Possible Uses of our Model

Our model can be used mainly for: $\bullet$

• Dynamic bandwidth allocation and dynamic contract negotiation: once a good traffic predictor is built, the BW can be allocated dynamically, instead of statically as it is currently done. In this case, an enhanced version of our proposed model can be used to dynamically renegotiate the contract with the service provider such that the clients can use only what they will consume.
• Dynamic traffic shaping: other direct use of our technique is within some large institutions or enterprises structured in groups where there a given protion of the BW is allocated for each group. For example in the university, one can find an organization where there is a given BW for the students, employees, staff, research, ..., which is statically allocated. Our model can be used so that each group can have roughly what it needs. The rest can be redistributed to the others who may starve for small BW. For example, suppose that the total BW is 100Mbps, the dean has fixed part of 10Mbps, and the students have 30Mbps. Not all the time the dean is working in his office, while the students' network may be congested from time to time, our model can be used to detect automatically that the dean is not working now and decide that the unused BW could be given to the students. Of course in such a scenario, a separate NN should be used for each group's local network.
• Anticipating the best bit rate and encoding parameters the source has to use to deliver the best real-time multimedia quality. Traditionally, the current congestion reacting protocols (with the aim of maximizing the QoS) operate based on the current network state (congestion level) and the receiver asks the sender to modify the sending parameters (bit rate in general) accordingly. Obviously, there is a certain time needed to react. During this time, the quality can remain poor. By using a traffic predictor coupled with the real-time quality evaluation we have presented in the previous chapters, and the rate control protocol we have presented in Chapter 8, the source can anticipate the state of the network and modify the parameters, if needed, in advance before getting the network congested and degrading the quality. This will improve both the quality of the multimedia flows and the network availability.