The notion of quality of service
The term QoS (an acronym of "Quality of Service") refers to the ability to provide a service (in particular a communication media consistent with requirements in terms of response time and band width).
Applied to packet switching networks (networks based on the use of routers), QoS indicates the ability to be able to guarantee an acceptable level of packet loss.
In fact, contrary to circuit switching networks, such as the switched telephone network, where a communication circuit is always dedicated to communication, it is impossible on the Internet to predict the path taken by the different packets.
So, nothing guarantees that a communication requiring constant bandwidth will be able to take place without incident. That is why there are mechanisms, called QoS mechanisms, making it possible to differentiate different network flows and reserve a share of the bandwidth for those requiring continuous service, without breaks.
The term "service level
" defines the requirement levels for the capacity of a network to provide a point to point or end to end service with a given traffic. Generally there are three levels of QoS defined:
- Best effort provides no differentiation between several network flows and provides no guarantee. This service level is therefore sometimes called lack of QoS.
- Differentiated service or soft QoS making it possible to define priority levels for different network flows without however providing a strict guarantee.
- Guaranteed service or hard QoS comprising of reserving network resources for certain flow types. The principal mechanism used to obtain such a service level is RSVP (Resource reSerVation Protocol).
Quality of service criteria
The principal criteria enabling quality of service to be assessed are as follows:
- Bandwidth: defines the maximum volume of information (bits) per unit of time.
- Jitter: represents fluctuations in the digital signal, in time or phase.
- Delay: this characterises the delay between transmission and receipt of a packet.
- Packet loss: this relates to the non delivery of a data packet, most of the time this is due to network overload.
- Desequencing: this is a modification to the arrival order of packets.