Update Propagation Strategies for Improving the Quality of Data on the Web

Dynamically generated web pages are ubiquitous today but their high demand for resources creates a huge scalability problem at the servers. Traditional web caching is not able to solve this problem since it cannot provide any guarantees as to the freshness of the cached data. A robust solution to the problem is web materialization, where pages are cached at the web server and constantly updated in the background, resulting in fresh data accesses on cache hits. In this work, we define Quality of Data metrics to evaluate how fresh the data served to the users is. We then focus on the update scheduling problem: given a set of views that are materialized, find the best order to refresh them, in the presence of continuous updates, so that the overall Quality of Data (QoD) is maximized. We present a QoD-aware Update Scheduling algorithm that is adaptive and tolerant to surges in the incoming update stream. We performed extensive experiments using real traces and synthetic ones, which show that our algorithm consistently outperforms FIFO scheduling by up to two orders of magnitude.