Resource Management Techniques, Video Surveillance Systems

Recent development of computer and network technologies has enabled a boom of video stream applications. However, how to realize the video application with the high quantity stiH faces many challenges: How to satisfy video flow's isochronous request in a general operating system; How to implement the real time transport over Internet; How to map the parameters of high layers to low layers, and to control and management the resources based on these parameters. Because traditional OS and network can't provide end-to-end QoS guarantee for video streams without real time extensions. This thesis is an effort to identify some of the technical challenges and to provide solutions.The thesis constructs a QoS model of the resource management. This model divides the video stream system into application, system and network layers. Resources are controlled and managed separately and QoS parameters are mapped among layers. Application layer accepts user's command and returns the results of lower layers. System layer controls and manage CPU, memory, and other device resources. Network layer provides real-time transmission and resource reservation by using proper protocols.The thesis presents the design and implementation of the CPU resource manager with QoS principles and mechanism. The CPU manager employs EDF to meet the real time requirements of video stream.Further more, this paper discussed a novel disk scheduling algorithm for Multimedia File Server, and puts forward an adaptive allocation Scheme of disk I/O service period according to application load.The solutions distinguish themselves by providing more practical methods for guaranteed video stream applications while leads to high resource use.