Week7 - Distributed Systems 600.437 Replication Department...

Info iconThis preview shows pages 1–7. Sign up to view the full content.

View Full Document Right Arrow Icon
Yair Amir 1 Fall 10/ Lecture 7 Distributed Systems 600.437 Replication Department of Computer Science The Johns Hopkins University Yair Amir 2 Fall 10/ Lecture 7 Lecture 7 Replication Further readings: * Distributed Systems (Second edition) Sape Mullender, chapters 7 and 8. (Addison-Wesley) 1994. * Concurrency control and recovery in Distributed Database Systems Bernstein, Hadzilacos and Goodman (Addison Wesley) 1987. * Papers from ICDCS2002 and DISC98 on our www.dsn.jhu.edu/publications web page.
Background image of page 1

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Yair Amir 3 Fall 10/ Lecture 7 Replication • Benefits of replication: High Availability. High Performance. • Costs of replication: Synchronization. • Requirements from a generic solution: Strict consistency – one copy serializability. Sometimes too expensive so requirements are tailored to applications. Yair Amir 4 Fall 10/ Lecture 7 Replication Methods • Two phase commit, three phase commit • Primary and backups • Weak consistency (weaker update semantics) • Primary component. • What happens when there is no primary component? • Replication using group communication.
Background image of page 2
Yair Amir 5 Fall 10/ Lecture 7 Two Phase Commit • Built for updating distributed databases. • Can be used for the special case of replication. • Consistent with generic update model. • Relatively expensive. Yair Amir 6 Fall 10/ Lecture 7 Two Phase Commit Server TCP/IP TCP/IP Server Forced disk write Lazy disk write Send decision
Background image of page 3

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Yair Amir 7 Fall 10/ Lecture 7 Primary and Backups • Backups are maintained for availability only. • Backups can improve performance for reads, updates are sent to the primary by the user. What is the query semantics? How can one copy serializability be achieved? • The user interacts with one copy, and if it is a backup, the updates are sent to the primary What is the query semantics with regards to our own updates? Possible options: Yair Amir 8 Fall 10/ Lecture 7 Primary and Backups (1) P B B
Background image of page 4
Yair Amir 9 Fall 10/ Lecture 7 Primary and Backups (2) P B B Yair Amir 10 Fall 10/ Lecture 7 Primary and Backups (3) P B B
Background image of page 5

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Yair Amir 11 Fall 10/ Lecture 7 Weak Consistency
Background image of page 6
Image of page 7
This is the end of the preview. Sign up to access the rest of the document.

This note was uploaded on 11/15/2011 for the course ECKOW 123 taught by Professor Riuyt during the Spring '11 term at University of Damascus.

Page1 / 18

Week7 - Distributed Systems 600.437 Replication Department...

This preview shows document pages 1 - 7. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online