This work proposes a new method to variable bit rate video traffic model used to capture the multi-level characteristics that are significant in determining buffering requirements. While lots of the previous work in video traffic modeling has paid attention on real-time multimedia systems that of necessity engage queuing models with small buffers, there are increasing interests in awareness that high quality video is downloaded in non-real-time from a source like the Internet, and then spread at a later time in a regional environment to a very low-cost, so resource constrained devices in which smooth playback and prevention of data loss are of vital importance. This paper is focusing on the tradeoffs between the quality of video and resources' buffers. It also provides an architecture that is directly relevant to a large group of up-and-coming applications.