The Prudentia project aims to measure how Internet services like Youtube, Google Drive, and others, share network resources on the Internet. In our experiments, we run services at the same time in a controlled environment to see whether they share network bandwidth fairly or whether some services 'win' while others 'lose.'

Latest Results

Service Rankings

8 Mbps

Top-5 Most Contentious Services

Top-5 Least Contentious Services

50 Mbps

Top-5 Most Contentious Services

Top-5 Least Contentious Services

We rank services based on their contentiousness: how much other services slow down when they run over the network at the same time. Consider two different users browsing the Internet, one using a video site called FunTube and another teleconferencing using a service called Boom. If the teleconferencing quality for Boom drops a lot while FunTube is also running, we would say that FunTube is very contentious. We measure slowdown relative to max-min fair share. For most Internet services, we would expect two services to share the network bandwidth 50/50. If Boom achieves a '60%' max-min fair share, that means that Boom only achieved '60%' of its fair share while competing against Funtube. More information on our methodology can be found here. Below, we rank services based on which are the most and least contentious; for each service we list the average max-min fair share for all tested services that competed against it.

Pairwise Results

Click on any square in the grid to zoom in on pairwise results.

These grids show whether a service achieves more or less than their max-min fair share of network bandwidth. A given square means that the incumbent service in that column received x% of their fair share competing with the contending service listed in that row. A blue square means that the service in that column achieved more than their fair share and a red square means that the service achieved less than their fair share.

Note: All measurements in Prudentia were made using the browser clients for services. The TV or mobile apps of these services may exhibit different fairness outcomes for various reasons. For example, the Netflix TV client uses fewer concurrent flows than the browser client.