Fast Convergence Counts
To reduce network bandwidth, video is highly compressed. Any loss likely causes visible artifact for a varying amount of time. Watch this short video and see the impact of packet loss, jitter and delay on video.
Video is made up of a series of pictures. Video compression formats typically use a technique known as ‘difference coding’ by comparing the difference between the current video frame with the preceding frame to ensure that information which does not change (e.g. static background) is not repeatedly transmitted. The most popular and widely deployed encoding techniques are MPEG-2 and MPEG-4.
Figure 1: MPEG Encoding – GOP
When encoding the video frames the encoder splits each video frame into specific types. These different types of frames are known as I-frames, P-frames and B-frames. The I-frame is encoded without any reference to any other frames. A P-frame uses a previous I or P frame. A bi-predictive B-frame makes reference to both a preceding reference frame or a future reference frame.
Network does have glitches from time to time. If one of the I-frames is lost, then subsequent B or P frames cannot be decoded. This dependency of P or B frames on data contained in other frames means that a single packet loss can impact the entire Group of Pictures (GOP). In fact, studies have shown that a 0.8 sec of loss can potentially translate to as much as a minute worth of bad video.
Switches play a key role in a video-centric enterprise network. Minimizing video packet loss provides huge benefits for user experience. Cisco switches have a variety of features to minimize packet loss:
- Layer two convergence (L2)
- Layer three convergence (L3)
- High Availability (HA) features
It is important to measure convergence times for unicast and mulitcast traffic as they can vary significantly. With these fast convergence features configured, you will find that all convergence times are significantly improved for all the Cisco switches.
Figure 2: Network Convergence matters
Video applications have stricter SLA requirements than non-real time traffic. Effects of losing video packets has an effect on visual behavior that exceeds the packet loss duration itself. So if you have real time traffic, enable these fast convergence features on your switches.
Learn more: http://www.cisco.com/go/medianet