A First Characterization of Anycast Traffic from Passive Traces

Danilo Giordano, Danilo Cicalese, Alessandro Finamore, Marco Mellia, Maurizio M. Munafò, Diana Zeaiter Joumblatt, Dario Rossi

Introduction

IP anycast is a routing mechanism where multiple physically distributed servers share the same IP address, with BGP routing directing client packets to the "topologically nearest" instance. Historically, anycast was restricted to stateless services like DNS due to fears that routing changes mid-connection would break TCP state. However, modern Anycast-enabled CDNs (A-CDNs) have challenged this convention. This paper provides the first large-scale passive characterization of IP anycast traffic, using traces from a major European ISP to reveal how anycast now supports a wide array of stateful web services.

Methodology

The researchers employed a two-step approach to identify and analyze anycast traffic:

  1. Active Census: Identification of ~1,600 anycast /24 subnets through global probing, narrowing to a high-confidence "conservative" set of 897 subnets.
  2. Passive Monitoring: Observation of actual traffic at a European ISP Point of Presence using Tstat for flow analysis and DN-Hunter for mapping IP addresses to Fully Qualified Domain Names (FQDNs).

The dataset, collected in March 2015, encompasses traffic from 20,000 households, totaling 270 TB of data and 2 billion TCP flows. Within this, the authors identified 59 million anycast-related TCP connections.

User Penetration and Traffic Volume

Anycast has achieved significant mainstream adoption. It accounts for approximately 3% of all web connections and 4% of total HTTP/HTTPS volume. More strikingly, the daily penetration is massive: over 50% of active users interact with at least one anycast-enabled service every day, showing a consistent diurnal pattern.

Daily reach of anycast services among ISP clients.

Service Diversity and Performance

The study dispels the myth that anycast is only for DNS. A-CDNs now host diverse services including social media (Twitter), search (Bing), and streaming (Soundcloud, Vine). Performance is highly optimized; major players like Edgecast and Microsoft maintain median Round Trip Times (RTT) under 10-20ms, indicating server placement very close to the ISP's borders.

Multidimensional characterization of major anycast providers across metrics like RTT, volume, and protocol count.

Path Stability and Insights

Despite theoretical concerns about BGP "flapping" breaking connections, the authors observed remarkable path stability. Most connections are short enough that routing changes do not occur, and even long-lived streaming flows remained stable over the month-long observation. Rare RTT jumps were observed, indicating route shifts, but these occasionally improved throughput by moving traffic to less congested, albeit longer, paths.

Conclusion

Anycast is now a mature, high-performance pillar of the modern web. While this study is limited by its single-ISP vantage point, it demonstrates that anycast effectively supports stateful, high-volume traffic with excellent stability and low latency. Future research will likely focus on deeper comparisons between anycast and unicast CDN performance architectures.

Summary updated

Abstract

Abstract—IP anycast routes packets to the topologically nearest server according to BGP proximity. In the last years, new players have started adopting this technology to serve web content via Anycast-enabled CDNs (A-CDN). To the best of our knowledge, in the literature, there are studies that focus on a specific A-CDN deployment, but little is known about the users and the services that A-CDNs are serving in the Internet at large. This prompted us to perform a passive characterization study, bringing out the principal A-CDN actors in our monitored setup, the services they offer, their penetration, etc. Results show a very heterogeneous picture, with A-CDN empowered services that are very popular (e.g., Twitter or Bing), serve a lot of different contents (e.g., Wordpress or adult content), and even include audio/video streaming (e.g., Soundcloud, or Vine). Our measurements show that the A-CDN technology is quite mature and popular, with more than 50% of web users that access content served by a A-CDN during peak time
BibTeX
@inproceedings{giordano2016first,
  author = {Danilo Giordano and Danilo Cicalese and Alessandro Finamore and Marco Mellia and Maurizio M. Munafò and Diana Zeaiter Joumblatt and Dario Rossi},
  title = {{A First Characterization of Anycast Traffic from Passive Traces}},
  booktitle = {TMA},
  year = {2016},
  url = {https://openreview.net/forum?id=e8yNL8W2rJ},
}
Figure