The Internet ecosystema traffic measurement analysis

Resumen

The analysis of the interconnection status quo between content providers and the different networks of the Internet Service Providers (ISPs) that transport Internet traffic to end users is essential in better understanding the evolution of the Internet ecosystem. In recent years, we have witnessed a spectacular increase in the consumption of Internet traffic, especially multimedia content, which has driven both content providers and network operators, including transit, access ISPs and third party CDNs, to rethink their interconnection models. Internet video services demand particular network characteristics for which the original Internet was not initially developed. During the last twenty years of commercial Internet, the research community has contributed multitudes of hardware and software improvements. From a hardware perspective, user devices have augmented their processing power to unimaginable limits, while prices have been placed within the reach of the vast majority. Meanwhile, Internet networks have increased their capacity and provided the gift of ubiquity thanks to wireless technologies. From a software perspective, Internet applications have motivated the appearance of new hardware and network protocols because of their requirement to provide innovative user experiences. Moreover, virtualization and cloud technology have dramatically decreased the servers¿ costs in facilitating the proliferation of new services. Surprisingly, and despite the implementation of optimal and specific Internet protocols by the research community, the industry has opted for using existing robust and flexible protocols to transport their low-latency Internet content, HTTP over TCP. Certainly, current TCP and HTTP are not the same as the originals. However, they maintain their original essence and provide interoperability among millions of devices around the world. In addition, the protocols of the TCP/IP suite have a strategic advantage over others: they are an intrinsic part of the operating system, and Internet browsers are present in almost all devices. Therefore, technology has evolved and matured when necessary, and we currently observe a scenario where new applications and new hardware solutions (networks and devices) fit perfectly into a stable environment (HTTP over TCP). Now, therefore, we wonder if the evolution of Internet traffic has influenced the interconnection structure of the Internet. Internet players have also evolved the ways in which they interact with the rest of the ecosystem because of the increase in Internet demand, and this fact has impacted the Internet topology. Originally, the Internet was configured by a set of operators organized in a multi-tier hierarchic structure, where users and content were located at the bottom. When they needed to reach each other, it was necessary to go up and down the different tiers of the hierarchy. The network expansion of access ISPs and large content providers motivated by the growth of Internet traffic caused an evolution of the Internet landscape to a more meshed model bringing end users closer to content. In this context, CDNs emerged as a technical solution to deal with the massive demands of popular content and to provide reliable services for low latency requirements. Many global content providers deploy their own CDNs, while others prefer to contract third parties with the objective of being closer to the end users. Moreover, the introduction of paid peering as an interconnection agreement for ISPs that want to terminate the traffic exchange in an access ISP has revolutionized the market, as peering was founded on a settlement-free basis. After this point, transit ISPs began to claim because they were required to pay to deliver traffic in an end network when traditionally the economic transaction was in the other direction. This new situation, in which employing CDNs or establishing peering agreements redefined the content delivery models, forced transit ISPs to reinvent themselves. Transit ISPs needed to adapt by introducing new services to their catalogue and addressing new market segments because global content providers began to bypass them to reach the access ISPs. All of these changes raise the question of what interconnection differences exist between different ISPs with different roles in the Internet ecosystem. In this doctoral thesis, we perform an extensive traffic analysis from two perspectives to better understand the rationale behind the different Internet players. First, we analyse the Internet traffic from the perspective of the evolution of the Internet protocols. In analysing the protocols, we attempt to observe whether the traffic pattern has changed as new applications have emerged and the Internet demand have exploded. Second, we collect a data set of Internet traces that allows us to evaluate the connectivity between access ISPs and the most popular content providers. By analysing the Internet traces, we want to identify the differences and correlations in the interconnection models used by different Internet players. The main contributions of this work are the empirical corroboration that Internet traffic is mostly formed by HTTP over TCP, two robust and flexible protocols that have found their best ally in the Web, and that from the perspective of web content, the Internet market provides sufficient interconnection alternatives for the different profiles of Internet players. Accordingly, after twenty years of commercial Internet, we confirm that the content delivery market is sustainable and offers plenty of opportunities.