Latency defines the delay in which the video stream will be delivered to the end point. This delay is essentially a buffer that is used for additional processing, thereby improving the quality of video stream. The latency should be considered based on the importance of immediacy. For example, if the video requires interaction (e.g. between a reporter in the field and an anchor in the studio), the latency should be set to a minimum so that the interaction does not suffer from any delay.
Latency can vary from tens of milli-seconds to several seconds.
ZiXi protocol can start provide benefit with a delay of a single frame (30ms), while increasing the latency will provide more robustness for the error-recovery.
In low latency (below 1500ms) the error recovery is based on a hybrid approach of FEC and ARQ. FEC will add overhead of up to 30% (adaptive–without congesting the network) .
To allow sufficient error-recovery it is recommended to set the Latency at least 3 times larger than the RTT (i.e Latency >= 3* RTT)
Latency can be set to any value, we can recommend two common modes:
- Interactive mode - 500ms
- Non-interactive mode – 4000ms
Latency below 1500ms will add an FEC overhead of up to 30%.