DASH Streaming Test

DASH is designed to measure the quality of tested networks by emulating a video streaming. This test is called DASH because it uses the DASH (Dynamic Adaptive Streaming over HTTP) streaming technique.

Running this test can be useful to understand the baseline streaming performance of a specific network connection. It measures video-related metrics as well as network metrics that are key to understand the reason of performance issues.

When you run the test, it emulates the streaming of a thirty-second video from an M-Lab server. The video is divided in fifteen two seconds segments. When the client requests a video segment, it must also specify the video quality (e.g., SD, HD, Super HD). Of course, the higher the request quality, the bigger the returned segment. During the streaming, the client seeks to use the higher quality that does not load the network, creating queues, so that the streaming can continue smoothly.

We say the player is simple in that it does not employ algorithms that real players (e.g. YouTube, Netflix) implement to keep the streaming quality stable and to avoid stalls. This simplicity is, however, key to understand the contribution of the network to streaming quality, which otherwise could be masked by smart players’ behavior.

As a result, we expect real players to be generally faster than this test, because they implement more optimization techniques. However, if the throttling of video is caused by congestion at interconnection points, this test may result faster when the network path from the client to the M-Lab server does not pass through the congested interconnection point.

This network performance test was originally developed by the Neubot project and later integrated into measurement-kit, the engine used by ooniprobe-mobile.

Disclaimer: DASH is a general-purpose performance test conducted against third-party servers provided by Measurement Lab (M-Lab). M-Lab’s services require the retention and disclosure of IP addresses for research purposes. Learn more about M-Lab’s data governance, see its privacy statement.