Off-campus UMass Amherst users: To download dissertations, please use the following link to log into our proxy server with your UMass Amherst user name and password.
Non-UMass Amherst users, please click the view more button below to purchase a copy of this dissertation from Proquest.
(Some titles may also be available free of charge in our Open Access Dissertation Collection, so please check there first.)
Optimistic processing for distributed database systems in a partitioned network
Abstract
This dissertation investigates the problem of supporting optimistic processing for distributed database systems in the face of network partitionings. Two new distributed protocols that provide an integrated approach to concurrency control and partition recovery through the use of timestamps are presented. These protocols offer automatic detection and resolution of conflicts while supporting serializability. They differ from one another primarily in their view of the network. The first supports optimistic processing only while the network is partitioned. The second does not distinguish between a partitioned and a non-partitioned network, and supports optimistic processing at all times. Neither relies on assumptions regarding the state of transaction processing at the time a network partition occurs. Both protocols can handle multiple, overlapping failures, without requiring that failures cannot occur during recovery. The performance of the protocol that supports optimistic processing only during the time of network partitionings was evaluated on the basis of backout rate and CPU overhead metrics. Its performance was compared to that of an existing graph-based optimistic protocol that also supports serializability while offering automatic detection and resolution of conflicts. Results of simulations of these protocols show that while the expected backout rate produced by the distributed timestamp-based protocol is higher than that of the graph-based approach, the CPU time required to detect and resolve conflicts is significantly less for the distributed timestamp-based approach. The difference in CPU overhead for the two protocols is sufficiently high to allow the distributed timestamp-based protocol to have an overall transaction throughput rate equal to, or larger than, that of the graph-based approach. Additional results from these simulation studies provide insight into the effects of the degree of copy replication and the size of partitions on the performance of optimistic processing protocols. In summary, this dissertation presents a new approach to the problem of supporting optimistic processing for distributed databases in the face of network partitionings. Simulation studies offer new insights into factors which affect the performance of optimistic processing protocols.
Subject Area
Computer science
Recommended Citation
Goodman, Ann Tracy, "Optimistic processing for distributed database systems in a partitioned network" (1990). Doctoral Dissertations Available from Proquest. AAI9022691.
https://scholarworks.umass.edu/dissertations/AAI9022691