Internship proposal 2015/2016
Topic: Collaborative home network troubleshooting Duration: 4 to 6 months
Hosting team: Muse, Inria Paris-Rocquencourt (https://team.inria.fr/muse/) Contact: anna.bednarik@inria.fr
Mentor:
Renata Teixeira, Senior Researcher, Inria (head of Muse team) Anna-Kaisa Pietilainen, Researcher, Inria
Keywords: Internet measurements, home networks, network diagnosis Description:
Home networks are the most frequent interface of people’s “Internet experience”. When Internet service quality is poor, home users — who are rarely experts on network management — find themselves clueless. If the service is critical (for instance, TV during a sport’s game), users may end up calling their ISP hotline. Call center operators, however, know very little about the home network and cannot be of much use unless the problem lies within the ISP network. Indeed, our discussions with several large access ISPs reveal that a large fraction of customer calls are caused by problems that have nothing to do with the ISP, instead the problems are in many cases related to issues inside the home. The goal of the Muse team at Inria is to develop an easy-to-use home-network troubleshooting tool that can reliably identify performance and functionality shortcomings rooted in the home and propose ways to fix them.
This internship will focus on diagnosing the causes of high delays. The work will build upon our recent work on a network measurement platform for the browser (called Fathom) [1]. Fathom is a Firefox extension that provides a wide range of networking primitives directly to in-page JavaScript. Using Fathom APIs one can get direct TCP/UDP socket access; higher- level protocol APIs such as DNS, HTTP, and UPnP; and access system and network configuration information and network performance measurement tools such as ping (which measures round-trip delays). In addition, Fathom collects periodically baseline measurements of the network performance such as the amount of cross traffic on the local host; the perceived WiFi signal quality; network delays to the home gateway and the access network; CPU and memory utilization; the system load;
and page load times of few popular domains.
The student will develop a tool within Fathom that performs in five main steps. First, the tool will discover the devices in the home network and which devices will respond to Fathom’s request to perform measurements. This discovery will use standard mechanisms (such as mDNS and UPnP) that are already implemented in Fathom. Second, based on the set of available devices the tool will devise a measurement plan. For example, say that one end device and the home gateway are both Fathom-enabled, then we can ask the gateway to perform measurements to some services that are experiencing high delay and have the end-device perform delay measurements to the gateway and other devices in within the home. Third, the tool will trigger the measurement plan remotely. Fathom already implement remote calls and we have an OpenWRT package that will also trigger measurements from the gateway. Fourth, the tool will collect and interpret the results of measurements.
Finally, the tool will display the results to the user. All these steps can be implemented directly in JavaScript using the Fathom API. This work is supported by a research grant from Comcast (an ISP based in the United States) and we expect the student to be involved in collaboration with the Comcast engineers during the internship. In particular, we hope to have a test deployment in the homes of some Comcast customers.
The student should develop scientific skills on Internet measurements and troubleshooting as well as scientific writing and presentation. If the student is interested, there is a possibility of staying for the doctoral studies after the internship.
Desirable skills:
- Comfortable communicating in English
- Good knowledge of networking protocols and applications (TCP/IP, HTTP) - Knowledge of network performance measurements and troubleshooting - Knowledge of web technologies (HTML, CSS, Javascript)
References:
[1] Fathom, https://muse.inria.fr/fathom