Davis Center, Room 3508
David R. Cheriton School of Computer Science
University of Waterloo
200 University Ave. West
Waterloo, ON N2L 3G1, Canada
E-mail: brecht AT cs DOT uwaterloo DOT ca
Phone: (519) 888-4567 x35892
In addition to working with people on current projects (some of which are described below), I often enjoy working with students or other researchers who have their own ideas for projects.
There are a large number and variety of services that stream video over the Internet. Some of these include: Amazon, Apple, CraveTV, HBO, Hulu, Microsoft, Netflix, Shomi, YouTube, and most major sports leagues and television networks. As a result of the popularity of these services, streaming video constitutes more than 50% of Internet traffic. Many of these services distribute data using content distribution networks (CDNs) like Akamai, Conviva or Netflix's own CDN. Our research is focused on understanding, designing, implementing, optimizing and provisioning the growing number of CDN and origin servers distributed throughout the world. The goal is to ensure that services are efficient and cost effective, and that subscribers are provided with the best quality of experience (QoE) possible.
We've been lucky to conduct some of this work with Netflix and to transfer some of our technology into their production servers. I am currently looking for students to continue this work and in particular to analyze significant amounts of data from their production servers.
To obtain streaming video data, many devices use a WiFi network to connect to the Internet via an Internet Service Provider (ISP). Recent work  has found that home WiFi networks are more often the bottleneck in delivering data than the ISP link. The increasing popularity of WiFi devices exacerbates the problem. A recent study  reports that the median number of access points seen from a home network in developed countries is 20. Obtaining peak throughput in such environments is difficult because radio spectrums are shared by WiFi devices (e.g., access points, computers, smart phones, tablets, TVs, and game consoles) and non-WiFi devices (e.g., wireless keyboards and mice and microwave ovens). It is therefore imperative that we understand how to optimize performance in situations that include dense deployments. However, because it is extremely difficult to collect information about or experimentally evaluate WiFi performance in these settings, they are often avoided by researchers. Our research is focused first on understanding and improving the performance of WiFi networks in such environments because they represent those in which many streaming videos are viewed. Secondly, we will examine techniques specifically designed to improve the quality of video streamed over WiFi networks.
We are working on the design and implementation of a cloud-based architecture that enables the collection, storage and analysis of data from IoT devices in a privacy preserving fashion. This work has significant applications in energy, healthcare, and surveillance sectors.
Last modified: Tue Feb 19 12:20:36 EST 2019