The application level services which are specified by the user through a browser specifying a application level profile in XML created by Aquila group. This partial information is then mapped to servicecomponent profile (again in XML) which are detailed requirements of the resources that we need to reservce from the network.
The mapping is static meaning the classfication of network services as is done above in above 5 categories is sufficient to meet almost all application level requirements. And for any given application requirement, we map it to one of these services that most closely meet its requirement.
Before commiting a resource reservation request, some admission control algorithm is used that ensures if sufficient resources are available to meet the request.
Above, method assumes DiffServ network. They have beutifully and effectively solved the problem of mapping SLA which may be specified using QML, HQML or using Aquila’s applicationprofile DTDs and service component DTDs into SLS (technical aspects of SLA requirements of applications) and then into 5 network services that most closely meet its requirements.
The 5 network services as done in Aquila architecture are:
1) Premium Constant Bit rate (PCBR).
2) Premium Variable Bit rate (PVBR).
3) Premium Multimedia (PMM)
4) Premium Mission Critical (PMC)
5) Standard Best Effort (STD).
The 5 network services are provided by creating 5 traffic classes (TCL) as follows:
1) First TCL has a queue and has the highest priority.
2) Rest 3 are respectively in smaller priorities and controlled using weighted fair schedulling.
3) Last one is of the smallest priority and is best-effort service. Yet, some resources are reserved for it too.
4) Different network parameters such as bandwidth, peak rate, average rate are controlled using leaky token bucket (and sometimes, dual leaky token bucket to control average traffic rate for a period of time, also).
The above details of traffic classes are then used to allocate resources in DiffServ network. The end to end state is ensured by maintaining the logical resource pool in the layer above network that keeps track of the resources. This logical layer is known as RCL (resource control layer). Thus, to implement above mapping schema, to MPLS networks, we would need to map network services to the MPLS TE. Thus, bringing changes to the RCL (resource control layer). Aquila Architecture has already been explained in detail in the earlier section of related works.
Extending Aquila : SLA to QoS mapping in MPLS network
In order to be able to provide proper SLA to QoS map-ping in MPLS networks, we need to create network services that could be mapped to the service component profiles of the applications. These network services can be crested by considering following issues which would make it a traffic engineering problem:
1) We need to be able to provide 5 network services as isdone for above DiffServ network.
2) We need to create a servicecomponent network levelprofile that could be mapped to the application profile which are the application/user specified requirement.
3) In MPLS networks, we need to map LSPs and labels and see how they can be used to provide 5 network services as mentioned above.
4) We can also explore possiblities for being able to provide given network services in DiffServ over MPLS networks.