Robust Cross-Layer Network Optimization for Diverse QoS Requirements: Work in Progress

V Marbukh
Conventional approach to cross-layer network optimization assumes elastic users adjusting their bandwidth requirements in response to the resource congestion prices. This assumption leads to Network Utility Maximization (NUM) framework with Lagrange multipliers associated with resource capacity constraints playing role of the congestion prices. However, often users can more naturally quantify their preferences in terms of the rate and high-level Quality of Service (QoS) requirements rather than
more » ... rements rather than networks level end-to-end bandwidth requirements. This paper suggests that replacing resource capacity constraints with constraints on the feasible QoS parameters may lead to cross-layer network optimization framework with elastic users adjusting their diverse QoS requirements directly in response to the QoS-sensitive prices. We illustrate the proposed framework on examples of end-to-end bandwidth allocation subject to the worst-case scenario and average end-to-end delay. Index Terms-cross-layer system optimization, quality of service, pricing.