In this paper, we propose an asynchronous cluster video server architecture, which is quite different from synchronous video server architecture in various aspects such as striping policy, placement policy, and server scheduling algorithm. This architecture has the benefits that the system scales unlimitedly with the number of disks and videos. We discuss the main design issues and compute the buffer requirement at both client side and server side, scheduling delay, and prefill delay. The result is exciting compared with synchronous server scheduling algorithm. It requires fewer client buffers and server buffers, consumes less system response time and can be scalable. To improve system performance, group sweeping scheduling (GSS) is applied.