Research On Resource Prediction And Allocation Of Video Surveillance Cloud Platform Based On Docker

Nowadays, with the large-scale deployment of video surveillance equipment, the traditional video surveillance system is unable to deal with the massive amounts of video surveillance data for effective calculation and analysis. In recent years, cloud computing technology is developing, and the related researches focus on the video surveillance system which is integrated with the cloud computing technology.Basically, the current mainstream video surveillance cloud platform provides services for user in the form of virtual machine, which results in the large resource fragmentation and the performance loss. The cloud platform allocates the resources by performing a static allocation in the initial stage, and then in the running period, it adaptively rebalances resources according to monitoring alarm or load predication strategy.However, the way of monitoring alarm does not consider the instantaneity of the resource requirements, and it adopts a passive telescopic way that will violate the SLA easily. Although the current load prediction method is proactive, it is unsuitable for video surveillance cloud platform to make the accurate prediction about the load of video services.This paper focuses on how to improve the resource utilization of the video surveillance cloud platform and realize an efficient and flexible scalable video surveillance cloud platform. Firstly, this paper analyzes the load characteristics of the video surveillance services, and proposes a resource prediction model for video surveillance cloud platform. The model construction process can be divided into two phases. In the first phase, we predict the initial resource demand based on the performance of video services. In the second phase, we predict the video workload based on the resource demand time series similarity analysis. Moreover,as for lackness in re-allocating resources adaptively and accurately in video surveillance cloud platform, this paper gives a strategy to optimize the existed video surveillance cloud platform by adding a resource prediction module and adjusting the resource allocation strategy to improve the resource utilization rate. At the end of this paper, we realize the optimization strategy of the video surveillance cloud computing platform based on Docker, and conduct the extensive performance experiments. The experimental results show that the resource prediction model proposed in this paper has higher accuracy and the adjusted resource allocation strategy can effectively improve the resource utilization of the video surveillance cloud platform.