Skip to main content
European Commission logo print header

SErvice QUality across Independently managed Networks

Risultati finali

SLA specification for QoS enabled networks aims at providing positive quality guarantees and setting out the limits of the services provided. Such SLAs move one step forward in the direction of traditional ones, in the sense that they do not only have to specify availability, security, quantity of allocated resources and a number of other quantitative values but also have to specify the values of appropriate quality parameters. In networks where QoS is inherently supported (such as ATM) the provision of SLAs comes as a natural delimitation of the relevant parameters. However, in IP networks where best-effort traffic has no quality guarantees, the introduction of QoS and associated services requires a thorough and accurate engineering of QoS metrics in the SLA specification on top of the guarantees for availability and characteristics of the transport medium, security, fault handling etc. The analytical computation of such metrics is extremely complex taking into consideration the extensive level of aggregation and more generally the nature of traffic flowing in large interconnection domains. Usually only upper bounds for the relevant parameters can be defined. Therefore, SLA specification for QoS enabled networks becomes a process where intensive testing and probing of the available infrastructure has to take place, before being able to quantify the QoS offering and include concrete parameters and values in the agreement. The Service Level Agreement specification for the Premium IP service, as a result of SEQUIN, is a corner stone in defining and implementing the service. It can be tailored to each instantiation between the involved domains for the provision of Premium IP and defines the framework for the provision of the service: the providers' commitments, the users' obligations, what happens when the rules are broken etc.
The requirements for and specification of a multi-domain monitoring infrastructure for QoS , which is able to monitor and verify SLAs established for Premium IP services is available. By installing the necessary monitoring infrastructure, Service Providers will be able to check the conformance of the network behaviour to the agreed SLAs and verify its compliance to the agreements with their customers. In the case of a network problem, monitoring nodes would immediately inform the Service Providers Network Operation Centres and gather enough information to locate the problem (in a network). Another important reason for installing monitoring infrastructure is that customers would accept being charged more for enhanced services than they are being charged today for a Best Efforts service, under the condition they can verify the provided QoS service from the network. Consequently, SLAs compliance verification is an important and significant challenge for Service Providers. This result provides a general framework for monitoring of SLAs for different Classes of Services (CoSs) in a multi-domain network.
Following from the definition of a Premium IP service able to operate in a multi-domain network environment, Implementation architecture has been specified. The implementation is based on the Expedited Forwarding Per Hop Behaviour (EFPHB) as specified by the IETF diff-serv Working Group. The implementation architecture is tailored for the GEANT and NREN networks that will initially implement the Premium IP service, but is extensible to other research networks and commercial ISPs. The requirements are to maintain simplicity and effectiveness. The main implication of this is that no dynamic signalling Protocol is used, instead static provisioning techniques are used. Given the number of users who will initially make use of Premium IP, this is expected to be manageable. As the number of users increases, and as technology matures, dynamic signalling may be added at a later stage (in 2 or 3 years). The Premium IP model requires that a number of functions be performed on the network nodes. These are admission control, classification, policing and scheduling. Each function must be performed only on certain nodes. Scheduling is performed on all nodes, based on the DSCP for Premium and is therefore applied to the aggregate Premium IP traffic. Policing is done close to the source of traffic and at network boundaries, in which case it is applied to the aggregate traffic from the domain sending the traffic. Admission control and classification is done only as close as possible to the source of the Premium IP traffic. The Premium IP service has been made available, via GEANT and NRENs, to three IST projects.

È in corso la ricerca di dati su OpenAIRE...

Si è verificato un errore durante la ricerca dei dati su OpenAIRE

Nessun risultato disponibile