Demystifying TCP Initial Window Configurations of Content Distribution Networks

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 recommendations-despite 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 network-dependent 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.

[1]  Farnam Jahanian,et al.  Internet inter-domain traffic , 2010, SIGCOMM '10.

[2]  V. Jacobson,et al.  Congestion avoidance and control , 1988, CCRV.

[3]  Sally Floyd,et al.  Increasing TCP's Initial Window , 1998, RFC.

[4]  Yasushi Shinjo,et al.  VPN Gate: A Volunteer-Organized Public VPN Relay System with Blocking Resistance for Bypassing Government Censorship Firewalls , 2014, NSDI.

[5]  Sally Floyd,et al.  Measuring the evolution of transport protocols in the internet , 2005, CCRV.

[6]  Jan Rüth,et al.  How HTTP/2 pushes the web: An empirical study of HTTP/2 server push , 2017, 2017 IFIP Networking Conference (IFIP Networking) and Workshops.

[7]  Harkeerat Bedi,et al.  Riptide: Jump-Starting Back-Office Connections in Cloud Systems , 2016, 2016 IEEE 36th International Conference on Distributed Computing Systems (ICDCS).

[8]  Amit Agarwal,et al.  An argument for increasing TCP's initial congestion window , 2010, CCRV.

[9]  Feng Qian,et al.  TCP revisited: a fresh look at TCP in the wild , 2009, IMC '09.

[10]  Jan Rüth,et al.  Large-scale scanning of TCP's initial window , 2017, Internet Measurement Conference.

[11]  Michael Scharf,et al.  Performance Evaluation of Fast Startup Congestion Control Schemes , 2009, Networking.

[12]  Oliver Hohlfeld,et al.  A QoE Perspective on HTTP/2 Server Push , 2017, Internet-QoE@SIGCOMM.

[13]  Nick McKeown,et al.  A network in a laptop: rapid prototyping for software-defined networks , 2010, Hotnets-IX.

[14]  Mark Allman Removing TCP's Initial Congestion Window , 2015 .

[15]  Jan Rüth,et al.  A First Look at QUIC in the Wild , 2018, PAM.

[16]  W. Richard Stevens,et al.  TCP Slow Start, Congestion Avoidance, Fast Retransmit, and Fast Recovery Algorithms , 1997, RFC.

[17]  Vikram Visweswaraiah,et al.  Improving Restart of Idle TCP Connections , 1999 .