Carrier policy control and making the customer king
Cisco’s latest mobile Virtual Network Index (VNI) report continues to show exceptional growth for mobile data, which in 2011 was eight times larger than the entire global Internet in 2000. Multiple devices are driving growth, ranging from smartphones to tablets and laptops. The amount of data downloaded per smartphone tripled in 2011 from 2010, and tablet users showed an even higher consumption, generating 3.4 times the traffic of the average smartphone. Laptops went one better, consuming 22 times the data of the average smartphone.
What is surprising, however, is that smartphones represent only 12 percent of global handset sales but already account for more than 82 percent of global handset traffic. As adoption of smartphones increases and more bandwidth is made available, we can expect more consumption per smartphone. Already we can see that 4G connections generate 28 times more traffic than non-4G connections, and average connection speeds for smartphones increased 39 percent, from 968 kbps in 2010 to 1,344 kbps in 2011. The expectation is that connection speeds will increase 9-fold by 2016, generating an 18-fold increase in global mobile traffic.
Informa has also seen similar trends as indicated in its recent “Mobile Content and Applications Forecasts” report, which predicts that in 2016 mobile phone users will, on average, consume 6.5 times as much video, eight times as much music, and nearly 10 times as much in games as in 2011. Another key finding in the Informa report was that capitalizing on this data traffic increase will prove a challenge to carriers – global mobile data traffic is forecast to grow 10-fold from 3.89 trillion megabytes in 2011 to 39.75 trillion megabytes in 2016, however global mobile data revenues are only forecast to double in the same period, from $325.8 billion in 2011 to $627.5 billion in 2016.
Herein lies the dilemma for carriers. Increasing bandwidth is, in the first instance, a cost, as most services are charged on a fixed monthly rate basis and often with an “all-you-can-eat” offering. Thus, an increase in bandwidth consumption per user does not result in an increase in revenue, only an increase in cost.
On the other hand, this challenge lies at the heart of a necessary transition that carriers are making from their traditional utility model of Supply to a retail-focused model driven by customer Demand. The utility mindset is based on a “take-it-or-leave-it” proposition, while a retail mindset is based on “the customer is king” proposition. As can be expected, making this mindset transition will take time for both carriers and their customers.
Phasing in “retail” bandwidth
It is not surprising that the initially proposed solutions to the increased bandwidth consumption versus flat revenue effect (often referred to as the “scissor effect”) lean more toward a utility mindset than a retail mindset. Traffic shaping and the throttling of “unwelcome” traffic are the first phase in a series of solutions to the problem, and many consumers today have mobile data plans with a set limit on data downloading; when the limit is exceeded, the download speed is greatly reduced.
However, the next phase of more intelligent services are already appearing and have shown a willingness to adopt a more retail mindset that offers services in a way that reflects customer preferences and service usage. For example, data plans are being introduced that increase bandwidth limits at certain times of day, in certain locations, or for certain applications. The most ambitious of these services provide customers with their own portal to “fine-tune” their service and preferences, such as more data for Facebook and less for email, more on Monday and less on Sunday, or more at home and less at work.
For all these options, be they traffic shaping or tailored services, the key underlying enabler is network intelligence and policy enforcement. As a carrier, you need to be able to profile network usage in real time to understand the dynamic usage of this shared medium. You need to know what kind of data is being exchanged, where it has come from, and where it is going. You need to be able to recognize what kinds of applications are being used. Only with this information can you make a policy decision on whether to filter certain traffic or enhance bandwidth allocation for certain users.
Solutions exist today that support network intelligence and policy enforcement within the existing network infrastructure. These include network probes for real-time data collection and analysis (otherwise known as “Deep Packet Capture”), dedicated Deep Packet Inspection (DPI) systems that combine Deep Packet Capture with application recognition, as well as policy servers that maintain and implement policy decisions. The challenge for all of these solutions in the coming years will be the ability to efficiently handle the vast amounts of data that are expected to be generated.
Processing for next-gen policy control
Most network connections today are either 1 Gbps or 10 Gbps, but with an 18-fold increase in mobile traffic expected over the next four to five years, carriers are already looking at 40 Gbps and 100 Gbps line rates in planning for the future. At these rates, an extraordinary amount of data needs to be collected, analyzed, and acted upon in very short periods of time. For example, at 10 Gbps a fully loaded connection can generate up to 30 million packets per second that need to be analyzed with as little as 67 ns between each packet. At 100 Gbps, this means 300 million packets per second with only 6.7 ns between packets.
The key for developers of network probes, DPI systems, and policy servers is to ensure that their solutions can scale to meet this demand with zero packet loss collection of data and efficient processing for analysis. Efficiency, in this regard, addresses the challenges of having enough processing power available to process huge amounts of data, but in a form factor that provides low power consumption and does not consume excessive space.
The strong roadmap provided by standard servers using x86 architecture chips promises to keep up with this challenge. With annual increases in processing performance of up to 60 percent, a 10-fold increase in processing power will be provided over the next five years. This does not quite match the 18-fold increase in data predicted for the same period, but with intelligent pre-processing adapters, like those provided by Napatech, it is possible to offload the processing burden and ensure that only relevant data is processed in the server. In addition, intelligent network adapters today can ensure zero packet loss at speeds up to 40 Gbps while conforming to PCIe standards for low power consumption and compact form factors. The combination of standard servers and intelligent network adapters thus provides a powerful hardware platform for network probe, DPI system, and policy server development that also addresses concerns with power and space constraints.
The changing landscape of mobile data communication provides carriers with an opportunity to complete the needed transition to a retail mindset and address the scissor effect. An efficient and scalable network intelligence infrastructure provides the engine for driving this process, and provides carriers a cost-effective strategy for keeping up with the challenges of increasing bandwidth usage.