Congestion in the RAN is a devilishly tricky problem to tackle, but it's not a problem that's lacking in proposed remedies. In its new report,
"RAN Congestion Control & the Road to QoE," Heavy Reading identifies 32 companies selling RAN congestion control solutions of one kind or another, and that number is most likely an underestimate. There's a mini-investment boom going on in this area.
It's easy to see why: Mobile broadband has made the life of the RAN engineer a lot more difficult. It isn't simply that there's a lot more data traffic; other headaches include highly skewed usage patterns (a few users streaming video can easily clog a cell), more random congestion events (history is no longer such a good guide to where congestion will occur); congestion in the signaling plane; and varying impact on proliferating user applications, some of which are more sensitive to congestion than others.
Many now believe that focusing on customer quality of experience (QoE) may be the key that will solve this puzzle. This has quite a few attractions. Operators can justify the investment in new controls because QoE demands an end-to-end view – including a view into the RAN, at cell level – if appropriate and cost-effective QoE remedies are to be applied. Moreover, a QoE view takes account of the differences among applications noted above, as well as differences among customers (such as subscriber value/tier, or stated preferences).
But a QoE approach raises tricky issues of its own. How do you measure QoE? What exactly is congestion, and when does it become unacceptable? And how do you make the information feeds as near real-time as possible?
To make matters worse, there's no overarching standards initiative in this area for operators to turn to – though 3GPP's recently formed User Plane Congestion (UPCON) group may help fill that gap.
Despite these obstacles, there's no shortage of ingenuity on show among solution providers. Some solutions are based around video optimization, inferring congestion from metrics such as video buffering and packet loss. Others rely on probes in the network to feed back information. And some are pitched between real-time and historic data approaches, for example providing a read on congestion every 15 minutes. Analytics is a strong (and strengthening) theme in many solutions. And one constant that nearly all agree on is that RAN congestion control systems need to feed information to policy servers to provide operators with the widest possible set of options for handling the congestion event.
It seems probable that a thousand flowers will persist for a while, not least because operators have quite different views about what's needed here. And while there are plenty of obstacles and uncertainties that may yet make this a slow burn, our surveys show that operators are highly receptive to the idea of better controls, and ready to spend on solutions that make a positive contribution to QoE.
In this exciting field, some fortunes are likely to be made, and in the ongoing battle to differentiate services, pioneering operators deploying innovative systems may steal a valuable march over rivals. There may not be that much to see in real networks as yet – but watch this space.
— Graham Finnie, Chief Analyst, Heavy Reading
For more information about Heavy Reading's RAN Congestion Control & the Road to QoE, please contact:
