Understanding and Improving Content Distribution Through Expansive Network Measurements

Warrior, Marc Anthony

doi:https://doi.org/10.21985/n2-rrv4-ye26

Work

Understanding and Improving Content Distribution Through Expansive Network Measurements

Public

Download PDF

Download All Files (.zip)

In response to exponentially increasing demand for digital media, today's Internet landscape has evolved into a multitude of diverse and interdependent distribution systems designed to move content as efficiently as possible. While many of these systems have \emph{individually} been explored in depth by both academic and industrial communities, a cross-sectional investigation of the \emph{relationships} between competing or coexisting content distribution systems and resources is generally absent from the current narrative. Further, when such expansive studies are given consideration, they are avoided due to the daunting challenges they present. Scope and vantage point concerns become non-trivial when designing experiments that span multiple network resources, and third-party systems may lack transparency for the curious researcher. In this thesis, I assert that expansive network measurements such as these are not only feasible, but \emph{essential} to our efforts to understand and improve modern content distribution systems. I demonstrate that anchoring cross-sectional measurements in client-side machines provides the real-world perspectives necessary for optimizing actual client experience. Rather than examine the performance of a single resource-client pair, I instead obtain, for each client considered, \emph{relative} measurements across the set of systems and resources visible to the client or its peers. Each additional considered Internet resource or system provides relative context that highlights otherwise unobservable outlier properties. With this approach, I achieve the following: First, I discover and resolve sub-optimal resource-client mappings using only a lightweight, client-side implementation. Next, I quantify the extent to which clients are exposed to the same network resources as each other, and I further leverage these results to systematically identify opportunities to improve client performance. Finally, I enable scalable assessment of a crowdsourced ecosystem's content aggregation and distribution patterns.

Creator