We can assume that no service provider will be ready to buy and deploy core network as it is today for the one simple reason — nowadays core networks are monolithic and centric while most OTT cloud infrastructure is distributed and build by using micro-services. Using the old fashion technology simply does not making any sense. Despite the fact that there are several vEPC products in the market, none of them has reached a production stage; most of installations are in labs. Looking forward vEPC will never replace the current EPC install base — the market is already looking for 5G technology which will introduce a new conceptual network design. But what this new design is about?
During 2015 several service providers published their own view on 5G core architecture which should support 5G requirements. The general consensus is that 5G should be a full redesign of old concepts and no one of 4G components should be reused.
AT&T in Nov 2015 5G event at Santa Clara provided their view:
“5G CN should be a clean slate solution and NOT an evolution of EPC
- A new 5G System and CN should fulfill our needs of RAN-CN decoupling and access neutral CN
- 5G CN should not carry legacy baggage from EPC. It should be based on access independent features like access neutral QoS model, AAA, mobility and policy framework
- Reuse of EPC concepts for 5G radio will impose legacy system onto 5G radio, making 5G
UEs and base stations dependent on EPC
- Developing a new 5G CN will have impact on RAN-CN interface and hence for any deployable 5G system this interface needs to be specified”
SKT published a detailed white paper about 5G architecture design and implementation guidelines in October 2015 and main message was:
- 5G is user-centric, as opposed to being network centric in the past
- Disruptive in nature, integrating different industries
- Core network needs to be re-designed and optimized (e.g., remove any potential traffic bottlenecks)
- Due to EPC network nodes are currently structured in a hierarchical manner, all user Traffic traverse through a series of serving gateways (S-GWs) and packet data network gateways (P-GWs), CN should become flatter and distributed
As could be seen above, all service providers recognize 4G limitations due to its centric and monolithic nature.
NEP recognizes this fact too, but the problem is the wasted capital which has been invested over the years in 2G/3G and 4G (build as patch). The network redesign (again) is a huge thing for network vendors.
When looking at recent activities in the industry the major focus is in 5G access components such as cloud RAN which is going to replace the compute unit attached to the wireless antenna and move the base band components to remote datacenters. Mobility and EU performance were mentioned many times by multiple vendors but no one of them is addressing the core network which supposed to support them. MassiveMIMO, CloudRAN, mmWave and SCMA are major topics and a lot of research has been done there.
The challenge could be also seen as an opportunity for both Telcos and service providers as they are straggling and fighting the same battle where creativity and innovation can be a game changer. It is especially true whilst Facebook and Google are planning to launch their own 5G network solution.
The question is how this solution will look like? One of initiatives for network evolution was to design a more flat network, as it is essential for increasing performance and reduction of involved components (remember every component in the old world is a weak link). So a flatter network is a network which consists of 2 layers — control layer and data layer.
The control layer should be responsible for system orchestration and global name resolution which uses a key value store for a user’s device authentication. The ID should be unique for all devices in the world. As an example, IP6 can provide is 2¹²⁸ unique ids. This will help to identify, authenticate and re-authenticate the device in case of dissection. Basically it should replace the GTP tunnel that responsible for EU connectivity. One of possible examples for key value store is Redis which has been used in many Telco systems but there are many other options.
The data layer is responsible for packet processing and routing. The new distributive architecture will replace all centralized components. The distributed components will be placed and managed in many different regions of the network. This architecture will allow the following:
- Launching Microservices to support routing
- Launching containers to support the Microservices
- A key value to authenticate the EU
- Routing based on the calculated next hop instead or predefine tunnel
- BGP as a protocol to discover the best route to support the next hop routing
- Full life cycle management for microservices
- Running containers on bare metal for performance
- Design for scale
All this will allow pure IP packet flow across the core without the expensive and dedicated EPC components. If this concept for you is science fiction try to think how Netflix has been designed. You will see a lot of similarities: Netflix uses Microservices and serve 3B services requests every day. The internal communication (between Netflix internal components) is 3 times bigger than external communication.
For Netflix It’s absolutely clear that if one will keep using development processes and application architectures which worked just fine ten years ago, this simply will not provide a required performance to capture the demand. The idea is very similar for the next generation of cellular networks (e.g. 5G).
In this case new players which are already familiar with running a web-scale system environment (with a great success) will develop and introduce their own core. What is the effort required from Google to create its own core and rent the access from an existing cellular providers? What will prevent us as google users (everyone is using Google) to use Goggle cellular services?
The threat leads to an opportunity which is in hands of the service providers and the network providers. The former should innovate and the posterior should accept the innovation. NFV is a good start but just virtualization of Telco apps will not close the gap between OTT and Telcos. This and other 5G innovations could remove from Telcos the “dumb pipe” label