Optimal resource allocation for next generation network services
Michael Devetsikiotis, Michalis Kallitsis · 2010
Advances in various networking and computing technologies that allow high bandwidth, low latency connections have made transport services offered by telecommunication service providers to become a commodity. In order to gain a competitive advantage, service providers seek to enable value-added, next generation network services layered on top of the commodity transport service. At the same time, businesses across industries realize the need to be flexible and adapt to change so as to succeed in today’s information-driven economy. A robust, efficient, scalable, and dynamic communication and integration infrastructure is necessary to support this trend. Next generation network services include services that are offered via the emerging service oriented network architectures, such as e-banking and e-commerce transactions, security services and communication services with users’ presence and location information. In addition, the evolution of the “Web 3.0” paradigm provides services like online social networks, virtual collaborative environments and cloud computing services. Moreover, services like voice over IP, video on demand, online gaming and high speed Internet services are brought through triple and quadruple play architectures. Those services have diverse quality-of-service requirements and are constantly evolving in size while being geographically distributed across the world. Hence, network and system designers encounter the challenge of allocating the scarce and limited network and computing resources in an efficient and fair manner. In this dissertation, we attack the problem of optimal resource allocation in the aforementioned networks under quality-of-service and resource constraints. We propose optimization frameworks and algorithms that seek to maximize customers’ quality of experience of differentiated network services. Our framework is dynamic in the sense that it continuously monitors the state of the system (e.g., arriving traffic) and the optimization problem is solved only when required. We have employed deterministic as well as stochastic end-to-end performance bounds to assess the performance of each service (through network calculus and network decomposition tools, respectively). Furthermore, in the case of services that involve social collaboration, we propose a system that captures both technical and social interactions and allocates the resources accordingly.