Diversified QoS Expectations

Different people may have different expectation to the QoS of a network. For example, a regular user may expect

the best quality at the lowest price,
the demanded quality at the lowest price,
the best quality at the acceptable price, or
the torlerable quality at the lowest price.

On the other hand, a network operator may have a total different QoS objective in offering network services depending on his/her strategic objective. For example, he/she may offer:

the best quality at the acceptable price,
the acceptable quality at the highest price, or
the tolerable quality at the lowest price.

Instead of assuming a speific QoS expectation/objective, the objective of our research is to design a flexible infrastructure for QoS management as well as associated QoS management tools so that they can be used in most networks. Operators can customize their network management systems to achieve their strategic objectives. A popular objective is to satisfy as many customers (or to make as much profit) as possible within the available resources.

UMTS Application Classes

DiffServ Services Classes

Quality Entropy

QoS requirements of users and service classes of an underlying network are all diversified as shown in the previous two sections. In order to make our system flexible to support most systems, we define a generic metric to measure the quality of a system, quality entropy, which is actually a reciprocal of quality level. It can be an arbitrary combination of concerned quality parameters such as delay time, packet loss rate, and jitter, depending on the need of network operators. To be consistent with the semantic meaning of these popular quality parameters, the metric is defined in such way that it is the lower the better.

In this reserach, we assume that the users of our QoS tools, who would probably be a network operator, will define the formula of quality entropy by themselves.

To use quality entropy, the network planners need to convert their quality parameters into quality entropy first, and convert the results of planning back to the original parameters.