Prioritized query shedding technique for continuous queries over data streams

Document Type


Publication Date

Summer 12-14-2009


In data stream applications, the unpredicted fluctuation of the arrival rate, along with continuous processing of posted queries; is one of the main problems that may result in an overloaded system. Load shedding is one of the commonly used approaches in data stream management systems to overcome this problem. Most proposed load shedding techniques in the literature are based on shedding parts of the stream in order to relax the processing load. Although these techniques have proven successful relaxation to the processing load, they do affect the quality of continuous queries’ results as most of these techniques use approximation. In this paper we investigate different load shedding techniques in the context of data stream processing systems and their effect on the quality of continuous query results. We also propose a new technique that handles the overloading problem when considering queries’ priorities. Unlike most shedding techniques that assume all queries are equally important, or even prioritized shedding techniques that are based on dropping input tuples according to the regions’ priorities, our technique considers the priority of a query as a whole. Our results have shown that query shedding is more effective when having queries with common regions, without affecting results quality.