Supporting Cloud IaaS Users in Detecting Performance-Based Violation for Streaming Applications

Abstract

Cloud infrastructure-as-a-service (IaaS) provides on-demand resources at low cost. IaaS users anticipate stable performance for their applications so that the applications' Quality of Service (QoS) constraints are satisfied. However, they generally experience performance variability in their deployed applications, primarily due to multi-tenancy. Such variability may lead to QoS violations for the applications. Existing cloud infrastructure solutions offered by different cloud providers (CPs) do not have facilities to detect such QoS violations. Hence, cloud users need to take responsibility for monitoring the performance of their applications in order to detect application-specific QoS violations. To support cloud users on this, researchers have proposed a few user-centric cloud monitoring frameworks. However, these frameworks do not detect application-specific QoS violations. In this paper, we propose a novel algorithm for detecting QoS violation for media streaming applications. At each detection instance, the algorithm compares the cumulative value of the expected streamed data against the cumulative value of the measured streamed data. Based on this comparison, the algorithm may raise QoS violation alarms. We evaluate the algorithm by deploying a media streaming application in a lab-based cloud set-up. Experimental results demonstrate that the proposed algorithm can reliably detect QoS violation for streaming applications deployed in the cloud.