Hogs & Bandwidth
iPhone highlights data network realities related to peer-to-peer transfers.
Two items regarding operation of the new 3G iPhone made the news recently, but garnered far less attention from the industry than I believe they deserve.
In one, AT&T announced that it would ban from its network any iPhone users that it catches doing so-called peer-to-peer file transfers on its 3G network. Then it became known that iPhones will not be able to download music from Apple’s popular iTunes site using 3G. Instead, as with the older version of the iPhone, users will have to wait until they are within service range of a Wi-Fi hotspot before they can spend their money on the latest hit songs.
Together, these limitations on 3G iPhone operation may portend a profound effect on the marketing of wireless data network services. This impact will not be limited to the iPhone, or even to smartphones in general, and won’t be cured by a future transition to 4G technologies. That’s because these limitations spring from the fundamental differences between wireless networks and other means of Internet access. If network operators fail to recognize these differences and market their services accordingly, they may find it impossible to economically meet customer expectations.
What peer-to-peer and music download applications have in common are large file transfers. A typical song in MP3 format is maybe a 5 MB file. That may not seem all that large, but it represents the entire capacity of a 2 Mbps packet data channel for 20 seconds. A peer-to-peer video file might easily be 10, or even 100, times as large. Clearly, AT&T is concerned that even a relatively small number of such huge file downloads among its users could throttle the network.
The bottleneck isn’t in network servers or routers, where upgrades can be made in a fairly straightforward manner, but rather in the capacity of the air interface. Herein lies the fundamental difference between wireless data services and such wire-borne services as DSL: In the radio network, the connection to the user is a shared channel. A 3G channel in a given network base station sector may be simultaneously serving a relatively large number of “active” users – that is, users who are actively engaged in activities involving data transfers. Whatever the nominal throughput capacity of that channel, it must meet the collective requirements of all of these active users.
SHARING TAKES A HIT
What makes the shared channel arrangement practical is that most users really don’t demand that much throughput when averaged over some reasonable period of time. A typical Internet surfer may want to look at a different Web page every minute or so. If the average page download is around 100 kb, and the channel is serving 100 such users simultaneously, the collective download throughput demand will average only 1.33 Mbps.
Because wireless data network protocols are geared to efficient shared use of the channel resource, a channel with a throughput of 2 Mbps will probably be able to deliver to each of these users a perception of very good download speed. But if just a couple of the users want to engage in large file transfers and are allowed to do so without constraint, perceived throughput rates for everybody will take a huge hit.
By the way, such performance degradation due to “channel hogging” is quite apart from, and in many ways more serious than, speed variability associated with RF channel quality. A user may reasonably expect speeds to go down in areas of low signal strength, but will likely be very irritated if things slow to a crawl when the indicator shows “four bars.”
In order to preserve the perceived performance of 3G data networks, carriers will have to rein in applications involving large file transfers. AT&T has done this rather bluntly by banning peer-to-peer apps and by preventing iPhone music downloads from iTunes, but there are plenty of other Internet applications that involve large file transfers.
The immensely popular YouTube, with its streaming video, is but one example. As 3G usage rises, with growing popularity of smartphones and even more with increased use of 3G modems on laptop computers, simply banning specific applications may not be enough to prevent network gridlock. A more effective approach would be to limit individual users to a maximum throughput over a given period of time – maybe 10 minutes or so.
The effect of such a constraint on large file downloads would certainly dampen consumer enthusiasm for abandoning wire-borne broadband services in favor of wireless Internet access. By some accounts, such downloads account for well over half of all Internet traffic.
You might think this problem will be greatly diminished with evolution to the much higher speeds of 4G networks, but I believe that without careful management of customer expectations exactly the opposite might occur. After all, from the user’s perspective, the primary advantages to higher throughput rates are faster downloads of large files and higher bandwidths for streaming media.
Wireless data networks are great for applications involving bursty data transfers with low average per-user throughput, particularly where user mobility is helpful or necessary.
But they are not so good for applications, mobile or not, involving high average throughputs. It’s critical that operators recognize this, and equally critical that they adjust customer expectations accordingly.
It may seem inviting to suggest that users can make use of 3G and 4G networks as they would a wire-borne service, but having to maintain service quality while meeting that promise will inevitably lead to disaster.
Drucker is president of Drucker Associates.
He may be contacted at firstname.lastname@example.org.