Future-Proofing Network Management
AT&T took some heat last fall for imposing limits (some of which have since been removed) on which subscribers can use FaceTime, Apple’s video calling service, over its cellular network. This week, the FCC Open Internet Advisory Committee published a report on the subject that I helped to author. Its technical analysis is well worth a look.
The OIAC report recognizes that mobile networks have limited bandwidth and that a high-bandwidth application like FaceTime can cause substantial performance problems on the network. But does that mean that network management techniques need to target particular applications? One view represented in the report, as well as in CDT’s most recent blog post on the subject, is that it would be much better for application developers and consumers if network management were to be application-agnostic. The growing popularity of mobile devices and applications is likely to continue to contribute to network utilization and congestion. Instead of focusing on a particular app-of-the-moment like FaceTime, it makes more sense to develop application-neutral network management solutions capable of dealing generically with changes in network usage.
There are a variety of application-neutral approaches available to network operators. One option is congestion-sensitive traffic management, which slows or limits the traffic of the customers using the most bandwidth during congested times, regardless of what application the customer is using. This is the approach that Comcast has been using for the last several years to manage the traffic of its 19 million cable broadband customers.
To make congestion-sensitive traffic management even easier, I’ve been collaborating with a number of network operators within the Internet Engineering Task Force (IETF) to create a standardized mechanism known as Congestion Exposure (ConEx). The idea is to give network operators a standard way to judge how much congestion each user is creating. Last month, the initial ConEx document was published as an official IETF RFC. Although ConEx-based solutions are nascent, the success of the Comcast solution shows that congestion-sensitive approaches can be viable, reliable ways of managing new traffic demands on growing networks even as application usage changes.
As the OIAC report notes, per-customer rate-limiting is a potential solution even when it isn’t triggered in response to actual congestion. Limiting the amount of bandwidth available to each customer in localized hot zones where congestion is likely to occur (stadiums or train stations, for example), or during congested hours of the day would have a limited impact on lighter users while allowing all users to choose their applications as they see fit. Over a longer time frame, carriers could scale up network capacity or apply other bandwidth management techniques (such as offloading cellular traffic to WiFi) in those same frequently congested zones.
Another potential option is to let the user decide which of their applications to prioritize during times of congestion. This approach benefits some applications at the expense of others, but it would leave the choice about which applications benefit in the hands of users, not carriers.
While the approaches above focus on how bits are managed on the network, similar outcomes are possible by adjusting pricing. If carriers were to charge more for data used during busy periods or in busy locations, it would give consumers incentives to shift their data usage to other times and places. Explaining fine-grained congestion pricing to customers may be difficult, but simpler approaches have been in use for years and are quite familiar to most consumers: just ask anyone who has ever waited until the evening or a weekend to make a phone call, or turned an air conditioner off in the middle of a summer day. These approaches provide targeted incentives – certainly more so than the monthly caps on many mobile data plans, which are oblivious to whether data is used during congested or uncongested periods – but are likely simple enough for consumers to understand and adjust their behavior accordingly. These approaches could also be bundled with real-time notification to users, so they can be more aware of when the network is congested.
When the public learns that a broadband provider has adopted a new network management strategy, the provider’s chosen approach is often treated as if it’s the only available option. In fact, there are many alternatives – including application-neutral and user-controlled ones – for carriers to choose from. Additional network engineering may be required to make some of them happen, but isn’t that what we pay broadband providers to do? Choosing an application-neutral strategy lays a foundation to deal with congestion challenges posed by whatever the hottest new app happens to be – this year, next year, and into the future.