THESIS
2002
xiv, 163 leaves : ill. ; 30 cm
Abstract
With the growth of the Internet, clients abandoning their connections due to excessive downloading delays translates directly to profit losses. Hence, minimizing the latency perceived by end-users has become the primary performance issue comparing to more traditional ones, such as server utilization. In this thesis we first survey research on the two more widely used techniques to improve Web responsiveness namely, proxy caching and server replication/mirroring. We argue on their complimentary roles for decreasing client perceived response time and proceed by illustrating the main topics that affect their successful deployment on the Internet. Concerning replication we turn our focus on allocation issues namely, what to store where. We formulate the problem as a constrained optimization...[
Read more ]
With the growth of the Internet, clients abandoning their connections due to excessive downloading delays translates directly to profit losses. Hence, minimizing the latency perceived by end-users has become the primary performance issue comparing to more traditional ones, such as server utilization. In this thesis we first survey research on the two more widely used techniques to improve Web responsiveness namely, proxy caching and server replication/mirroring. We argue on their complimentary roles for decreasing client perceived response time and proceed by illustrating the main topics that affect their successful deployment on the Internet. Concerning replication we turn our focus on allocation issues namely, what to store where. We formulate the problem as a constrained optimization one and provide fast heuristics that achieve good solution quality both when the access patterns remain static and when they change in time. Moreover, we investigate the potential of minimizing the page retrieval delays by downloading the various page components in parallel from multiple servers. Of particular interest is the case of multimedia repositories which we examine in more details. In proxy caching we identify two significant limitations of current schemes. The first one is the inability to cache dynamic pages carrying query results. We use an active caching framework that enables caching such pages at current Web proxies. Intuitively, the benefits are expected to be higher in the case of computationally expensive queries. Therefore, we turn our attention to On Line Analytical Processing (OLAP) queries that often involve aggregations of millions of rows. Such queries are typically issued by decision makers towards the Data Warehouse of an enterprise. Finally, we motivate the need for adaptive strategies that organize proxies in dynamically reconfigured cooperative groups. Two cases are distinguished, depending on whether the neighborhood relations are symmetric or not. We formulate the problem as a second level caching and experimentally prove the viability of our framework.
Post a Comment