Performance Evaluation of Image and Video Cloud Services

With the continuous improvement of cloud computing technology and the increasing demand for image and video applications, a large number of media services have emerged in the cloud service market. Cloud services have high stability and performance. However, the current service level agreement of media cloud services is not comprehensive, and cannot fully represent the actual performance requirements. In this paper, through the comprehensive consideration of cloud QoS and media application requirements, we first design a performance evaluation method for the image and video services provided by different cloud platforms. By analyzing the technical status quo, we select the mainstream image and video cloud services, which include facial recognition, image analysis, OCR, VOD, live streaming and video transcoding. We define the representative evaluation workloads according to the characteristics of the selected cloud services, and design the performance metrics from the perspective of cloud services and specific media service types. Then, we conduct the extensive evaluation experiments of the media cloud services on the different cloud platforms including Google cloud, Microsoft Azure, AWS, Alibaba cloud, Baidu cloud and Tencent cloud, and derive some insights into the media cloud service selection.