The web serving protocol stack is constantly changing and evolving to tackle technological shifts in networking infrastructure and website complexity. For example, development of QuiC to tackle security issues, HTTP/2 to tackle loss, and BBR to tackle high throughput. As a result of this evolution, the web serving stack includes a plethora of protocols and configuration parameters that enable the web serving stack to address a variety of realistic network conditions. Yet, today, most content providers have adopted a “one-size-fits-all” approach to configuring the networking stack of their user facing web servers, despite the diversity in end-user networks and devices.
In this paper, we demonstrate through empirical evidence that this “one-size-fits-all” approach results in sub-optimal performance and argue for a novel framework that extends existing CDN architectures to provide programmatic control over the configuration options of the CDN serving stack. To this end, we designed ConfigTron a data-driven framework that leverages data from all connections to identify their network characteristics and learn the optimal configuration parameters to improve end user performance.
ConfigTron uses multi-armed bandit-based learning algorithm to find optimal configurations in minimal time, enabling a CSP to systematically explore heterogeneous configurations while improving end-user page load time by as much as 16% (750ms) on median.