Quality of Service

Quality of Service (QoS), when measured in communications networks, indicates the overall performance of the connectivity. For measuring the network QoS, the most common QoS statistics include delay, jitter (delay variation), and packet loss. As statistics already indicate, QoS is always a measure carried out between two points of interest in the network. Qosium measures QoS from network traffic level, indicating the QoS experienced by applications.

1. Why QoS Is Important #

Applications have a diverse mix of QoS demands. For example, a real-time video stream like a camera stream for remote controlling requires low delay at all times while the video codecs are built to be resilient to slight packet loss. Industry automation applications typically require both very low delay and zero packet loss to work safely and efficiently. On the other hand, the over-the-top streaming media services such as Netflix and YouTube demand throughput capacity rather than low delays and packet loss levels. Content buffering in the end-device takes care of sporadic QoS variation in the connectivity.

Is maximum throughput performance QoS? No. Tools that allow you to measure your connection’s speed only tell how much capacity there’s in reserve in your network connection at that time. It may give you a hint that the capacity likely allows your online game or a high-quality video stream to run smoothly. But, the throughput capacity tends to change over time, depending much on the network load and signal quality when employing a wireless network. A Ping test often accompanies maximum data speed measurements. That’s QoS, but it tells the QoS only for that Ping application. Some other applications over the same connection likely get different QoS.

In telecommunications terminology, QoS is often misleadingly referred also to as data prioritization. For example, many wireless systems are equipped with QoS class definitions to manage different traffic flows differently. For example, there are dedicated bearer definitions for different types of traffic and subscriptions in mobile networks. DiffServ is one commonly known standard for classifying and managing IP traffic. While these solutions attempt to provide the application traffic with satisfactory QoS, they cannot guarantee that. For example, data prioritization alone cannot battle low radio link quality efficiently. Moreover, what is missing in these QoS class definitions is that they do not report the realized QoS for the traffic. Thus, they don’t indicate what the QoS has been over the network path.

For end-users, often, the only thing that matters is that the application used works as it should. The application works when it gets the QoS it demands for from the network. The application does not care about the underlying network technology. It can be wired or wireless, as long as delays, delay variation, and packet loss stay within the tolerable boundaries. Real-time QoS measurement enables you to monitor how well the network can serve your applications and how usable your applications are.

2. Typical QoS Statistics #

The table below introduces a few main QoS statistics. Application QoS demands to determine which of them play the essential role. As the statistics indicate, they are such that you always need to have a reference point over which they can be calculated. For example, calculating delay for an IP packet necessitates that you know when that packet has been sent on the other end of the measurement path. As this is not simple to be carried out, the typical way has been to measure delay as a round-trip-time (RTT), like what Ping tells you. The round-trip-time cannot tell you what has the delay behavior been in the sent and receive directions, which can sometimes vary much. If the packet involved in the round-trip measurement is lost, you cannot know which direction caused this loss. Having one-way statistics helps you analyze network problems in more detail.

Statistic Description
Delay Time difference between the time data is transmitted to the moment it is received by the other measurement point. Sometimes delay is also called latency.
Jitter Time how much delay differs between sequential packets
Packet loss ratio Ratio between successfully transmitted and lost packets on the measurement path
Connection break Time duration from the moment a packet loss is detected to a moment packet is successfully transferred