Problem Formulation

With the rise of the internet and its success, the distributed systems in heterogenous environments have become a norm. These distributed applications are usually built over distributed middleware that makes the task of the programmer quite easy. This middleware handles the complexity that orginates from the distributed and heterogenous nature on behalf of the programmer and makes the application easier to implement and allow them to inter-operate.

There are few applications which work sufficiently well in present networking and operating platforms. But, there are enough applications which need to specify their QoS requirements to the underlying network and recieve it. Best-effort networking platforms are not sufficient for them. Thus, applications that require a certain level of QoS must be able to specify their requirements in a clear and accurate manner by using QoS parameters. The values of these parameters reflect the requirements of the application. These parameters can be stored as a pre-defined user profile or can be obtained through direct interaction with the user.

The acheivement of desired QoS is done by resource reser-vation of the underlying networks and systems which include bandwidth, processing time, memory, hardware etc. Such mechanism to reserve resources at the network level is done by different resource reservation protocols like RSVP, Diff-Serv. But, the middleware needs to handle the complexity of mapping the application level specification and requirement to the underlying network QoS specification. These middlewares refered to as QoS.

Architectures are responsible for providing mechanisms for specification and enforcement of QoS that make use of the resource reservation protocols provided by the underlying system. QoS architectures deal with issues such as the translation of QoS parameters comprehensible at the application-level into the parameters understood by the
underlying reservation protocols that control access to the resources provided by the system. Without the services provided by a QoS architecture, these issues would have to be dealt with by the application which would make the programmers as well as users task complex and difficult

Research into QoS architectures ([CCH98], [Siq99] and asdiscussed in later sections) show folowing features :
_ Most architectures need a higher level abstraction at the application level. And this what they aim at.

_ Most architectures assume a particular underlying network and then build a middleware that provides QoS to the applications abov like Aquila (European IST funded project) assumes DiffServ network, QoS-A assumes ATMnetwork etc.

_ Usually, the mapping is done in a static manner tosome network services created/assumed at the underlyingnetwork.

_ Resource allocation modules co-ordinate at the middle-ware in a distributed enviroment using architectures likeCORBA etc.

Also, to most appropriately assign QoS resources network wide, the technology needs to have a mechanism which calculates and provides the required resources based on the network state and application requests.

Multi Protocol Label Switching (MPLS) is choice of future technology as it has capability to perform traffic engineering. Details of MPLS are explained in earlier section.

Thus, the objective of the thesis is to come up with an efficient mapping mechanism and QoS architecture for MPLS networks. The core issue in the architecture will be the mapping between SLA at the application level to the QoS parameters at the network level. We should be able to create a standard language to specify application QoS requirements and then be able to map them to the network parameters.

Our objective henceforth should be to be able to provide such a mapping function that does the following: Mapping QoS parameters (which are mentioned in SLAs) to NPMs (Network Performance Metrics) for different application specification as
these mapping function would be dependent on applications for MPLS networks. We might also, explore possiblities and mapping mechanism in DiffServ over MPLS networks.