Hash Based Caching Mechanism for Web-OLAP

In recent years, the advent of client-side data caching techniques has brought significant performance improvement in highly interactive web applications. In this thesis, we focus on the client-side data caching mechanism in the web-based OLAP system. Previous work or methods on this area lack the universality of dealing arbitrary shapes of the cached data. For instance, in 2 dimensional data, it can only calculate the overlaps between two rectangles. While in reality, in most cases, the cached data can be any sizes and shapes. Therefore, we propose a new client-side hash based caching mechanism that can deal with arbitrary shapes of the dataset. Also, we come up with the idea of downloading the inflated sub-cubes in background and serve the users first. By combining this idea with our hash based caching mechanism, we can not only serve the users at the very first time, but also provide high performance for the subsequent cross-tab or drill-down operations by utilizing the cached inflated sub-cubes. Our result shows that this new method has great performance improvement compared with the previous method and when the size of the inflated sub-cubes is not so large, our mechanism also outperforms the no-cached one.