FUTURE of mobe tech: Today will be dry with chilly gusts of 12Mbps
Mobile web 'weather reports' predicted as networks overload
Internet monitoring isn't just for the US government as mobile network operators turn to real-time analysis to work out what we're doing, and what we're going to be doing, too.
The live data processing comes from Actix, which reckons we should expect connectivity congestion reports from mobile operators (eg: there's heavy data downloading in town but it'll otherwise be a fine, dry day) so we can avoid making unreasonable demands of our networks - such as expecting them to work while one is lost in a busy city centre.
Telefonica, on the other hand, plans to avoid the problem by delivering data before we knew we wanted it.
Five per cent of mobile phone mast cells carry half of all mobile traffic, according to Acrix, which makes a living monitoring such things for the operators. Delivering that data to the right cell is overloading the infrastructure so operators have an interest in balancing the load.
"Mobile operators may have to start issuing their own network traffic reports to subscribers if they hope to avoid ‘mobile rush hours’ where calls drop and data communications crawl," says the company, suggesting that customers entering an overloaded area could be warned to expect slow data and act accordingly, unless they had paid enough for priority access:
"[operators] are considering varying charges for data downloads based on whether subscribers need it immediately, even when the network is busy, or are willing to wait and download or watch later ... When network demand is lower, subscribers would be given the green light for cheaper per download costs."
Mobe biz Three started throttling connections back in 2009, and ensuring its most-profitable customers got the best access, but a better approach would be to cache the data near the customer before the customer knew they wanted it, which is what Telefonica's Tailgate project tried to achieve.
The idea was to monitor social networks and work out who shares data with whom, and ensure the data was cached nearby before it was needed - thus allowing it to be quickly served to the punter rather than all the way over the infrastructure to the public internet. The result managed to decrease YouTube latency by half - requested videos started up in half the time - as well as making it cheaper to deliver.
The project ran through 2012 and used Twitter as the dataset is open, and analysed 41.7 million accounts for behavioural information. The team was able to get a rough location on just over eight million twitter users who then formed the core of the study which focused on how followers responded when a YouTube video was tweeted. The concept is simple - if someone in Scotland, who has a lot of Twitter followers, posts a link to a YouTube video then it's worth ensuring that video is cached somewhere north of the border where her followers are, but the improvement gets even better when user-generated content is optimised.
Imagine our proud Scot is across the pond exploring the Rockies. She posts a video of a jackrabbit escaping his sgian and promptly uploads it to YouTube (in ignorance of mobile coverage or indigenous wildlife), the TailGate software notices this and knows where her mates live, ensuring the Leith network has a local copy of the video before her mates are out of bed.
The Telefonica chaps were joined by boffins from the Polytechnic University of Turin, the Swiss École Polytechnique Fédérale de Lausanne, and the American Northwestern University, with a chunk of EU cash helping to pay for the whole thing, but the technique developed by TailGate was surprisingly effective.
TailGate also demonstrates the advantages of analysing the content of traffic, as well as its existence. Operators have long known that traffic exists, and there's often too much of it, but knowing what's being said can be equally valuable, and not just to overly paranoid security services. ®
Sponsored: Hyper-scale data management