Twitch’s CDN as an Open Population Ecosystem

Abstract

The quality and continuity of the video services such as Twitch depend on the scale and well-being of their content distribution networks (CDNs). Each CDN may consist of 1000s of servers, physically feeding the videos to the clients. Opting for a better understanding, researchers have attempted to measure and analyze the CDNs of popular video services [10, 11, 12, 19]. These works are, however, one-time effort. Given the widespread use of Twitch, we find continuous survey of its CDN an important subject of study. The challenge lies in the cost of performing the Internet-scale scans – the probing traffic. The larger the CDNs and the more frequent the scans are, the higher the overhead. Instead of performing full scans repeatedly, we envision a cost-effective alternative that samples and estimates the CDN size (i.e., the number of servers). Only when the size change is significant, does the system trigger a full scan. To this end and inspired by Capture-Mark-Recapture (CMR), a methodology widely used in Ecology to estimate animal population with little human effort, we propose two mechanisms to estimate the CDN size with lightweight traffic. Using a data set collected in Nov 2019, we find a 7.25% average estimation error. Provided an estimation error bound, we can identify as well the best parameter combination to minimize the probing traffic.