A study released in June by Cisco Systems presents a sobering picture for those of us in the wireless industry concerned about meeting the growing demand for data services. According to this study, wireless data throughput in North America will increase nearly five-fold between now and 2018. This will considerably outstrip the anticipated growth rate for traffic over wired Internet access, so that by 2018 it is expected that consumers and businesses will send and receive more data on their wireless devices than over wire- or fiberoptic-borne services. It should be noted that the forecast for “wireless” lumps together traffic over public Wi-Fi systems and cellular networks, but even taking this into account it is clear that Cisco expects many Americans and Canadians to ditch their wired broadband service for wireless the same way we did for telephone service a decade or so ago.
If you dig into the report a bit, the biggest impact of a shift to wireless Internet access is expected to be a huge increase in the use of wireless networks for watching streaming video. In 2013 streaming video from services like Netflix and YouTube accounted for 66 percent of global consumer Internet traffic, a figure which is expected to grow to 79 percent by 2018. No matter how you cut it, if 50 percent (or more) of Internet traffic is going to be delivered over wireless networks, and 79 percent of Internet traffic is going to be streaming video, there is going to be a whole lot of video on our smartphone screens.
So, assuming the Cisco numbers are a reliable peek into the future, we come face-to-face with two pretty significant facts for the wireless industry to deal with. First, we will need to grow the capacity of wireless data networks in North America by around 500 percent in the next four years or so. I think it’s reasonable to assume that the amount of spectrum available for wireless broadband won’t quintuple in that time (or ever, for that matter), so we are going to have to significantly improve the throughput efficiency of the spectrum we’ve got. Second, much of the traffic will be streaming video, so it would probably be a good idea to focus on improving the efficiency with which wireless networks carry that particular type of data. To do that we may have to abandon some of our current beliefs about how data should be carried over the air interface.
Within the Internet, streaming video is of course carried in packet format using the Transmission Control Protocol/Internet Protocol (TCP/IP, or just IP for short). IP is not a very spectrally efficient way to transport streaming media, partly because of the substantial per-packet data overheads required. That’s not a huge deal on the Internet, where the vast majority of traffic is carried on fiberoptic cables of gargantuan bandwidth. But when packets of streaming video arrive at a wireless network base station and have to be transmitted over the very bandwidth-limited air interface, spectrum efficiency suddenly becomes the holy grail.
In today’s “all IP” wireless networks, a number of techniques are employed to improve the efficiency of packet data transport over the air interface, the most important of which is IP header suppression. By its very nature, though, the structure of packetized data makes it less than ideal for transport of streaming media on wireless networks. Not only does it require a lot of overhead control channel signaling, it also makes it tougher to manage mutual interference, so spectrum reuse density, and thus overall spectrum utilization, suffers.
If we can free ourselves of the notion that data on the air interface has to retain packetized format, we could develop a channel structure optimized for streaming media – perhaps as part of the future “5G” technology. I tend to favor broadband spread spectrum with very tight, closed-loop transmit power control – sort of like CDMA 1x on steroids – but there are perhaps even better technologies for maximizing streaming media transport efficiency. In any event, if Cisco is right about our wireless data future we should probably start thinking outside the “all IP” box.