Priority-Based Continuous Query Scheduling in Data Stream Management Systems
Lory Al Moakar
Wednesday April 11, 2012
1:00 pm - Sennott Square - Seminar Room 5317
AbstractThe emergence of Data Stream Management Systems (DSMS) facilitates implementing many types of monitoring applications via continuous queries (CQs). However, the quality-of-service (QoS) requirements of detecting events are different for different monitoring applications. For example, the CQs for detecting anomalous events (e.g., fire, flood) have stricter response time requirements over CQs which are for logging and keeping statistical information of physical phenomena. Traditional DSMSs treat all the CQs as being equally important in the system and attempt to optimize their overall performance. In particular, they typically employ a CQ scheduler to decide the execution order of CQs to achieve a global performance goal and thus perform badly in an environment where CQs have different importance levels.
The hypothesis of my research is that there is a need for a suite of schedulers that optimizes the response time of critical queries while satisfying the requirements of the other, less critical classes and taking into consideration the underlying processing environment. Towards this, we first developed the Continuous Query Class (CQC) scheduler for single-core / single-process systems which is assumed by many of the current prototypes, including our own, AQSIOS. Then, we propose the Adaptive Broadcast Disks scheduler (ABD) which is more suitable for dual-core environments. Currently, we are extending our work to multi-core environments to take better advantage of modern machine architectures and their processing capabilities. The goals of the multi-core scheduler are to optimize the response time of the critical queries while maintaining acceptable performance for non-critical ones. However, it also needs to utilize the cores efficiently and provide better performance using the new platform. We demonstrate the effectiveness of our schedulers using new metrics under AQSIOS, our prototype DSMS, and SimAQSIOS, a simulator that mimics AQSIOS very closely.
Dissertation AdviserDr. Alexandros Labrinidis and Dr. Panos K. Chrysanthis, Department of Computer Science
Committee MembersDr. Kirk Pruhs, Department of Computer Science
Dr. Mohamed A. Sharaf, School of Information Technology and Electrical Engineering, University of Queensland