Thursday, August 24, 2017

Is throttling evidence that U.S. mobile networks are capacity constrained?





by James E. Carroll

Verizon is taking heat in the media this week for changes to its “unlimited” data plans. The new throttling policy that should sharply cut down the volume traversing its network. Customers of Verizon’s lowest-cost unlimited plan will find the video streaming is now limited to 480p resolution (approximately 1.5 Mbps) and that data tethering is limited to 600 Kbps. Higher priced unlimited plans support streaming at 720p on smartphones and 1080p on tablets. A business-class tier provides a higher standard of service. AT&T, T-Mobile and Sprint also have throttling policies in effect.

We know streaming video is data intensive. We also have seen mobile operators boasting of how fast their 4G LTE access speeds have become. Now it looks like the carriers want to restrict the performance experienced by the majority of their subscribers. In the case of 480p video streaming for subscribers with the latest smartphones, the visual experience could be worse when they use the LTE than when connected by Wi-Fi.  Why would the carrier’s do this?

When word of Verizon’s newly restricted unlimited service plan got out, Net Neutrality advocates immediately warned once again that we are witnessing the end of the open Internet: service providers eager to launch special content bundles with favored partners, are now pushing other content into a throttled and capped slow lane.


But there is another way to look at this.  Over the past few months due to competition from T-Mobile and Sprint, AT&T and Verizon have jumped into offering unlimited data plans too, perhaps before their networks were truly ready to support the traffic.  Unlimited plans must have increased traffic on their networks.  Before the launch of unlimited, mobile data traffic was already surging.  Now it must be growing even faster.  Have investments in the network been sufficient?

A recent study from OpenSignal found that a significant decrease in 4G speeds for both Verizon and AT&T in the months after they introduced unlimited speeds. OpenSignal found that Verizon’s average LTE download test fell 2 Mbps to 14.9 Mbps. T-Mobil and Sprint, which have been offering unlimited plans for a while, have only seen their average download speeds increase over the same period. OpenSignal now says that T-Mobile now offers the fastest download speeds with an average of 17.5 Mbps and an overall speed of 16.1 Mbps.

Lots of talk about super-fast access

Rarely a week goes by without a press announcement from Qualcomm, a network equipment vendor or a major carrier about a field trial delivering super-fast access to a mobile handset. For example, a few days ago, Verizon, Ericsson and Qualcomm Technologies announced a new peak downlink speed of 1.07 Gbps using the Qualcomm Snapdragon X20 LTE Modem, the first announced modem to support Category 18 LTE speeds. The Ericsson lab trial used 12 simultaneous LTE streams, which allow for up to 20 percent increase in peak data rates and capacity with a corresponding improvement in average speeds. Multiple technologies were combined to achieve the improved performance:

·         12 LTE streams with 3 cell carrier aggregation of FDD spectrum

·         4x4 MIMO per carrier (multiple in, multiple out), which uses multiple antennae at the cell tower and on consumers devices to optimize data speeds

·         256 QAM per carrier

Verizon said this 1.07 Gbps achievement builds on its recent announcement about Gigabit LTE with support for License Assisted Access (LAA). Also of significance, the 1.07 Gbps speed was achieved using only three 20MHz carriers of FDD (Frequency Division Duplex using separate transmit and receive frequencies) spectrum, achieving new levels of spectral efficiency for commercial networks and devices. These efficiencies will enable the delivery of the Gigabit class experience to more customers and lead to new wireless innovations.


In Australia, Telstra says it already has gigabit LTE running in some neighborhoods of Sydney.  Before long, peak downlink speeds for the big four U.S. mobile operators could also break the 100 Mbps barrier and zoom up into gigabit-class service.

But meanwhile, if Verizon is capping mobile video for most users at 480p that means that most of the traffic for most of their users is flowing at around 1.5 Mbps - far below the potential of current 4G LTE, not to mention these superfast access speeds on the horizon.

Why bother to restrict or block 1080p streaming at 4.5 Mbps if you are capable of delivering 100 Mbps to that same user? What gives? Where is the bottleneck? All four operators publicly say that they have sufficient spectrum, at least for now. Could it be not enough bandwidth to the towers or remote antennas? Not enough small cells? Are the mobile networks capacity constrained in the metro? Or are they lacking sufficient caching at the edge?

If mobile operators are throttling because they are capacity constrained, the weakness in the network could be at any one of these levels. We know that the mobile operators are engaging in network densification, another term for deploying small cells. But the intensity of these deployments is perhaps too low. Anecdotally, we've heard of small cell deployment schedules numbering in the thousands or tens of thousands for a national network. In comparison, Comcast recently boasted of having 18 million Wi-Fi hotspots available to its users. Maybe that's the size of small cell installation really needed to support a truly unlimited mobile plan.

Capacity constraints could be growing at the metro level. It's not hard to see how these could become saturated with millions of consumers on unlimited plans watching Netflix all day. As the 1G backhaul links to each base station become saturated, the next move would be to scale up until reaching to 10G links.  A robust metro transport network is needed with the ability to carry lots of 100Gs and perform intelligent switching. Are the mobile networks sufficiently prepared at this level?

Network equipment vendors have been saying for some time that service providers are underinvesting in their network infrastructure, at least compared to historic CAPEX-to-revenue levels. Maybe there is a case to be heard.

Protecting the 5G fixed line for residential video service

One further possibility is that mobile operators really fear video streaming will move from the small screen to the big screen without their permission. Truthfully, 480p video streaming to a 5” smartphone looks OK.  Streaming at 720p to the same device is better, especially on newer smartphones. At 1080p to the handset or higher, you reach the point of diminishing returns. The pixels are too small to make a big difference. However, on a 50" flat screen TV, there is a huge difference in quality. Tethering 480p streaming video from your phone to your TV is not visually compelling. But with a Gigabit LTE connection and an unlimited plan, streaming video at 4K could be very nice.

In Texas, AT&T is already testing residential, fixed-line 5G for delivering its DIRECTV NOW service. From a marketing perspective, maybe the plan is to preserve serious video watching as a premium service and to position mobile viewing as a casual service regardless of how fast the handsets become or how the networks perform.


OpenSignal mobile test report – August 2017 - https://opensignal.com/reports/2017/08/usa/state-of-the-mobile-network 

See also