A significant trend in the Network Policy Control industry is that network intelligence devices are moving increasingly close to the subscriber access edge.  This trend is true both in fixed access networks, where elements such as our own Policy Traffic Switch are deployed alongside the aggregation devices (BRAS in DSL and CMTS in Cable), and in the mobile space, where devices are moving ever-closer to the base stations (NodeB or eNodeB).

Being at the edge has many benefits, including more efficient and accurate subscriber awareness, measurement, and management of network traffic.  Ultimately, these characteristics mean that Communications Service Providers (CSPs) gain more precise control over network traffic, which lets them manage congestion precisely, charge accurately, and deliver differentiated subscriber services.

Focusing on mobile networks specifically, we’ve invested in the technology to allow our network policy control solutions to be deployed in the packet core network, for instance between the SGSN and GGSN, which yields:

• More reliable subscriber profile and session information via GTP-C than conventional methods that acquire this information from RADIUS accounting
• Policy control of all packets, including localized handset-to-handset traffic
• Accurate per-subscriber reporting and policy enforcement, even if the GGSN is configured to provide NAT functionality
• Control and reporting of inbound and outbound roaming traffic, which is critical for accurate bill shock warnings and deployment of innovative services (like unlimited day passes for roaming subscribers)

The “to the edge” trend comes at the same time as 100GE is beginning to emerge in the interconnects of networks around the globe.

It bears noting that the forces driving 100GE (primarily economic) are not the same as those that drove the adoption of 10GE (primarily the need to address Internet growth challenges).  Service providers, in particular, felt the Internet’s growing pains, so 10GE was of major interest; in fact, Sandvine launched and deployed the industry’s first 10GE policy control solution.

By contrast, the motivators and use cases for 100GE are very different, as they focus on economic reasons like cheap wholesale, Internet peering and datacenter interconnection.  There is no imminent demand for 100Gbps of Internet bandwidth at the edge of the network. In other words, 100GE is a technology that applies to use cases that are in the network core, rather than at the edge.  Since the motivators for 100GE are primarily economic, to determine the merits of investing in such technology it makes sense to evaluate the usefulness of a 100GE network policy control device from an economic standpoint.

Let’s suppose such a system is deployed at the network’s transit links, to address the transit traffic management use case (note that it would not be able to address any business intelligence use case since CDNs exist below the transit links, so any business intelligence use case would require a deployment that can intersect the CDN traffic).  Using simple, but by no means unreasonable, assumptions, a carrier with 100 Gbps of transit will save $1k-$10k per month for a capital investment of ~$500k for a redundant system deployment.  Excluding operational expenses (which are definitely not trivial, and will likely run in the ~$100k per year range), this means that a carrier can expect to see a return on investment within 60 to 600 months.  I suspect few carriers would invest in such a use case, with the possible exception of extremely remote islands with enormous bandwidth requirements (of which there are none).

Now, one could examine such a 100GE system in a core deployment using real-world examples, such as South Korea (since it has the highest penetration of LTE on the planet) and AT&T in the USA.  Even carriers of such magnitude, with aggressive deployments and bandwidth-hungry subscribers, experience approximately 20-30Gbps of total mobile traffic within their network given the dynamic nature of mobile usage with even large GGSNs handling only up to 4Gbps of bandwidth.  In an edge deployment, the bandwidth needs are even less, so this 100GE unit would be running at ~4% capacity if deployed access-side of a GGSN, or at 20%-30% when deployed highly-centralised in the network.  Again, this is impractical, and does not deliver a positive return on investment for the service provider.

We can therefore conclude that a 100GE network policy control system not only fails to address any compelling service provider use cases for Network Policy Control, but also fails to deliver an acceptable return on investment for the network operator.  When it comes to effective network and service management, it is imperative to choose the network intelligence device that can deliver the most functionality at the best price-point to meet the current and future needs of the CSP.