Advance reservation in distributed computer networks.

摘要

The advance reservation of network resources and connections is an area of growing interest, particularly when large volumes of data are involved. Most notably, the scientific research community is presenting some very pressing challenges in this regard, both in terms of bandwidth scalability and service flexibility. In particular, a wide range of collaborative research activities are being carried out in key areas such as high energy physics, genetic modeling, climate change, astronomy, etc. In many cases here, related "e-science" data volumes are readily exceeding the capacity of currently-deployed networks with immediate reservation capabilities. Moreover, owing to the increasingly global nature of modern research collaborations, much of these data sets need to be transferred and processed at widely-dispersed laboratory facilities and institutes across the world. Indeed, this mandates meticulous scheduling of available network resources.;In light of the above, this dissertation presents a comprehensive set of solutions for service provisioning in advance reservation bandwidth networks. Specifically, a novel linear programming model is first proposed to address the computational complexity and limitation of existing optimization schemes. Next, a heuristic rerouting algorithm is developed to further reduce run-time complexity in "on-line" operational settings and improve overall service provisioning performance. Furthermore, to address real-world scalability and reliability challenges, a comprehensive distributed advance reservation framework is also presented for single-domain and broader multi-domain routing networks. Overall, detailed performance evaluation results are presented for each of these proposed solutions, and wherever possible, the findings are compared with those from other existing schemes.