direkt zum Inhalt springen

direkt zum Hauptnavigationsmenü

Sie sind hier

TU Berlin

Inhalt des Dokuments

Es gibt keine deutsche Übersetzung dieser Webseite.

Traffic and Internet Characterization

DNS in the Wild

DNS is a fundamental building block of the Internet. During the evolution of DNS new uses have emerged. E.g., content distribution networks (CDNs) direct client requests to content host through DNS. Moreover, DNS resolvers run by third parties, e.g, OpenDNS or Google Public DNS, made their appearance. This motivates us to look into their performance compared to the local DNS resolver.

We collected DNS traces for about 10.000 hostnames from 60 globally distributed vantage points within 50 commercial ISPs. We identify typical patterns of DNS deployment indicating the relative distance of resolvers towards the client, as well as the use of load balancing across multiple DNS resolvers.

While the local DNS server is not necessarily better than a 3rd party resolver, the use of the latter deteriorates the quality of the content server selection.

Researchers: Bernhard Ager, Wolfgang Mühlbauer, Georgios Smaragdakis, Steve Uhlig

Locality Aspects in Content Delivery

The immense growth of Internet traffic on the one side and the constant user demand for better Internet service leads to a challenge for providers to find a solution coping with both problems. One solution that has worked well in the past is caching. However, caching has become unpopular due to the increasing dynamicity of content. Due to the changing shape of Internet content we see the need to re-investigate content cacheability.

Our study is based on two weeks of traffic of the four most prominent protocols of 20.000 residential customers. On the one hand our results indicate very high potential cacheability for P2P protocols, which however, are responsible for only a small fraction of the overall traffic volume. On the other hand, HTTP cacheability is limited while being responsible for the majority of the traffic. The most important limiting factors are cache control, URL parameters and load balancing. A possible mitigation strategy is the use of opportunistic heuristics.

Researchers: Bernhard Ager, Juhoon Kim, Fabian Schneider

Selected Publications

This section is still under contruction. Please see full list of publications or search publication.

Zusatzinformationen / Extras


Schnellnavigation zur Seite über Nummerneingabe

Under Construction

This page/section is
still under construc-
tion. Please try again