Computing resource minimization with content-aware workload estimation in cloud-based surveillance systems

As cloud computing platform provides computing power as utilities, it is important to develop a mechanism to adaptively adjust the resources needed for handling cloud service. In this paper, a computing resource minimization framework for cloud-based surveillance video analysis systems is proposed. Videos streams are divided into clips and multiple processing nodes are used to handle clips. While the quality-of-service (QoS) is maintained, the proposed framework dynamically adjusts the number of processing nodes based on a proposed content-aware workload estimation mechanism. Experimental results show that the proposed mechanism successfully predicts the variability of system workload while QoS is maintained and outperforms other mechanisms in terms of average virtual machine (VM) quantity and job failure ratio.