Deployment of ~100 probes as of Jan 2017. Each probe is connected to a dual-stacked network. Most of these probes are deployed in residential settings.

This topic is taken and not available anymore.

Motivation

ISP networks (such as Deutsche Telekom) deploy nodes inside their network to cache popular content from content provider (such as Google and Facebook) networks. These cache nodes help reduce response times to improve overall experience of the ISP' subscribers. With the rapid deployment of IPv6, we want to know how widely dual-stacked is the caching infrastructure inside ISP networks today. A caching infrastructure that is not dual-stacked today can increase response times when services are delivered over IPv6 thereby hampering the user-experience of the ISP' subscribers.

Approach

In order to pursue this research question, we leverage a deployment of ~100 measurement devices. These devices (similar to a TP Link router) are connected to dual-stacked networks. We have been running traceroute and dns measurements from these devices for the past 3 years. As of Jan 2017, the collected dns dataset is ~300G in size while the traceroute dataset is ~900G in size.

Goal

The goal of this thesis would be to analyse these datasets to determine how often do users hit a dual-stacked content cache when accessing popular websites. Moreover, in situations, where the cache is not available over IPv6, how inflated are the paths to fetch the service over IPv6, how do the IPv6 paths compare to that of IPv4 and what is the penalty in terms of latency that a user has to experience when accessing content over such inflated paths.

Requirements

  • A basic understanding of network protocols and basic statistics.
  • Knowledge of Python and SQL.

Contact

Vaibhav Bajpai, bajpaiv at in.tum.de