Shape and offload traffic, but mind your combo

While flat-rate, all-you-can-eat data plans has been critical to the growth of mobile broadband, it has caused mobile data traffic to surge while revenues remain almost flat. Carriers need to add a large amount of extra network capacity, yet profits from these services are insufficient to fund the necessary upgrades.
 
Operators now are turning to a cocktail of new remedies based on traffic management – but the wrong mixture can prove poisonous. For many, the solution to this decoupling of revenue and capacity appears obvious: LTE, which will provide greater capacity at a lower cost than existing 3G networks. But with mobile broadband traffic doubling every nine months, it has become clear that LTE alone will be insufficient to meet demand.
 
Operators are now increasingly turning to traffic management, specifically two techniques – traffic shaping and traffic offloading. While all traffic is essentially treated equally, traffic shaping allows operators to prioritize some services over others, thereby improving the consumer’s experience. This is being increasingly delivered using deep packet inspection (DPI) technology.
 
Simply put, DPI looks “deep” inside data packets and classifies them according to a variety of criteria – all in real time. Operators can then prioritize, rate limit and in extreme scenarios, block different types of traffic. DPI capabilities allow highly accurate identification of network traffic such as BitTorrent, YouTube, Skype, viruses, spam and others, even if these applications use port-hopping, encryption, masking or other behaviors to hide their identity.
 
Operators can then implement rules-based management and traffic shaping to allow allocation of bandwidth based on criteria like the application (such as VoIP), roaming versus non-roaming, time-of-day, volume or per user. As such, operators can prioritize time-sensitive services such as voice or media streaming.
 
The other technique, traffic offloading, allows operators to offload internet traffic from their networks at the earliest opportunity so that they don’t have to invest further in increasing capacity at the network core. DPI is again frequently used to differentiate between the traffic that should be offloaded and that which should continue into the core network.
 
For instance, an operator might wish to transport a revenue-generating VoIP service over its core network to guarantee quality of service and offload an over-the-top, non-revenue generating service such as Skype. In simple terms, DPI is being used to classify traffic and then offload part of it according to policies set by the operator with the rest passing through the network core.
 
These two new traffic management techniques have separately allowed operators to dramatically improve the consumer mobile broadband experience without having to provide costly additional capacity. However, news from some early deployments has been that they can create a poisonous combination.
 
The problem with introducing a traffic offloader with a currently deployed traffic shaper is that traffic shaping can stop working properly. But why? When traffic shaping and traffic offload are deployed together, traffic is offloaded before the shaper is able to classify it. This means that traffic has not been rate limited or prioritized so that it causes severe congestion on the radio network on its return trip back to the user.
 
The solution is to move the traffic shaper out of the core network into the same unit as the traffic offloader in the radio access network. Not only does this vastly improve traffic management, it also allows traffic shaping to become adaptive. This means that shaping can be carried out locally based on activity at a specific base station, allowing operators to improve the quality of their networks in a much more targeted manner.
 
The travails of fixed fee tariffs
 
The travails of fixed fee tariffs
 

Mobile data traffic is increasing fast yet corresponding revenues are tapering off

 
Manish Singh is VP for product line management at Continuous Computing