Demystifying TCP Initial Window Configurations of Content Distribution Networks

Abstract

Driven by their quest to improve web performance, Content Delivery Networks (CDNs) are known adaptors of performance optimizations. In this regard, TCP congestion control and particularly its initial congestion window (IW) size is one long-debated topic that can influence CDN performance. Its size is, however, assumed to be static by IETF recommendationsdespite being network-and application-dependent-and only infrequently changed in its history. To understand if the standardization and research perspective still meets Internet reality, we study the IW configurations of major CDNs. Our study uses a globally distributed infrastructure of VPNs giving access to residential access links that enable to shed light on networkdependent configurations. We observe that most CDNs are well aware of the IW's impact and find a high amount of customization that is beyond current Internet standards. Further, we find CDNs that utilize different IWs for different customers and content while others resort to fixed values. We find various initial window configurations, most below 50 segments yet with exceptions of up to 100 segments-the tenfold of current standards. Our study highlights that Internet reality drifted away from recommended and standardized practices.

An Empirical View on Content Provider Fairness

“…Studying actual Internet traffic is motivated by the observation that CC research often neglects the complex parameterization possibilities. For example in a previous study [10], we found that CDNs use different initial window configurations and some utilize pacing. To this end, we suspect that not only the initial windows might be different, thus choose two URLs for Akamai (named AkamaiA (then using IW32) and AkamaiE (then using IW16)) mapping to these different settings.…”

“…These increasing demands have led to a logical centralization of the content-serving Internet where a few big players serve the majority of the content [8], [9]. In previous work [10], we have shown that CDNs specialize in serving such content by tuning their TCP stacks beyond RFC-recommended values in hope for higher performance and user satisfaction. Fundamentally, such observations raise the question of fairness, and in fact, from an economic standpoint being unfair to a competing CP might be advantageous (e.g., by being able to deliver data with more than a fair bandwidth share).…”

Assessing the overhead of post-quantum cryptography in TLS 1.3 and SSH

“…The last -highly-debated-change in the recommended window size by IETF took place in 2013 [13] due to the increasing page sizes and Internet speeds [19]. Recent studies [53] show that initcwnd customization is not uncommon (up to 100MSS) for some Content Distribution Networks, while the majority uses the standard 10MSS value. A future increase in the recommended window value could aid the adoption of more PQ schemes in TLS and SSH.…”

Application layer classification of Internet traffic using ensemble learning models

“…However, the direction of transmission is now inverted, and outgoing byte windows would pertain to the transmitted data size limit set by recipient machine. In summary, the attributes of outgoing and incoming window of bytes dictate the data rate of the application involved 52 . It is visible that most labels have similar or relatively persistent range of values, with the exception of the Apple label.…”