Comment by aspenmayer

When I found out that Booking[.]com of all companies is moving major traffic, I started to look at what companies are even buying or selling anymore. I clearly had no idea.

In the following paper, CPs refer to content providers, as defined in the paper.

https://estcarisimo.github.io/assets/pdf/papers/2019-comnets... [pdf]

(more at https://estcarisimo.github.io/publications/ )

canonical link for above paper, which is the lead researcher's GH from what I can tell:

https://www.sciencedirect.com/science/article/abs/pii/S01403... ( https://doi.org/10.1016/j.comcom.2019.05.022 )

> Studying the Evolution of Content Providers in IPv4 and IPv6 Internet Cores

> Esteban Carisimo, Carlos Selmo, J. Ignacio Alvarez-Hamelin, Amogh Dhamdhere

[I have edited out some hyphens that made this really hard to read but were helpful due to the layout of the original document as typeset. If that bothers you, I'm sorry in advance. Links are included above.]

> Our goal is to investigate what role CPs now play in the Internet ecosystem, and in particular, if CPs are now a part of the “core” of the Internet. Specifically, we motivate this work with the following questions: How can we identify if a CP does or does not belong to the core of the Internet? If the core of the network does indeed include CPs, who are they?As the overall adoption of IPv6 has been slow, do we notice that delay on IPv4 and IPv6 core evolution? As the AS ecosystem has shown striking differences according to geographical regions [15], do we also see geographical differences in the role of CPs and their presence in the “core” of regional Internet structures? Finally, as more CPs deploy their private CDNs, can we detect “up and coming” CDNs that are not currently in the core of the network but are likely to be in the future?

> We use the concept of k-cores to analyze the structure of the IPv4 AS-level internetwork over the last two decades. We first focus on seven large CPs, and confirm that they are all currently in the core of the Internet. We then dig deeper into the evolution of these large players to correlate observed topological characteristics with documented business practices which can explain when and why these networks entered the core. Next, we repeat the methodology but using IPv6 dataset to compare and contrast the evolution of CPs in both networks. Based on results, we investigate commercial and technical reasons why CPs started to roll out IPv6 connectivity.

> We then take a broader view, characterizing the set of ASes in the core of the IPv4 Internet in terms of business type and geography. Our analysis reveals that an increasing number of CPs are now in the core of the Internet. Finally, we demonstrate that the k-core analysis has the potential to reveal the rise of “up and coming” CPs. To encourage reproducibility of our results, we make our datasets available via an interactive query system at https://cnet.fi.uba.ar/TMA2018/

[…]

> Finally, we study the core evolution of nine other remarkable CPs that belong to the TOPcore but were not included in the Big Seven. Seven of the nine selected ASes are the remaining ASes in Bottger et al.’s [47] TOP15 list, except Hurricane Electric (AS6939) which we do not consider as a CP since it is labeled as Transit/Access in CAIDA’s AS classification [80]. These seven ASes are OVH (AS16276), LimeLight (AS22822), Microsoft (AS8075), Twitter (AS13414), Twitch (AS46489), CloudFlare (AS13335) and EdgeCast (AS15133). The other two ASes are Booking.com (AS43996) and Spotify (AS8403). Interestingly, Booking.com or Spotify are not normally considered among the top CPs, however, they are in both TOPcores.