CRII: NeTS: Characterizing, Quantifying and Modeling Network Complexity

点击翻译按钮获取中文摘要

Networks with higher degrees of complexity generally require more manual intervention to manage, are more difficult to reason about, and are more prone to human errors. Despite its critical importance, complexity remains one of the least understood aspects of a network, and has not been well characterized or modeled. There is a general perception that for the same target network multiple designs often exist to meet the same set of operational objectives (e.g., security and resiliency requirements), and that some designs could be significantly easier to manage than others; however the decisions on complexity are typically made subjectively or via qualitative terms. This project aims at developing a fundamentally new understanding of, and metrics and models for, network complexity. The project studies how a network's design characteristics impact the complexity in its management, through a novel white-box approach that involves reverse-engineering the network design from device configurations, identifying design patterns that differ from conventional wisdom, and iterative discussions with network operators to capture perceived complexity. A suite of bottom-up metrics will be developed to accurately quantify the complexity of a given design. These metrics enable comparisons of alternative design proposals and "what-if" analyses of how design changes might impact the management complexity. A top-down framework will be developed to model the interplay among network design, operational objectives, and the resulting management complexity. The framework formally abstracts the objectives of a given design task thus allowing reasoning about whether and how a combination of design primitives will meet those objectives, decomposes the design task into its constituent primitives, and applies the metrics to quantify the complexity of individual primitives.If successful this project will advance the state-of-the-art in understanding, quantifying, and modeling network complexity. The modeling framework enables complexity-aware top-down network design by guiding the search through the design space. The insights obtained from this project may inform the design of future network architectures, management applications, and configuration languages. The project's PI will actively interact with the research community, operator community, and Internet Engineering Task Force to disseminate the research findings. Results from the project will be incorporated into graduate-level network management classes. As Florida International University is a designated Hispanic-serving institution, this project will serve as a vehicle to advance the involvement of underrepresented minorities in computer science research.