YouLighter: A Cognitive Approach to Unveil YouTube CDN and Changes

Abstract

YouTube relies on a massively distributed Content Delivery Network (CDN) to stream the billions of videos in its catalog. Unfortunately, very little information about the design of such CDN is available. This, combined with the pervasiveness of YouTube, poses a big challenge for Internet Service Providers (ISPs), which are compelled to optimize end-users' Quality of Experience (QoE) while having almost no visibility and understanding of CDN decisions.This paper presents YouLighter, an unsupervised technique that builds upon cognitive methodologies to identify changes in how the YouTube CDN serves traffic. YouLighter leverages only passive measurements and clustering algorithms to group caches that appear co-located and identical into edge-nodes. This automatically unveils the YouTube edge-nodes used by the ISP customers. Next, we leverage a new metric, called Pattern Dissimilarity, that compares the clustering results obtained from two different time snapshots to pinpoint sudden changes.By running YouLighter over 10-month long traces obtained from two ISPs in different countries, we pinpoint both sudden changes in edge-node allocation, and small alterations to the cache allocation policies which actually impair the QoE that the end-users perceive.

SeLINA: A Self-Learning Insightful Network Analyzer

“…The cluster set is also given as input to the model training phase of step (3), where a classification model is built by exploiting clusters as classification labels. The model is able to self-learn how to assign each network flow to the proper cluster.…”

“…We focused on two real-world use cases. The first one consists of a dataset of YouTube flows in which we know the CDN had changed over time, causing possible issues to both end-users and ISPs [3]. The second case deals with the understanding of P2P traffic, for which, instead, little knowledge is available.…”

“…Datasets from May 2nd to May 31st are used instead to run the online model update phase and highlight traffic changes possibly suggesting the automatic model rebuilding. For this dataset, we know that during the second part of May 2013 the YouTube CDN had relevant changes affecting end-user quality of experience [3], [1]. Hence, we consider this as ground-truth information that allows us to verify if SeLINA correctly identifies interesting events.…”

“…Big data and machine learning approaches have emerged to build systems that aim at automatically extracting information from the raw data that the monitoring infrastructures offer, and a significant effort has been devoted to apply them to network traffic analysis. Most of the proposed systems target a specific problem, e.g., monitoring of a CDN [1], [2], [3], detecting anomalies [4], [5], or simply offering scalable platforms [6].…”

“…SeLINA has been implemented in a state-of-the-art Big Data framework, Apache Spark, and has been applied to two real-world large use cases: a YouTube video streaming dataset and a peer-to-peer traffic dataset. Experimental results show that SeLINA is able to provide insightful network traffic models, e.g., pinpoint different groups of YouTube servers with different properties, and suggest the presence of changes in the infrastructure that have caused well-known issues to end- users [3].…”