How to use MOS Score to Monitor VoIP Video and Audio Application Performance?
The mean opinion score (MOS) measures the quality of experience of real-time telecommunications such as video, audio, and audiovisual applications. MOS has become a useful metric in predicting VoIP call quality, despite weaknesses in accurate measurements based on its subjective nature.
How is MOS measured?
MOS is typically collected in a survey style, where respondents assign an opinion about call quality on a predefined scale, generally 1.0–5.0. The table below is the accepted scale for the popular G.711 audio codec based on extensive testing performed in sound controlled rooms with countless subjects. The maximum score for the G.711 codec is 4.4, where the quality of a perfect 5.0 score is synonymous to the clarity of an in-person conversation.
|G.711 Codec MOS scale|
|Some users satisfied||3.6–4.0|
|Many users dissatisfied||3.1–3.6|
|Nearly All users dissatisfied||2.6–3.1|
How to calculate MOS?
The accepted method for calculating MOS is to take the arithmetic mean of all the opinion scores collected from test subjects. (R = individual ratings from N = subjects).
Application Performance and MOS Score
Because MOS is a subjective call quality indicator, by connecting it with other objective data, insights can be gathered that will support performance decisions.
In the case of choosing a codec, the G.711 is popular because it is a non-compressed codec, a characteristic that bolsters its signal against packet loss, which is the main challenge in real-time communications. Compared to the popular G.729 codec, which offers signal compression, the G.711 codec requires much more bandwidth. However, due to this bandwidth compression advantage, the G.729 signal will degrade more sharply as latency increases. The comparison chart shows at a latency of 100 ms the G.729 codec begins to degrade significantly. By 200 ms, the G.711 codec is still considered by most users as very satisfactory, whereas G.729 is in danger of dissatisfying nearly all users.
MOS is an imperfect measurement, but by providing the right context, it’s a measurement that can be extremely useful in making configuration decisions.
What factors impact MOS?
Generally, any factor that helps reduce call quality will positively impact MOS. The following are specific areas that network administrators can address, improvements in any of these areas will increase the network’s MOS.
- Packet Loss — Packet loss is the occurrence of missing data packets in a network transmission resulting in an incomplete transmission, which is dire enough can noticeably degrade real-time transmissions. Incomplete file transmissions typically trigger a recall of missing data packets, which can be noticed, for example, when webpages load slowly. In real-time video and voice, packets typically are not recalled if they are not available when it is time for their playback (known as deadline expiry), this is what gives weak real-time transmissions their “stuttering” characteristic. Packet loss can be improved by ensuring network traffic remains uncongested. Solutions such as Cisco Application Visibility and Control (AVC) help to visualize, understand, and control network traffic with the aim of anticipating and preventing network traffic issues.
- Propagation Delay — Propagation delay refers to the distance that a signal must travel in order to reach its destination. The greater the distance, the longer the transportation time. By using efficient technology, like fiber optics over coaxial cables, propagation delay can be reduced. However, the maximum limitation is the limit of the speed of light, the fastest a signal could theoretically travel.
- Packetization Delay — Packetization delay is the duration of time it takes to digitally encode an analog signal, i.e. voice into digital signal. For example, the G.729 codec uses compression which consumes less bandwidth than non-compression codexes, but also adds packetization delay to the overall signal delay. .
- Packet Delay Variation or Jitter — Network jitter, a.k.a. packet delay variation (PDV), is a stuttering like effect in signal quality because of inconsistent packet delays in a data transmission. Each packet in the transmission may be routed differently to its destination causing packets to arrive out of order or not at all (called packet loss). Though technology will handle this situation and put the packets back in order, it does cause delays. To illustrate the impact, in cases of high jitter video calls or VoIP, users will experience stuttering video, intermittent voice or dropped calls when speaking to others over the internet.
- Bandwidth Demand Fluctuations — Bandwidth is the amount of data that can simultaneously travel across a network. To illustrate, the diameter of a hose is its bandwidth, the wider the hose the more drops of water can go through the beginning and out the end in one go. Using the same analogy, network latency would be the time it takes for the first drop of water to go from the beginning to the end of the hose.
- Outdated Hardware — Network hardware becomes outdated and glitchy over time. If the traffic load on the network increases significantly then the firewalls, routers, and network switches that support that network must also grow to ensure the free flow of traffic.
The Cisco Application Visibility and Control (AVC) solution deploys a holistic approach for managing quality of service (QoS) technologies. It intelligently prioritizes traffic for critical applications while reducing or preventing traffic from noncritical or unwanted applications in an attempt to improve network and application performance over a wide area network (WAN).