Abstract of the thesis proposal
The term “quality” as it is commonly understood in the context of QoS is “something” by which a user of the service (in a very large meaning) will judge how good the service is.
Today, a lot of different services from many distinct application areas ask for some side needs, sometimes preferable to provide, but sometimes strictly connected to the service itself. Consider, for example, a streaming video transmission on a network: surely it is important that the data-packet with a given video-frame arrives at destination, but it is also important, at the same time, that the packet arrives within precise time bounds, otherwise it would be useless to reproduce it.
The application fields of this framework can be represented by i) Networks and principally Constraint-Based Routing, ii) Web services, for the negotiation of Service Level Agreements, and iii) Trust Management and trust inference decision However the application areas are not exclusively restricted to these three classes: think for example to an extension concerning the QoS management for Workflows.
Examples of QoS metrics for the network links can be the bandwidth, delay, jitter the number of hops, probability of packet loss, for Web Services can be the cost or the availability, and for Trust Management, the trust decision depends on the chosen trust metric.
Our wish is to model the entities that ask and provide the resources connected to services, and the negotiation interactions among them, for example by using Soft Concurrent Constraint Programming.
For this reason, we reckon that the use of Soft Constraints as the fundamentals of our formal framework, could be promising a decision. As a matter of fact, we think that the c-semiring is a general structure that can be perfectly instantiated to represent and model the composition of most of the QoS metrics. This feature provide to us all the expressiveness and the flexibility that a framework needs to be adapted to many application fields, while classic crisp constraint could show evident limitations.
Moreover, the negotiation of quality could be slightly over-constrained, since the requests of the client/consumer could be not fully satisfiable (by the service provider) at the present moment, because of other pending requests; on the other hand, the best possible solution, that gets closer to the consumer needs, can however be found, proposed and estimated by the potential client.
We are currently planning to enhance the QoS model, by including all its possible features and by finally translating the QoS satisfaction (and/or optimization) problem in a corresponding Soft Constraint Satisfaction Problem.
At last, we want to describe Shortest Path and Steiner Tree problems with Soft Constraint Logic Programming (i.e. in a declarative fashion), in order to model unicast and multicast QoS routing.