Packet core

The Service-Oriented Mobile Core

The industry is making decent progress on virtual mobile core networks. At Light Reading's Big Telecom Event last month, AT&T's SVP Architecture & Design, Andre Fuetsch, said the operator is now running live traffic over a virtual EPC. And in a recent Heavy Reading operator survey, 70% of respondents said they expect to move to live deployment within three years. (See Virtual Mobile Core: Time to Get Real.)

To identify what needs to happen to move from proofs-of-concept to live services, earlier this year I wrote a white paper on Commercializing Virtual EPC at Scale.

One critical point -- and one we've consistently argued -- is that to unlock the virtual core opportunity, mobile operators need new service models, with associated revenue, to drive the business case. These new models will, inevitably, drive new network architectures -- perhaps not radically different at first, but evolving in phases in tandem with software-centric networking technologies.

The idea of the service-oriented core network is central. By creating a "workflow" or virtual "network slice" configured in software, dynamically, according to the needs of the application, or user group, the operator can create a core network instance that is optimized for that service type. This virtual core network instance can be private and secure, and can make use of important LTE features, such as prioritization and QoS.

The most common use cases today are machine-to-machine (M2M) services, which have particular traffic profiles that drive different core configurations from mass-market consumer broadband. Other substantial opportunities include virtual private enterprise networks, public safety, or mobile virtual network operators (MVNOs). You could also include "service chaining" in the Gi-LAN.

The idea is to build a software-defined core dedicated to these users/functions/services and deploy it in parallel to the main backbone EPC. This is a relatively low-risk way to introduce NFV and prepares the ground for a more fundamental core network transformation at a later stage. And it comes with a business case!

One area that needs attention to really make this model fly is traffic separation in 3GPP mobile networks. There are various techniques available for this, but they tend to be either coarse-grained, such as using PLMN (public land mobile network) identities, or are micro-techniques embedded in the existing core, such as APNs (Access Point Names) and bearers. To enable dedicated, customer-specific core networks, a third way is needed.

An initiative to address this is the 3GPP work Architecture Enhancements for Dedicated Core Networks, known as DECOR (see TR 23.707). The aim is to specify how users connect to dedicated cores in RANs without using PLMN-IDs, and without requiring handset changes.

It is not directly linked to virtualization or NFV, and is not dependent on them, but in practice I expect there to be a strong link. DECOR is part of the Release 13 specifications, due early next year, and so conceivably could be in use, in some form, by the end of 2016. There is strong demand for this use case from operators, so the chances of progress are good.

Finally, the service-oriented mobile core is also linked to innovation in the RAN and is an important concept in the Evolved 4G network, and in the 5G system architecture currently under development. How this transition to 5G occurs is a topic for another article and a lot more work. But for now, one thing we do know is that operators will need radical thinking, matched with real-world pragmatism, to make it a success.

— Gabriel Brown, Senior Analyst, Heavy Reading

Sign In