In video grid system, network services are considered as grid resources. Every network management domain has own management strategies and the parameters of quality of service. A video-on-demand service is a composition services which composed of sequential network services and a video server node service. A video-on-demand service aggregate all the network QoS parameters and the video server's QoS parameters. Selecting the most suitable service for user task is a problem of multi-objective global optimization. We employ the expansion and supplement of fuzzy analytic hierarchy process (FAHP) to solve this problem. In our system, the video data packets are transported by TCP protocol. According to the TCP acknowledgment mechanism and utilizing one-point passive monitoring, this paper realizes to measure the network performance parameter of user's video stream on client, such as delay, bandwidth, jitter, and late rate. Finally, an example was presented to demonstrate reasonable and feasibility of these proposed methods. (4 pages)