AT&T developed a mobile video testing system that can measure the quality of streaming video content to mobile devices and ensure that content is matching expectation while not overloading network capacity, an increasingly important network management opportunity that can help network providers better manage services and costs.

The platform is the Automated Mobile Video Objective Testing System (AMVOTS), which measures video quality of experience (QoE) on mobile devices in “realistic network conditions.” This information can then be used by network operators can content application providers to manage bandwidth and network resources.

Eric Petajan, principal member of AT&T’s technical staff, explained that the system uses “prediction models for video applications, which will allow us some level of accuracy depending on the granularity of network traffic we have to predict the quality of experience for those sessions.”

These prediction models are based on actual network traffic, an adapter source on a device that can measure the actual screen experience, and an old-school way of asking people how they feel about their video experience.

The platform can access some network insight, mostly between the source material and the application streaming the content and what’s actually happening on the user’s device. What it does not currently have access to is what is happening within the application.

Petajan noted that the AMVOTS platform is targeted at adaptive bitrate applications “because that's an opportunity to actually optimize the performance, … deliver good enough quality to the user without overspending on the bit rate.”

The system does needs access to the highest source of that streaming data content to monitor against. This can act as a reference tool that can tell if a poor video stream could be due to a poor source.

The platform also has to take into account how a user sees the content. For instance, if someone is just passively watching video content, they might not need as high a resolution as someone intently watching something. That is also the case for people that might have lower-quality eyesight or are in difficult viewing environments.

Petajan noted that this is accomplished through sessions with customers at an Austin, Texas lab where AT&T collects mean opinion scores data to “help drive our prediction models.”

“We do a lot of subjective testing, because the ground truth for quality experience has to start with how people feel,” Petajan explained. “We do our own subjective testing, and we're also trying to use the standards from the ITU (International Telecommunications Union) that are also based on subjective testing. But subjective testing is expensive, it's hard to do, it's hard to get at how somebody actually feels after they're watching a three-minute video, there's a lot of variance in those results, so there's a whole subset of the science around getting that right as well.”

The platform is currently focused on mobile devices, as “that’s where the cost is and that’s where people’s attention is right now, staring at their phone,” Petajan said, though he did note AT&T recently worked with network partner Ericsson on some low-latency use case scenarios, including remote-render gaming (cloud gaming).

Need and future

These trials also provided insight into the platform’s ability to help operators optimize traffic depending on specific streaming content needs. For instance, streaming video content is not as latency sensitive at the packet level as a video conferencing service.

“Trying to optimize how the network is used, especially the mobile network is used between those two classes of visual applications is also the task at hand here,” Petajan said. “We don't want to over deliver latency from streaming and then take that latency budget away from the video conferencing system. How we do that is really a research topic, but the first step is to actually be able to collect the data, which is a big challenge that we cracked.”

While the platform’s mechanics sounds complex, that complexity is targeted at a specific need.

AT&T recently reported that its global network carried more than one exabyte of data across its network everyday by the end of last year. That’s a 1,000% increase compared to the 91 petabytes of data that was streaming across its network 10 year ago, and an 11,000% increase over the past 20 year.

Operators have a Gucci bag full of ways to deal with this capacity increase, though most of them are quite expensive.

AT&T, for instance, recently spent $23 billion in spectrum assets that is allowing the carrier to more aggressively target spectrum-dependent fixed-wireless access (FWA) services for the consumer and enterprise space. The carrier is also in the midst of a $14 billion mobile network overhaul program that includes switching network components that has allowed the carrier to unlock new service opportunities, with the broader move toward slashing the overall cost of network deployments.

Petajan explained that these resources are increasingly being stretched by growing video consumption, and that being able to minimize that impact has a significant financial benefit.

The next question for AMVOTS is can it be something that is scalable outside of AT&T, something Petajan said is “kind of an open question.”

“It's good to know, for the world to know, that it is possible to predict quality of experience from network traffic, and it's necessary for everybody's good to manage that traffic, whether it's encouraging the content application provider to be more efficient or other means, it's important to do that for the good,” Petajan said. “The mobile network is a limited resource. It’s very expensive to put up, and people rely on it. If you really need to make a call, you want to make a call, you don't want some other application to interfere with that. So we need to manage how the network is used and share it appropriately across users so everybody's happy.”