Master Thesis - Past Projects - Abstract

Generalized Performance Management of Multi Class Real-Time Imprecise Data Services

ID: LiTH-IDA-EX--05/063--SE

With the increasing demand of real-time services, the applications become more complex in their data needs. Real-time databases usually provide the management of data for real-time data services. However the workload applied to dynamic systems is not always known and the database might face overloads. Then deadlines can be missed and the freshness of the data might be violated. Among the proposed approches to resolve this problems, Robust Quality Management of Differentiated Imprecise Data Services proposed a QoS model that allows the differentiation of services according to the importance of the transactions. Further, the database operator can set the desired nominal and worst-case QoS requirements of the transactions. However, only one QoS requirement can be specified per importance level. In this thesis a QoS model that allows the setting of several QoS requirements for transactions of equal importance is proposed. The performance of the proposed algorithms is evaluated in different experiments. The simulation studies show that the proposed solution satisfy the requirements even during overloads and inaccurate run-time estimation errors.

Keywords:

Author(s): Nicolas Chaufette

Contact: Mehdi Amirijoo