Paper Review: Quality Adaptation for Congestion Controlled Video Playback over the Internet

Reviewer: Kenneth Chin

The main idea of the paper is to use buffer on the receiver to improve the smoothness of streaming video playback over the Internet.

Good points:
  1. Introduced (not by the author) hierarchical encoding where an video frame is encoded into multiple layers. The base layer is a coarse outline of the frame that is so small that can be sent over the Internet very fast. The subsequent layers are like supplements that improves the quality of the frame. Obviously, the more the layers, the better the quality of the frame.
  2. Good buffer management scheme in the sense that the buffer are efficiently used. I appreciated the idea of efficient buffer usage, but it is not efficient in improving the smoothness of streaming video. (see bad point 1)
  3. Hybrid use of network and buffer to improve smoothness of streaming video, but it is again not the optimal solution (see bad point 2)
  4. Included a smoothness parameter in terms of the anticipated maximum number of backoffs.
Bad points:
  1. Dividing proportionately among different layers; the lower the layer, the biggest is the allocation of the buffer. This buffer allocation scheme is not necessary because even though we divide evenly the buffer among all layers, the amount of buffer needed is at most twice of that used in the paper. However, with the use of more buffer, the video smoothness can last longer in the case of severe network congestion.
  2. The author made an optimistic assumption that by the time the buffered data is used up, the network would recover from congestion and the remaining data can catch up with the buffered data to provide a smooth video playback. However, it is not always the case because during the drain phase, there could be multiple backoffs which significantly lowers the transmission rate.
  3. The paper emphasized a lot on the buffer management techniques that sacrifice performance and quality as well.
  4. Packets of a frame can be lost in the course of transmission. The loss rate may be so high that even base layers cannot be played. Besides, the packets can arrive out-of-order and the situation can be as bad as stalling playing all frames following the frame that has problem. All these things reflects the deficiency of the buffer adaptation scheme proposed in the paper.
Buffer management alone cannot smooth out the video streaming, it is suggested that a hybrid scheme of TCP-friendly congestion control protocol and buffer management is going to work things out.

It is a 1st grade paper which gives no significant contribution.