The hype, and in many cases concern around DPI has always been strong. It has triggered provocative debate on the Internet around privacy, end user rights, the role of the operator, and the extent to which they can monitor what applications we send and receive on the Internet.
The truth is that DPI isn’t just an emerging technology; it’s actually a reality, showing up on traditional fixed line networks, enterprise networks and most recently on mobile networks. Recently, Telefonica announced the deployment of Sandvine’s network policy control solutions to provide visibility across the network for some 250 million subscribers across 20 countries.
We have a unique solution for testing DPI technology, based on our direct experience working with service providers and networking vendors that are building and deploying DPI solutions. One of the more recent use cases for DPI is around billing and charging on a per app and user basis on 4G/LTE networks.
With the release of Mu Studio for LTE testing, we are seeing mobile customers recreate thousands of mobile applications over GTP tunnels. As we showed in previous App Quadrant reports for Netflix and Skype, the end device on which the application runs on has a huge impact on both operator resources and on end user bandwidth consumption.
Spirent (formerly Mu) TestCloud content
Spirent (formerly Mu) TestCloud contains hundreds of application flows that are specific to the end device they run on and represent the exact payload, connection types and number of connections as seen on Android, iPhone and iPad devices.
The mobile customers use the mobile application flows on Spirent (formerly Mu) TestCloud to:
Test the accuracy of DPI-based detection, and
Test the accuracy of billing and charging systems
To do this, they configure the application flows with unique sets of IMSI, EBR, Application ID, Tunnel ID and source IP addresses along with specific destination IP addresses. As these users and applications belong to different rating groups the billing and charging policies are varied, and in some cases quite complex. For example, access to adult websites is charged at double the rate than other websites. A more sophisticated example is when access to an application at certain times of the day is free, while at other times it is charged a differential rate for best effort data versus video.
From a topology standpoint, the Spirent (formerly Mu) solution wraps around an SGw-PGw combination and acts as many eNodeB stations, simulating hundreds of thousands of UEs on the S1-u interface as well as a set of servers on the SGi interface.
This sort of testing with real application content not only gives them a better idea of how the network behaves under realistic application flows, but also helps to test application- or content-specific billing and charging policies.
Many of these activities are currently occurring outside the U.S. but with the rapid rollout of LTE in the U.S. it may not be long before we see some of this here as well.
Back to the controversy
As a consumer, I think this is a mixed blessing. On the one hand, I would enjoy guaranteed performance for applications that I can depend on being delivered whenever I need them. On the other hand, I might have to pay more for getting that level of service. Ultimately though, like with anything you will usually get what you pay for. If I need faster bandwidth for a video-conference with colleagues or for game 7 of the World Series, I should be able to pay for it and get what I paid for.
DPI, with the right regulations in place, can be great for both consumers and operators.