In real-time communication [1], the quality of communication depends on the time at which messages are successfully delivered to the recipients. Network Quality of Service (QoS) requirements are typically formulated in terms of performance metrics such as deadline, delay jitter, and loss rate. The deadline is the specific maximum delivery delay bound, which is an application-layer, end-to-end timing constraints. Another important performance metric is delay jitter, which is the maximum variation in delay experienced by packets in a single connection. Some real-time applications such as noninteractive television and audio broadcasting require bounds on jitter but not delay. The delay jitter is usually controlled, at cost of additional delay, by using bufferjs.
In recent years, a number of new classes of distributed applications have emerged, such as remote video, multimedia conferencing, data fusion, visualization, and virtual reality. The development from data-oriented to multimedia services, i.e., from non-real-time to real-time service, requires the QoS guarantees instead of best-effort-delivery. A number of real-time communication protocols have been developed that allow the network service to provide the applications with the required QoS guarantees. The basic paradigm used in such protocols is that of the "performance contract" [9]; during a separate connection establishment process, the network service and the application enter an agreement that the network service guarantees the required QoS as long as the traffic generated by the sender does not exceed a specified amount. If necessary, the network service can police the traffic stream, that is, drop the excessive traffic.
The mechanisms that enable such a performance contract are the following [9]:
With these mechanisms in place, the process of determining whether a connection can be admitted to the system is done during a connection establishment step, which precedes the transmission of data on the connection.
A new connection can be accepted (and therefore established) only if enough resources are available in the system to guarantee the QoS requirements for the new connection on one side, and existing connections on the other. If so, the necessary resources are reserved for the newly established connection.