The internet ecosystem is becoming increasingly complex, and is matched by the unprecedented growth of users - from 1 billion in 2005 to a current 3.5 billion. An EU-funded project has developed new techniques to assess the state and health of the internet so as to improve content delivery in this rapidly expanding and dynamic climate.
© pixelrain - fotolia.com
Researchers within the CDN-H project have focused on the performance and cost of content delivery networks (CDNs) in a hyperconnected world. These networks are responsible for half of all internet traffic on the internet. Moreover, because of its global reach, a CDN is a unique vantage point to track the evolution of the internet ecosystem and study peering strategies, address activity, and performance characteristics at an internet-wide scale, points out Georgios Smaragdakis, lead researcher with the project at the Technical University of Berlin.
Overhaul for the peering ecosystem
Internet peering is a reciprocal agreement whereby separate internet networks connect voluntarily to exchange traffic between network users and provide access to each others customers. With the emphasis on cost-performance trade-offs to deliver voluminous content traffic, the CDN-H researchers looked at the different interconnection possibilities at internet exchange points (IXPs) and other peering facilities.
CDN-H also looked at ways of harnessing new developments in virtualisation (the technology that powers cloud computing), creating virtual rather than an actual version of an operating system.
With scalable, secure content delivery within networks a focus, the team identified potential for new revenue streams and collaboration between internet stakeholders. The team has developed new techniques to map peering interconnections at the level of a single building. The developed techniques significantly expand existing maps of the internet and have practical value, for example, for locating the physical location of attacks, congestion, and vulnerabilities in the internet, explains Smaragdakis.
Capturing the pulse of internet activity
On a daily basis, a large CDN replies to trillions of user requests for content. The CDN-H researchers analysed user requests to one of the largest commercial CDNs from 2008 to 2016. They found that after years of linear growth in active IPv4 addresses, they stagnated after 2014, following exhaustion of the IPv4 address space.
This has significant implications for the operation of content delivery systems as more users share the same address and, thus, new algorithms have to be developed to better map user requests to the appropriate CDN servers, says Smaragdakis. For the first time, it was also possible to measure the population of the active IPv4 addresses with high accuracy and calculate the churn of IPv4 addresses across different regions and networks.
Alternative paths and protocols for more efficient content delivery
Looking at a large content delivery network, CDN-H, in collaboration with other labs, assessed performance characteristics before developing a fully decentralised, open-source analytics system, Datix. What makes the system innovative is its smart, partitioned data storage schemes that support fast algorithms and efficient processing to filter queries.
Other milestones have included the development of new techniques for internet traffic exchange without direct access to networks, and a breakdown of Web applications on the network to give details of bandwidth and user choice of application.
Conventionally, a user contacts a server for content delivery. However, to optimise efficiency, a server may contact another CDN server as well as servers hosted at different networks. Traffic between different networks is difficult to track but the CDN-H team used multiple internet vantage points to identify and investigate content delivery. The project found that a large part of todays internet is exchanged between servers, which has significant implications for internet architecture of the future, as well as efficient content delivery.
Smaragdakis sums up the significant implications of the project research, which was carried out in collaboration with researchers at the Massachusetts Institute of Technology and Akamai Technologies: As the internet continues to evolve and expand, the tools developed in the CDN-H project will be used to track the evolution of content delivery in the internet and provide useful insights for network operators, researchers, and regulators alike.