The ever-growing popularity of smart devices greatly promotes content demand in cellular networks. It was projected by Cisco that mobile data traffic will grow sevenfold in the upcoming years. Such an enormous demand challenges the content hosting infrastructure that delivers content to massive numbers of users.
Content hosting infrastructure is largely shaped by cost, network policies, and the local regulations where they are deployed. In a recent study, we examined mobile content hosting infrastructure in China, which has the largest mobile Internet population in a single economy, and, perhaps more interestingly, has unique local regulations and network policies.
We analyzed 55 billion DNS replies collected from all recursive resolvers of a cellular ISP in China. All domains requested by mobile users through cellular networks were considered. Note that the DNS dataset contained no information of individual users, nor the IP addresses where the queries originated.
China’s cellular content infrastructure is mostly concentrated in a few ISP ASes rather than CDN ASes
Our analysis shows that China’s cellular content infrastructure is mostly concentrated in a few ISP ASes rather than CDN ASes. This is illustrated by the volumes of DNS queries that were mapped to individual ASes (that is, the destination ASes) in Table 1 (see first column), where the top 10,000 second-level domains were considered.
Two metrics — content potential and content monopoly — were further used to characterize the content hosting infrastructure (see the second and third column in Table 1). The content potential metric is the fraction of domains that can be potentially served by an AS, while the content monopoly metric is the extent to which an AS hosts content that others do not have. Both metrics lie between 0 and 1 — a larger value reflects higher content potential/content monopoly.
| Rank | AS name | % of queries that resolved to the AS | Content potential | Content Monopoly |
| 1 | ISP-AS1 | 40.99% | 0.59 | 0.18 |
| 2 | ISP-AS2 | 24.59% | 0.33 | 0.12 |
| 3 | Alibaba | 6.32% | 0.22 | 0.19 |
| 4 | Apple | 4.88% | 0.08 | 0.05 |
| 5 | Chinanet-BJ | 3.91% | 0.21 | 0.13 |
| 6 | ISP-AS3 | 2.19% | 0.20 | 0.09 |
| 7 | China169-back. | 1.38% | 0.28 | 0.11 |
| 8 | ISP-AS4 | 1.33% | 0.94 | 0.26 |
| 9 | ISP-AS5 | 1.05% | 0.19 | 0.10 |
| 10 | ISP-AS6 | 0.94% | 0.14 | 0.07 |
| 11 | Chinanet-back. | 0.81% | 0.31 | 0.13 |
| 12 | Akamai-ASN1 | 0.79% | 0.09 | 0.06 |
| 13 | Akamai-AS | 0.76% | 0.07 | 0.05 |
| 14 | Chinacache | 0.67% | 0.07 | 0.06 |
| 15 | CNIX | 0.56% | 0.06 | 0.09 |
| 16 | Chinanet-SN | 0.54% | 0.08 | 0.06 |
| 17 | China169-BJ | 0.54% | 0.14 | 0.09 |
| 18 | Yahoo-5G | 0.52% | 0.05 | 0.03 |
| 19 | Tencent | 0.50% | 0.12 | 0.11 |
| 20 | 0.40% | 0.07 | 0.05 |
Around 94% of the popular second level domains are available in ISP-AS4. The low content monopoly index further confirms the content of popular domains has been well replicated in multiple networks. These observations imply a significant locality of cellular traffic, and that cellular users can often get their content mostly within only one AS hop.
Alibaba cloud hosts over 20% of the popular domains
Another interesting observation from Table 1, is that the Alibaba cloud hosts over 20% of the popular domains. This is relevant to its successful websites hosting service. In fact, when considering all domains regardless of their popularity, the Alibaba cloud hosts the largest number of domains in a single network, showing that it hosts plenty of non-popular domains. This implies cloud providers have already taken the niche market of content hosting.
The analysis reveals that major providers slice up their infrastructures to host different kinds of services. As to the tracker hosting infrastructure, the analysis reveals that over 20% of the tracking queries are mapped to foreign ASes. This last observation raises privacy and cybersecurity concerns as the trackers may collect sensitive information.
For more information, please read our paper ‘Mobile Content Hosting Infrastructure in China: A View from a Cellular ISP’ [PDF 242 KB], which we presented at the recent Passive and Active Measurement conference, PAM 2018.
Zhenyu Li is a Professor at the Institute of Computing Technology, Chinese Academy of Sciences. His research interests include data-driven network analytics and networked systems.
The views expressed by the authors of this blog are their own and do not necessarily reflect the views of APNIC. Please note a Code of Conduct applies to this blog.