Tech Insights-Streaming Video: The (Network) Killer App
Early cheerleaders for 3G failed to understand or chose to ignore that radio channels carrying data between base stations and individual users are shared resources. 4G networks will not save the day.
Back when wireless operators were first deploying third-generation networks and before the advent of smartphones, there was an ongoing quest to find the "killer app." This was to be the signature data application for nascent 3G networks that would excite traditional voice users sufficiently to get them to part with a few more dollars per month for data services.
The big problem was that the very limited user interfaces on basic cell phones made most interactive applications cumbersome. But one application emerged that didn't require all that much user interactivity and could take advantage of the higher bandwidths offered by 3G. That application was video, and in particular streaming video. Sure, the tiny screen on the typical handset didn't afford a particularly satisfying user experience, but the novelty of being able to watch YouTube videos and – let's be honest about it – "adult entertainment" on their phones appealed to a certain portion of subscribers. If there ever was a "killer app" prior to widespread uptake of smartphones, it was streaming video.
Unfortunately, while streaming video was a good 3G application from the standpoint that it didn't require very much user interaction and thus worked reasonably well on phones with limited user interfaces, it was just about the worst possible application in terms of network operation and performance. That its use was actually touted and encouraged in carriers' early advertisements for 3G service speaks to a fundamental difference between what those networks could deliver and what the public was led to believe they could.
The most obvious problem with streaming video on wireless data networks is that it is a bandwidth hog. That's hardly an earthshaking revelation, of course. But the impact of high bandwidth applications was clearly not well understood in the early days of 3G network operation, at least not by the folks charged with marketing 3G services. A typical streaming video with modest resolution might require a downlink (forward) channel throughput rate on the order of 300 kbps. A typical 3G channel can support downlink speeds well in excess of 1 Mbps. The DSL services available at about the same time from local wireline telephone carriers typically delivered about 1 Mbps downlink speed, and streaming video worked fine on DSL. "So," the marketers apparently thought, "streaming video should be no problem for 3G, right?"
WRONG! What the early cheerleaders for 3G services failed to understand, or perhaps chose to ignore, is that the radio channels carrying data between base stations and individual users are shared resources. That is, whatever the capacity of a channel might be, it has to be shared among all active users that the channel is serving at any given time, and a single user watching streaming video is consuming a significant portion of the available downlink capacity of a 3G channel. DSL service, by contrast, provides dedicated links to each user. From the user's perspective, comparing the performance of a wireless data channel with that of a wire- or fiberoptic borne channel of equivalent capacity is like comparing apples and watermelons.
Another brutal impact of streaming video on wireless data networks is that the throughput speed available to an individual user at a given time is hugely dependent upon that quality of the serving radio channel. Users located at the "cell edge," where received signal strength is lower and interference is greater, will typically get a fraction of the "up to" speed that the channel can deliver to close-in users. Unfortunately, the fraction of channel capacity consumed by a given user does not decrease when its throughput speed is reduced by channel quality degradation. Instead, it's the channel capacity that is reduced. This is another characteristic of wireless data networks that is apparently not well understood. All users with less-than-perfect signal quality will negatively affect the capacity of the serving radio channel, but those that are streaming video will have a comparatively huge impact because they consume such a large portion of total channel capacity.
Finally, there's the "streaming" part of streaming video. Streaming is much more impactive on network quality of service (QoS) than, say, downloading video for watching later because the constituent data packets for streaming have to be sent with high priority. Streaming services generally utilize a data buffer at the receiving end so that a momentary hesitation in the communications data stream will not cause interruption in the viewed video, but the transmitter (i.e., the serving base station) still has to keep things moving to prevent the buffer from emptying. So as the channel becomes congested, the QoS for other users engaged in more benign applications like Web browsing may degrade in order to maintain uninterrupted video streaming. By the way, this suggests that video conferencing apps will be even worse than streaming video because they are much less tolerant of the delays associated with data buffering.
4G WON'T EASE BURDEN
Unfortunately, while smartphones have made practical an endless array of more network-friendly applications, streaming video remains immensely popular and threatens to strangle some 3G networks. And recent moves by network operators to constrain throughput demand with monthly data caps will probably not provide much of a solution. A user with a 3 Gbyte monthly cap will likely think twice about spending a third of that allowance for downloading a movie, but have no problem with streaming a couple hundred YouTube videos at 5-6 Mbytes each. The increased channel capacity of 4G networks likewise won't ease the burden of streaming video applications, since economics will dictate that each channel serve a larger number of active users. In fact, the greater bandwidths afforded by 4G may encourage users to seek higher quality streaming video, which will make matters that much worse.
I believe that wireless data network operators are going to have to become much more proactive in stemming the channel congestion associated with streaming video apps. This will likely require a two-part strategy. First, carriers will need to make available to their subscribers an acceptable alternative to streaming video. One approach might be the use of on-demand video "pre-positioning" as provided by Opanga Networks (www.opanga.com). The second part of carrier strategy will be to come up with service pricing plans or terms of service that make streaming video either unpalatable or impractical, at least during periods of channel congestion.
It's high time that we in the industry admit to ourselves, and then to our customers, that unfettered streaming video is an application fundamentally incompatible with the economically prudent operation of wireless data networks. Then we need to do something about it.
Drucker is president of Drucker Associates. He may be contacted at firstname.lastname@example.org.