Latency is the time it takes for packets to travel across a network from a source to a destination.
ITU recommendation G114 recommends that network latency of 400ms should not be exceeded, and real-time traffic latency should be no longer than 150ms.
Network latency can be broken down into four categories:
- Fixed propagation delay
- Fixed serialisation delay
- Fixed processing delay
- Variable delay
Propagation delay is the time it takes for a packet to travel from the source to a destination at the speed of light over fibre-optic cables or copper wire.
The speed of light is 299,792,458 meters per second in a vacuum. These conditions do not exist in a fibre optic cable or copper cable so the speed of light is slowed down by a ratio known as the refractive index. The larger this index the slower light travels.
In fibre optic cables, the average refractive value is around 1.5.
Serialisation delay is the total amount of time it takes to place all the bits of a packet onto a physical link. It is a fixed value that is dependant on the link speed. The faster the link, the lower the delay.
The processing delay is the mount of time for the networking device to take the packet from an input interface and place the packet onto the queue of an output interface. The total delay is dependant on a few factors such as CPU speed, CPU utilisation, IP packet switching mode, the router architecture and any configured features on both the input and output interfaces.
Jitter is the difference in latency between packets in a single network flow. Causes of variable delay in packets within a flow can be a queuing delay, de-jitter buffers, or variable packet sizes.
Jitter is caused by packets involved in network congestion. It can be part of a queuing delay due to other packets in a interfaces queue, a slow link speed, or a queuing mechanism. There are unequal delays for each packet, introducing jitter.
Voice and video technologies usually contain de-jitter buffers that help smooth out changes in packet delay from jitter. It can adjust for around 30 milliseconds changes in arrival times of packets. If a packet exceeds this approximate 30ms delay, it is dropped – this affects the overall voice and video quality.
It is recommended to use technologies such as low latency queuing to allow matching packets to be forwarded before any other lower priority traffic during congested periods.
Packet loss is usually a symptom of a congested interface. It can be usually be resolved with one of these solutions:
- Increase the link speed where congestion is being detected
- Configure Quality of Service congestion avoidance and management technologies.
- Configure traffic policing to drop lower priority packets, and permit higher priority traffic onto the link
- Configure traffic shaping to delay packets instead of dropping them. This is not recommended for real time traffic as it can cause jitter.
Traffic shaping is not a recommended solution to handle data bursts that occur on microsecond time intervals. A specialised low-burst shaping technology is required for micro-bursts that need to be smoothed out by a traffic shaper.