Here’s how to build extremely flexible 5G core networks
This blog is by Volker Held, head of Innovation Marketing at Nokia Networks.
As a quick refresh, here’s what we see as the 3 basic drivers for 5G:
* Existing consumer demand for video and immersive multimedia will continue to grow
* Internet of Things (IoT) will require 5G to connect everything from robots to jet engine diagnostics which will produce vast amounts of data and therefore require exceptionally wide bandwidths
* Super low latency in the networks will be needed to enable virtual reality, control of robots and autonomous driving so that there’s no discernible delay
Not only will the demand be ad hoc and on-demand – but the traffic is anticipated to grow by a factor of 10,000 between 2010 and 2030, flanked by the need to support billions of sensors and machines. On top of that, we cannot say for sure what kind of traffic will run over the networks in the future. So how are we going to build extremely flexible, cost-effective 5G core networks to meet all these diverse requirements?
New Proof of Concept paves the way for 5G core networks
Nokia and Docomo EuroLabs have spent the last three years jointly analyzing the potential of applying cloud computing and SDN to mobile networks. Special focus was given to the packet core’s gateways as those nodes provide some unique challenges: specificially, the control and data planes are closely mingled in one physical node. State-of-the-art virtualization principles call for both the control plane and the data plane to be located in the data center cloud. In the future, this would mean that the data centers would be hit by massive 5G data traffic.
To come up with a sustainable solution for 5G, the research project opted for an alternative design. The concept focuses on running the control plane as an application in the cloud, while simultaneously running the data plane, with its massive traffic volumes, in decentralized and distributed transport network nodes such as aggregation switches. Typical gateway functions like GTP tunneling or policy enforcement are realized by employing OpenFlow based SDN. An SDN controller resides in the cloud and acts as the “glue” between the data center based control application and the SDN based transport nodes. This is an elegant way to steer the data traffic away from the data centers.
This concept was applied to a use case requiring the utmost flexibility – mega events. Large concerts and football games demand significant gateway and transport resources, but only for a limited period of time. By using the above design principles, it is possible to assign additional transport capacity and additional gateway control plane applications without over-provisioning, all on demand and within minutes. Once the event is over, the whole network immediately reverts to its original configuration, again within minutes. This use case was also shown as a live demonstration at MWC 2014.
Eye-opening Total Cost of Ownership (TCO) analysis
Next up was a TCO analysis. Which solution makes economic sense? 5G Nokia and Docomo EuroLabs compared the all-in-data center option, with both control and data planes in the data centers – and the split approach, with control in the datacenter cloud and the data plane located in the SDN-controlled transport layer. The study was done for topology and traffic patterns that would be typical for a country like Japan with appropriate extrapolations for 2020.
Costs were 4-5 times higher for the all-in-data center approach compared to virtualization, which keeps the data plane in the transport network. And this imbalance multiplies with increasing traffic. For example, depending on their locations, traffic has to be conveyed to and from the data centers, thus creating up to 4 times the traffic load on the transport network for the all-in-data center approach. Neither geographical distribution nor variations in the traffic mix helped to decrease this disproportionate impact.
Conclusion: The findings clearly showed that placing the entire packet core into the cloud is not a sustainable option to meet the massive traffic volumes of the 5G era. Instead, a cost-effective combination of SDN and cloud computing is the answer. Furthermore, the low-latency use cases like connected car and remote control of robots won’t allow for traffic to be conveyed to central data centers given the propagation delay.
Would you like to experience our demonstrations? Please contact Nokia Networks Solution Experience Center.
Please share your thoughts on this topic by replying below – and join the discussion with @nokianetworks on Twitter using #5G #Innovation #FutureWorks #IoT #NetworksPerform #Cloud.