Few things in life ever work out quite as anticipated. For mobile operators in recent years this truth has been best illustrated by the sudden growth in demand for mobile data, which has provided a number of unexpected developments, not all of them pleasant. The impact of what have become known as “chatty apps” on operators’ signalling loads has been one of the most challenging surprises.
Smartphone penetration has exceeded 50 per cent in a number of advanced markets and the emergence of truly low cost devices from Far Eastern vendors is increasing global penetration fast. Give a person a smartphone and they will immediately visit the relevant application store and start downloading apps. Some people are insatiable, others more reserved—but within a fairly broad demographic swath, you can expect a number of obvious selections.
Facebook has well over half a billion active monthly mobile users. While it is not known exactly how many of these use the app rather than browser interface, it is safe to assume, given the penetration of smarpthones, that they are in the majority. Twitter is also hugely popular; there are more than 400 million tweets each day, and the majority of users are mobile. Social gaming, messaging applications and email are equally sticky—and all of these applications are “chatty”; they are constantly pinging the network to check for updates.
In one famous instance, says Chris Hoover, vice president for product management and marketing at Openet, an Angry Birds update that served an ad on completion of every level of the game placed such strain on one operator’s network that, for an uncomfortable period of several days, 90 per cent of voice call attempts failed because the signalling needed to establish the calls could not get through.
There is no way that anyone in the industry, operator, supplier or developer, could have anticipated that.
In fact, says Hoover, operators were planning for data uptake that reflected existing subscriber behaviour, most of which was browser-based. They were expecting handsets to ping the network when a user was looking for a train time, trying to book a cinema ticket, or in real time when a user launched an email application.
“I think what surprised most operators was the unpredictability of application usage,” says Hoover. “It was kind of hard to predict that so many different apps could be deployed on the same device and would, asynchronous of one another, be querying the network very frequently. This confluence of factors—the quantity of apps, their popularity and the unpredictability of their queries—caused this problem.”
It is a problem that seriously needs addressing. Operators can ill-afford downtime; it angers users and damages reputations. In the UK earlier this year, O2 was forced to offer compensation to tens of thousands of users for a network outage that lasted the better part of a day.
The most frustrating characteristic of the chatty app for any operator is the fact that so many of its requests for data are unnecessary. Do users really need to be alerted to Facebook status updates as they happen? Not really; it would be perfectly functional for the app to retrieve the information only when fired up by the user. But developers of social apps in particular are desperate to drive urgency, so the frequent, needy pinging will remain.
In an ideal world it might be possible to manage the problem at a developer level; to set acceptable frequencies for requests and to require developers to synchronise their activities. “There are some attempts to do this but, really, what incentive is there for these developers to co-ordinate their work?” says Hoover. “The number of developers and the number of apps out there mean that people just pursue the path of least resistance.”
An alternative but expensive approach might simply be to deploy more network capacity. This is already happening for many operators involved in, or on the cusp of, the latest generational change. LTE has been identified as a panacea for a number of industry headaches, says Hoover, and the refreshed signalling solution it brings with it offers a potential solution.
“The optimised signalling with LTE could well address this problem, in the long run. But the honest answer is that nobody really knows. A lot of operators are looking to LTE as the answer to their congestion problems but I’m a little sceptical that it will be that easy,” Hoover says.
“All of the familiar challenges remain with LTE in terms of spectrum fragmentation and the rollout of the network itself. Most importantly, in the short and medium term the signalling problem will still exist anyway, as 3G networks are going to continue to be very important for operators until LTE networks are built out on a wide scale.”
A simpler, cheaper solution exists, says Hoover, which uses the smartphone itself as a control point, marshalling the activities of each user’s stable of applications so that the device’s impact on the network, in terms of signalling, is minimised.
Just as the problem was unanticipated, says Hoover, so was Openet’s development of this particular solution.
The firm was working on a solution that installed a client to the user device that helped users understand their data consumption in real time and buy top-ups and new services accordingly. This involved pulling information to the device from the operator’s Policy Charging Control (PCC) system, as well as the PCC reaching out to the device, he says.
“We thought that if policy could reach down into the device, maybe we could use the device as an enforcement point and do some interesting things,” he says. “Around this time the signalling issue had started to come to the surface. One thing led to another and the policy turned out to be pretty simple.
“We’re not restricting the app’s access to the network but we’re co-ordinating it so that, instead of multiple apps querying the network in isolation, at random times, our client cache’s the queries and sends those queries simultaneously. There’s no difference in terms of updates or notifications from a user perspective but, from a network perspective, they’re much more predictable.
Operators are free to set their own rules as to how often the queries happen, or to apply different rules to different applications. So far, Hoover says, they are preferring not to try and assert control over primary apps, ones that are actively used by the consumer. Instead they are looking to manage the output of apps that run in the background, of which there can be many.
For now the solution is restricted to Android only, because Openet wanted to work in a more open environment to see whether or not it had come up with a winning idea. In a recently concluded trial with a large North America operator, which ran to the thousands of users, the signalling burden was reduced by 30 per cent, which led to a forecast saving on Node B capex of $45m over three years. Buoyed by the outcome, Openet is now looking to develop a similar solution for iOS and Microsoft-powered devices.
The benefit is increased Hoover says, through the operator being able to use the client in conjunction with other policy functions such as smart charging and advanced notifications.
The solution may work technically but there is an obvious potential problem: The extent to which users are prepared to accept it remains to be seen. Consumers are understandably wary of any client-side intervention in their mobile behaviour, especially in the wake of the logging activities that emerged on iPhones last year. The extent to which they can be made to understand the issue—and even then whether they care about it—is a significant unknown.
Hoover argues that operators’ best chance of success in this regard is simply to be as open as possible with their customers, and give them the chance to opt out of the programme. They can also be tempted by an additional benefit: Openet claims to have derived savings in battery life on certain devices of 50 per cent when the activity of installed apps is properly managed.
But consumers are perhaps not the only stakeholders who might be uncomfortable with the idea of a device-side solution. Conventional operator thinking is that the network is the best place for important decision to be made. How ready will they be to embrace the alternative.
“The distinction between the device and the network is artificial,” says Hoover. The device is simply the end of the network. The comfort level that operators have with network based solutions is because that’s where the control used to be. Device-level controls are new and there will be a time of getting used to them and understanding their potential and their limitations. But in time it will just be another entity that the PCC systems interact with , no different from any other system in the network.”