Dynamic core provisioning for quantitative differentiated services
IEEE/ACM Transactions on Networking (TON)
| Hi-index | 0.00 |
Over the past decade network bandwidth has become a commodity item putting pressure on Internet Service Providers (ISPs) to differentiate their service offerings to customers in order to maintain market share. However, realizing service differentiation in IP networks is a broad, multi-dimensional and challenging problem. This thesis addresses this problem and proposes new approaches for bandwidth service management for the Internet and its wireless extensions. First, we propose a unified formulation of band width utility functions for application aggregates including TCP, small audio flows, and individual video flows. We discuss experiments using the online generation of utility functions from video traces and present a utility prediction algorithm that addresses the time scale mismatch that exits between video content changes and network adaptation time-scales. Next, we present two groups of utility-based link allocation algorithms that provide a foundation for utility differentiating and utility maximizing bandwidth management. The utility maximizing algorithm leverages the piecewise linear quantization of utility functions and uses the Kuhn-Tucker condition to significantly reduce the algorithm execution time. Our utility differentiating algorithm supports utility fair allocation that allows individual utility functions to have different maximum utility values. We extend these results to the problem of multi-hop utility-based flow control by augmenting the max-min flow control algorithm to support utility functions. We study, propose and evaluate a utility-based max-min fair allocation and renegotiation protocol in the context of an edge-based wireless access network taking into consideration convergence speed, protocol state reduction, and the management of application adaptation states. Third, we present a dynamic bandwidth provisioning model for quantitative service differentiation in core networks that comprises node and core provisioning algorithms. The node provisioning algorithm prevents transient violations of Service Level Agreements (SLAs) by predicting the onset of service level violations based on a multi-class virtual queue technique, self-adjusting per-class service weights, and packet dropping thresholds at core routers. Persistent service level violations are reported to a dynamic core provisioning algorithm, which dimensions traffic aggregates at the network ingress taking into account fairness issues not only across different traffic aggregates but also within the same aggregate whose packets can take different routes in the core IP network. We solve the problem of rate regulation for point-to-multipoint flow aggregates with the use of matrix inverse operations. We demonstrate that our model is capable of delivering capacity provisioning in an efficient manner and providing quantitative delay-bounds with differentiated loss across per-aggregate service classes. Finally, we propose incentive engineering techniques and design two incentive-based allocation service classes that effectively constrain the strategy space of subscribers to a set of cooperative behaviors that include the truthful selection of a service class and truthful declaration of bandwidth demands. Our design minimizes protocol messaging overhead imposed on wireless subscribers while possessing a number of beneficial properties including Nash bargaining fairness for the instantaneous allocation service, and incentive compatibility for mobile users promoting the truthful declaration of their service preferences. (Abstract shortened by UMI.)