests include fault-tolerant and real-time distributed systems, group membership and communication, replicated data management, publish-subscribe systems, peer-to-peer computing and mobile computing. He has more than 100 publications in these areas. He is co-author of two textbooks on distributed computing. He is a member of the Ordem dos Engenheiros, ACM, and IEEE.

Jos Pereira graduated (1995) and has a Master (1998), and a PhD (2002) from the Universidade do Minho. He is an assistant Professor in Computer Science at Universidade do Minho, teaching distributed systems foundations, transactional systems, and data replication topics to Masters and Doctoral Programmes. He is currently the Technical Manager of the GORDA Open Replication of Database european research project. His current research interests include dependable distributed systems, namely, on database replication and large scale group communication.

6 Requirements
A beamer is required. A second one would be very useful during demonstration and hands-on activities. Although not strictly required, attendants with laptops equipped the Java Virtual Machine may also experience the hands-on activities in their own machines.

7 Supporting Material
An annotated version of the tutorial slides will be provided to the workshop participants. The handouts will include an extensive reference list, with pointers to text-books, surveys, and technical papers where the interested reader may nd additional information covering the subjects and technologies addressed by the tutorial.