The MESCAL Project

 
Management of End-to-end Quality of Service Across the Internet at Large

Keywords: Internet, Inter-domain, Quality of Service, Traffic Engineering, Service Level Specification

Home

About MESCAL:
Abstract
Rationale
Objectives
Organisation
Partners
Contact
Results:
Results roadmap
Glossary
Deliverables
Publications
Restricted docs
Standardisation
Presentations

Other:
EEQoS workshop
Links
CFPs

Project only

Inter-domain service management

The MESCAL solution to the problem of inter-domain QoS delivery relies on interactions at the service layer between adjacent providers for establishing agreements for QoS traffic exchange, pSLSs in MESCAL terminology, to allow providers to expand the reach of their offered QoS-based services beyond the boundaries of their domains. Following the establishment of pSLS, q-BGP message exchanges occur at the IP layer for finding, establishing and maintaining suitable QoS routes.

pSLS Management is a layer of the overall MESCAL architecture, per provider domain, responsible for handling the establishment of pSLSes (from the perspectives of a single provider domain) and mediating the information that needs to be exchanged with the TE functions –mapping service-related to network-related information, intra- and inter-domain, and vice versa. In particular, it includes the following aspects, which are regarded as results on their own right.

Service management information model: An information model representing all notions and entities pertinent to QoS-based service, pSLS management functionality, also capturing the outcome of all interactions between the service management layer and the business and the TE layers has been specified and implemented in a data-base schema. The model represents entities such as l-QCs, e-QCs and o-QCs, peer providers and pSLS contents, QC-mappings, bindings and implementation, inter-domain and intra-domain resources (from service perspectives), traffic demand matrices and availability estimates. Based on this model, all pSLS-related functions operate, proving the validity of the MESCAL specifications.

Modelling of pSLSes and QoS-based connectivity services: Suitable templates -set of parameters with clear semantics- for describing the contents of c/pSLSes have been specified. MESCAL firmly believes that there is a need for standardising c/pSLS for the benefit of Internet service deployment and provisioning; standardised c/pSLSes would provide a common informational basis for the interactions between end-customers and providers and between providers, as well as for building the required service provisioning functionality; thus, enabling the automation of the respective processes. A two-level modelling approach was adopted. First, a general, open, detailed service model for describing pSLS as well as any QoS-based service was specified. Subsequently, different types of pSLSes, as appropriate to the various types of business relationships holding between the providers and the type of QoS to exchange, were specified by appropriately restricting and/or summarising the information identified in this open service model. The modelling work has been built on the TEQUILA SLS template.

pSLS validation, mapping and translation: On their establishment request, pSLSes are validated, mapped and translated to appropriate network-related information, inter- and inter-domain, as required for the configuration and the operation of the lower level service management and TE functions (including q-BGP), which are responsible for ensuring that pSLS traffic can access the network and be delivered according to their agreed requirements. In essence, this work realises the process of provisioning QoS-based connectivity services, proving the validity of pSLS specifications and the feasibility of the general MESCAL solution –from pSLSes to q-BGP configuration.

Demand derivation and aggregation: For calculating the anticipated traffic demand per QoS class per destination in the Internet, given a set of pSLSes that a provider domain has established on request from its peers. The granularity of the traffic demands per QoS class depends on the granularity of the reachable destinations in the pSLSes offered by a domain, which for reasons of scalability and feasibility should be in a coarse/aggregate form. For a given set of Internet destinations common to a set of pSLSes, demand derivation and aggregation on behalf of these pSLSes is a linear function, based on determined multiplexing factors and aggregation weights per type of traffic. The calculated demands, accordingly extrapolated, are fed to the off-line TE functions for dimensioning the network in terms of intra- and inter-domain resources.

pSLS admission logic: For determining whether to admit a requested pSLS or not, perhaps counter-proposing alternative pSLSes, given the current status of the engineered network to gracefully sustain new traffic –further subscriptions on QoS traffic. The admission logic relies on estimates of the availability of the network as engineered by the TE function to accommodate QoS traffic within and out from the domain. By formulating an appropriate linear optimisation problem it determines whether a feasible solution can exist in accommodating the demand of the newly requested pSLS as well as the demand of each of the already established pSLSes within the available intra- and inter-domain resources determined by TE. Clearly, this function, exerted at pSLS establishment request epochs, constitutes the first level of control to apply for settling the trade-off between network utilisation and service agreements at desired levels. The value of this function should be appreciated taking into account inter-domain aspects –inter-domain link capacity may not be available as great as in the core network- also, by considering that pSLSes are agreements, which if not honoured, negative impact and consequences will be created.

Evidently, in addition that they were required by the MESCAL inter-domain QoS delivery solution, the above results automate the process of service provisioning; highly desired by providers. Their validity and scalability of the above aspects has been verified through experimentation.

Further reading:

M.P. Howarth, P. Flegkas, G. Pavlou, N. Wang, P. Trimintzios, D. Griffin, J. Griem, M. Boucadair, P. Morand, H. Asgari and P. Georgatsos, "Provisioning for Inter-domain quality of service: the MESCAL approach," IEEE Communications Magazine, June 2005. [pdf document]

MESCAL deliverable D1.3, "Final specification of protocols and algorithms for inter-domain SLS management and traffic engineering for QoS-based IP service delivery", Chapter 9. [link]

MESCAL deliverable D3.2, "Final experimental results: validation and performance assessment of algorithms and protocols for inter-domain QoS through service-driven traffic engineering", Chapter 5. [link]

[back to MESCAL results roadmap]


MESCAL home -- IST Programme -- Contact MESCAL
Page updated by David Griffin September 2005