Strategies for provisioning end-to-end QoS-based services in a multi-domain scenario

Abstract

Providing end-to-end performance guarantees for QoS-based services, such as interactive voice and video, is a challenging task in the current Internet. In most cases, it involves the cooperation of multiple administrative domains, for the correct resource provisioning along the end-to-end path. Although some QoS standards have emerged in the last years, together with a trend for covering Internet services with performance guarantees through a SLA, endto-end QoS provisioning remains presently an open research problem. The main reason, advocated by this paper, is the current domain interconnection structure, which does not provide enough financial incentives for QoS deployment. In this paper, we tackle the interconnection problem by giving special attention to interdomain dynamic service negotiation, which is an important step that must be considered prior to the physical resource provisioning. Three strategies for ISP interaction (called negotiation models) are presented and evaluated through simulation and analysis.

Algorithms for SLA composition to provide inter-domain services

“…Beside the alliance framework, distributed models has been proposed by the authors of [1], [11]. Other authors opt for a centralized model [23], [4], [19], [7].…”

“…Authors of [14] predict the creation of alliances between domains: an alliance is a place of collaboration wherein confidentiality rules may be relaxed. The administration of 1 This work is partially funded by the ANR ACTRICE project. an alliance can be achieved through a third party which is responsible of the end-to-end contract: it monitors each contract and applies penalties.…”

A protocol for QoS contract negotiation and its implementation using Web Services

“…In [5,6,9], the authors study QoS specification issues, a prerequisite to cross domain negotiation, and they propose an end-toend distributed verification of the Service Level Specifications (SLS) with regard to domain service classes. Concerning the negotiation problem, [1] describes a centralized process to select QoS commitments satisfying a service QoS budget. In [12], the authors derive end-to-end global quality constraints from queuing models of composed services.…”

“…Here, we focus on negotiating global contracts from local QoS commitments, since the necessary information for building more sophisticated models are not available; moreover, negotiation cannot be centralized in our setting. Our distributed and nested approach extends the cascaded view advocated in [3,1]. In [3] authors focus on the QoS classes configuration problem so do not consider the cumulative effect of contracts.…”

“…In [3] authors focus on the QoS classes configuration problem so do not consider the cumulative effect of contracts. In [1] authors use traffic prediction to split the initial requirements into estimated QoS levels and then check if each participant can provide the estimated QoS level.…”