Cam Cullen is the Vice President of Global Marketing at Procera Networks. He is responsible for Procera's overall global marketing and product management, and is an active evangelist for Procera's solution and general market trends as well as an active blogger for Procera. He joined Procera as VP of Product Management to implementproduct strategy and to expand the company's product offering.
Prior to Procera, Mr Cullen held senior Product Management and Marketing roles at Allot and Quarry Technologies/Reef Point Systems, where he was VP of Product Management and Marketing. He also held various roles in business development, marketing and sales at 3Com. Mr Cullen was a captain in the US Air Force, where we worked at the National Security Agency and the Air Force Information Warfare Centre.
Cam Cullen holds a Bachelor of Science degree in Electrical Engineering from the University of Alabama.
The rapid mobile data growth necessitates prioritisation of broadband resource by personal preference and QoE, i.e. personalised charging. Operators are seeking value-charging, beyond ‘fair usage’ and beyond volume/time based charging. Great improvement of QoE perception is achieved by letting consumers select application priorities, so that their favourites will perform better. This requires intelligent policy based infrastructure that can detect the service type, define the policy per user, enforce it in both the ‘last mile’ and the backbone, and charge for it accordingly. The Policy server and the charging systems must scale to support numerous transactions in real-time. They must enable introducing novel charging-based services without impact on performance.
The broadband market has matured considerably from the days of email and simple web browsing. Consumers today want quality access to streaming video, social networks, and real-time news feeds. The expectation of a consumer is that their access will be fast, with minimal latency and reasonably priced. It is not uncommon for news to break on Twitter and Facebook before any news outlet (print or video) manages to break the story – a good case in point for why broadband access has become a utility and not a luxury.
However, as bandwidth consumption is growing, the challenge has also grown for broadband providers to meet consumer expectations in a cost effective manner. The “last mile” is always a potential bottleneck in broadband deployments (whether they are fixed, mobile, or wireless), and the backhaul is also an issue as it is routinely oversubscribed. In order to allow broadband operators to manage their networks, many new technology options have been explored to optimise networks. For example, operators have implemented versions of “fairness” (often on fixed networks), explored usage limits on both mobile and fixed networks and rudimentary “power user” traffic management.
However, operators have quickly realised that none of these solutions meet the challenge of quality broadband access at a reasonable price. Consumers judge their internet experience by the applications that they use: a social networker uses picture uploading and tweets or status update refresh, whereas a video streamer judges by how many times they see “buffering” on their video screen. ‘Fair use’ does not take this into account, as the consumer cannot choose what application that they want prioritised during congestion, or manage what applications can consume what portion of their usage limit.
As a result, broadband operators have begun investing in Policy Management and Policy Enforcement solutions. Policy Management enables operators to manage policies in their network and enables a ‘personalised’ service for each subscriber on the network. The linchpin to the success of Policy Management is Intelligent Policy Enforcement (IPE), which allows the creation and enforcement of services at the subscriber, device, location, and application perspective. IPE enables operators to create and enforce these service plans as well as help operators with analytics on the types of attractive service plans that they should offer to consumers.
IPE enables operators to better manage congestion by allowing Quality of Experience (QoE) expectations to be controlled during congestion. An IPE-based congestion management solution allows a consumer to choose what applications that they want prioritised out of their “fair use” allowance – so the video streamer can prioritise video, the social networker can prioritise Facebook and Twitter. This ensures that their QoE expectations can be met. Analytics can also reveal the result of congestion management – what delay is introduced due to congestion management, what applications are negatively affected (QoE), and which users are being impacted by congestion management. This is key in countries where regulatory systems will require this information to justify congestion management techniques.
IPE is also a great enabler of creating innovative usage-based charging e.g. Personalised Charging. By using detailed analytics to understand the applications and content that are driving usage, broadband operators can offer service plans that reflect the usage consumption and the application mix of consumers. Offering usage plans that are appealing to consumers will enable operators to shift service plans from unlimited (which many are struggling to do) to consumption-based models. This cannot be done with existing infrastructure equipment, as these systems do not have the intelligence to perform analytics or enforce services based on applications or content on a per subscriber basis.
The challenges of personalised charging or charging based on applications have generated much discussion. The key challenge in any type of service like this is the proper setting of expectations with the customer. There have been a few examples to date of mobile operators attempting to offer prioritised services (mainly gaming) or “enablement” services (i.e. Pay US$5 to use Skype on the mobile broadband), but application or even site-based charging will require even more techniques. Many GGSNs, CMTSs, or even BRASs in the past have offered limited zero-rating capabilities for ringtone downloads, system updates, etc. based on IP addresses or a small number of URLs. There are many cases of IPE systems being used for charging today, however many of them are volume-based or time-based charging to help existing charging systems, or have been installed for future application-based charging use cases.
On the other hand, ’application-based‘ charging or even “site-based” charging in today’s internet environment brings all new challenges to the table. The biggest challenge is that the user experience on a site is very different than it used to be. Facebook, for example, is now also a container for YouTube videos, ads, games, and chat. Or YouTube can serve up videos in formats other than flash video for Apple devices. To determine this, the operator needs to understand the user experience, and will need to use tools from an IPE to ensure that the user experience is matched through the proper application of signatures, policies, and charging models.
Operators looking to offer new charging models need tools to ensure that the service that they offer matches the user expectations for a site. Operators need the ability to look at a customer’s connectivity in real-time to troubleshoot billing discrepancies, much less to test service ideas in their labs. They need systems that can use properties of applications or sites – for example, the ability to key off the referrer field in http. They need to be able to log user connections to specific sites to have forensic details for value-based or zero-rated charging. They need quickly updated signatures when new versions of applications come out, and the ability to retroactively zero-rate traffic (unknown or incorrectly identified) if a signature changes and is identified as incorrect.
Beyond the features required, another huge challenge for personalised charging looms – signalling load. This is potentially the biggest scaling challenge and it needs to be architected carefully from the beginning. When a user initially accesses the network, the network must be provisioned with the correct service plan for that user, which can require interaction between multiple systems for each service that the customer has purchased. If architected correctly, signalling load can be minimised and only refreshed as necessary, but this can also operate on an on-demand basis, which would create transactions every time a new piece of content is accessed. The IPE, PCRF, and OCS/OFCS systems need to support tens of thousands of sessions per second in a mobile network of any size. Any less will result in severe limitations for operators who want to offer mass market personalised charging plans. On the PCRF/OCS side, a new class of product (diameter routers) is designed to allow network signalling to scale. The IPE solutions will need to scale independently, since they are doing both the heavy lifting for charging as well as the reporting.
Personalised charging is one of the most intriguing new services for broadband operators, but will require a system that can adapt rapidly to service changes and can be adjusted in real-time by the operator. It will need to scale on multiple fronts without losing performance or functionality, and certainly without losing accuracy – both for billing and for application identification.