Performance Study of Window Queries

Abstract

Notwithstanding recent tremendous improvement in access times of both memory and harddisk, the ratio between the two remains at about 4–5 orders of magnitude. With the ever increasing complexity of applications, and ever increasing volume of data that need to be retrieved from disk, the number of disk accesses becomes a very important performance parameter to optimize. Data must be organized on the disk in such a way that relevant data can be located efficiently. Such requirement calls for the use of indexing structures that are designed to provide fast retrieval of a small set of answers from the collection of data quickly. An index structure is recognized as the most effective mean in identifying the storage location of data quickly [13]. However, maintaining and searching the index is a cost in itself, and it has been shown that in such cases the cost of using an index may incur higher cost than a straight forward sequential scan. Hence, it is important to verify the performance of a new index against existing methods and sequential scan if appropriate.