Mu's Quadrant of Miscreant Apps

12:00 PM -- There's not much debate around the fact that video is the culprit for driving data and encumbering wireless networks. But not all video is created equal.

As part of Mu Dynamics 's new application-testing platform, TestCloud, the vendor has access to data on individual apps and the effect they're having on the network. Its initial report suggests that popular scapegoat YouTube Inc. is actually the most friendly video app to both consumers and wireless operators in standard or high definition. (See Mu Helps Operators Shape App-Aware Networks.)

Mu defines consumer friendliness as consuming the least amount of bandwidth, which will be especially important as usage-based pricing plans become more common. Operator friendliness, or the amount of network resources consumed, is measured by the number of connections established by an app within a given session.

Hulu LLC , which streams full-length TV shows and movies, was the least operator friendly in SD and HD, while Netflix Inc. (Nasdaq: NFLX) was least friendly to consumers in both resolutions. Amazon's video content was found to be more friendly to operators in HD than it was in SD.

Developers of apps like these should play a role in building services that are network aware and don't hog resources, but most of the responsibility for mitigating the effects of the traffic falls to the wireless operators. Knowing how certain apps affect them differently will help operators determine what policies to set on apps, such as giving consumers incentives to watch full episodes in off-peak times or on carrier-managed Wi-Fi.

Video traffic is going to continue to increase on Long Term Evolution (LTE), but being aware of what's causing it at least puts the operators a step ahead of where they were when 3G apps caught them by surprise.

— Sarah Reedy, Senior Reporter, Light Reading Mobile

sarahthomas1011 12/5/2012 | 5:08:03 PM
re: Mu's Quadrant of Miscreant Apps

I'm not surprised that YouTube was the most operator and consumer "friendly," if not just because most videos viewed on YouTube are shorter than on Netflix and Hulu. Future Mu reports will be more interesting, because it'll start comparing devices and operating systems too.

The SD versus HD differences are interesting too.

tmaufer 12/5/2012 | 5:07:47 PM
re: Mu's Quadrant of Miscreant Apps

Actually, our methodology enabled us to take samples of the traffic from each video source. After some experimentation we discovered that after sampling the streaming video for 5+ minutes, we reached a point of diminishing returns. Our algorithm involves comparing rates not absolute numbers, so the size of a particular capture is not factored in to the comparison.

It was necessary for us to define a content-independent comparison algorithm because it's difficult to find the same movie on more than one video streaming service. Not only is our algorithm content-independent, it's time-invariant. This means that each traffic capture needn't be the same length or the same number of packets.

sarahthomas1011 12/5/2012 | 5:07:46 PM
re: Mu's Quadrant of Miscreant Apps

Thanks for the response. That's interesting - would seem to suggest that there is more developers can be doing, in addition to operators, to make their apps more network "friendly." What is YouTube doing so much better than Hulu?

tmaufer 12/5/2012 | 5:07:41 PM
re: Mu's Quadrant of Miscreant Apps

It's unclear "why" the apps are different...but they are measurably different!

NetFlix opens connections to 36 different hosts during the first minute of a video...app servers, CDN hosts, ad servers, back-end support servers, etc. (we infer their roles based on their names).

On the other hand, YouTube opens very few connections and creates very few transactions over which they send the whole video stream. Their fundamental design is simply different and there was never a feedback loop to show them that their app design choices affect the network, and possibly users (i.e., certain CODECs may be more bandwidth-intensive).

We're trying to provide concrete metrics to show which apps are relatively more costly on the network so people (consumers, operators *and* developers) can make informed choices.

Sign In