Techniques of data prefetching, replication, and consistency in the Internet

Internet has become a major infrastructure for information sharing in our daily life, and indispensable to critical and large applications in industry, government, business, and education. Internet bandwidth (or the network speed to transfer data) has been dramatically increased, however, the latency time (or the delay to physically access data) has been reduced in a much slower pace. The rich bandwidth and lagging latency can be effectively coped with in Internet systems by three data management techniques: caching, replication, and prefetching. The focus of this dissertation is to address the latency problem in Internet by utilizing the rich bandwidth and large storage capacity for efficiently prefetching data to significantly improve the Web content caching performance, by proposing and implementing scalable data consistency maintenance methods to handle Internet Web address caching in distributed name systems (DNS), and to handle massive data replications in peer-to-peer systems. While the DNS service is critical in Internet, peer-to-peer data sharing is being accepted as an important activity in Internet.;We have made three contributions in developing prefetching techniques. First, we have proposed an efficient data structure for maintaining Web access information, called popularity-based Prediction by Partial Matching (PB-PPM), where data are placed and replaced guided by popularity information of Web accesses, thus only important and useful information is stored. PB-PPM greatly reduces the required storage space, and improves the prediction accuracy. Second, a major weakness in existing Web servers is that prefetching activities are scheduled independently of dynamically changing server workloads. Without a proper control and coordination between the two kinds of activities, prefetching can negatively affect the Web services and degrade the Web access performance. to address this problem, we have developed a queuing model to characterize the interactions. Guided by the model, we have designed a coordination scheme that dynamically adjusts the prefetching aggressiveness in Web Servers. This scheme not only prevents the Web servers from being overloaded, but it can also minimize the average server response time. Finally, we have proposed a scheme that effectively coordinates the sharing of access information for both proxy and Web servers. With the support of this scheme, the accuracy of prefetching decisions is significantly improved.;Regarding data consistency support for Internet caching and data replications, we have conducted three significant studies. First, we have developed a consistency support technique to maintain the data consistency among the replicas in structured P2P networks. Based on Pastry, an existing and popular P2P system, we have implemented this scheme, and show that it can effectively maintain consistency while prevent hot-spot and node-failure problems. Second, we have designed and implemented a DNS cache update protocol, called DNScup, to provide strong consistency for domain/IP mappings. Finally, we have developed a dynamic lease scheme to timely update the replicas in Internet