|M.Sc Student||Yaron Banai|
|Subject||Aggregation and Tunneling Schemes for Scalable QoS|
|Department||Department of Electrical Engineering||Supervisor||Full Professor Orda Ariel|
The Internet was originally designed for “best effort” service. Recently, the desire for new network applications has lead to the definition of a scheme for assuring Quality of Service (QoS). This scheme, based on per flow reservation, defined in several IETF standards and known as Integrated Services (IntServ), was not widely accepted in the Internet due to scalability problems. New standards such as Multi-Protocol Label Switching (MPLS) and Differentiated Services (DiffServ) allow classification and treatment of aggregates instead of individual flows.
In order to utilize these new standards so as to overcome the scalability issues, tunnels for the aggregates must be set up, with the purpose of reducing reservation state and messages treated at the nodes. Such tunnel setup is the focus of this work. Specifically, the scheme should minimize the state in the most congested node and the number of messages it treats during session set up and tear down. The tunnel layout can be static or dynamic. Our work focuses on the dynamic case. As we indicate, our results can be used to treat the simpler case of static layouts. This work defines a corresponding optimization problem in mathematical terms and, based on mathematical analysis, it establishes a simple heuristic approach to dynamically set up and tear down aggregates.
Previous work, related to ATM networks, treated scalability issues of connection set up. However, that work considered a different optimization criterion, making the schemes presented in previous studies unsuitable for our purpose. This work is also unique in that it treats four different classes of the state optimization problem,controlled by two parameters:
By treating these four different classes of the optimization problem, this work establishes an aggregation scheme that can be used to design scalable QoS, multicast routing with reduced information, and efficient optical networks.