Exploiting Laziness for Improving Performance in Data Replication Management

In this dissertation, we study how to best explore laziness in data replication for the purpose of performance optimization without sacrificing the ability for fault-tolerance. We also study techniques to deal with various types of failures to ensure transaction durability. First, we propose an Optimized 2-Safe (o2-safe) approach for multi-processor primary-backup systems that combines the best features of both lazy and eager designs by careful addition of partial synchrony in the log transfer protocol. The approach presented achieves high throughput during normal processing (similar to lazy methods) without risking loss of transactions in case of primary failures (similar to eager methods). The failure-handling techniques developed are able to handle both partial as well as complete system failures and guarantee continued availability of a database partition as long as both the primary and its corresponding backup do not fail simultaneously. We then explore a new Optimized Eager-Master (OEM) approach which extends the o2-safe technique to general data replication problems. The resulting OEM approach ensures transaction durability with much improved performance over eager approaches. Furthermore, the non-blocking commit protocol, along with the techniques developed for handling site failures, enables a backup to take over in the event of a failure to provide continued availability. Performance studies further confirm the viability of the proposed approaches.

Name thesis:degree_name

Ph.D.

Level thesis:degree_level

Dissertation

Discipline thesis:degree_discipline

Computer Science

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2015

Author dc:creator

• Hu, Kexiang

Contributors dc:contributor

• Sharad Mehrotra

• Computer Science

Language dc:language

eng

Identifier

(MiAaPQ)AAI9990024

OAI identifier oai:identifier

oai:www.ideals.illinois.edu:2142/81986