A QoS Management Scheme for Real-Time Data Services

University Of Virginia Main Campus, Charlottesville VA

Investigators

Abstract

The demand for real-time data services has been increasing recently, as many real-time applications are becoming sophisticated in their data needs. A real-time database, a core component of many information services for real-time applications, can be a main service bottleneck. Current databases are not time-cognizant, are poor in supporting temporal consistency of real-time data and real-time execution of transactions, and therefore they do not perform well in these applications. To address this problem, a scheme for QoS management for real-time data services is proposed that provides guarantees on transaction timeliness and data freshness, which are considered two key performance metrics for real-time data services. The fundamental premise motivating this research is the need for real-time data service systems that are autonomous, data-intensive, and operable in unpredictable environments. The novel idea of using feedback control based QoS management to provide robust performance guarantees against unpredictable workload, incomplete knowledge, and load disturbances will be pursued. The scheme will be evaluated to validate the contributions using simulation. Success of this research would benefit an ever increasing set of applications, including several national critical infrastructures, that require real-time data services.