IoT on the edge: avoid IoT network overload
Carsten BrinkSchulte, Core Network Dynamics
Had Mark Twain been reporting on the rise of the Internet of Things, he might have commented wryly that ‘the reports of its growth have been greatly exaggerated’, writes Carsten BrinkSchulte, the chief executive of Core Network Dynamics.
Certainly, forecasts of growth vary considerably with Gartner predicting more than 25 billion connected devices by the end of this decade, whilst Cisco believes there will be a whopping 50 billion. Either way, it is no exaggeration that unprecedented numbers of connected devices are coming on stream, deluging the networks with disparate data. The embedded sensors already keeping tabs on electricity meters or tracking shipping containers are just the tip of the IoT iceberg. The problem is that today’s global centralised cellular network architecture was designed to handle millions of smartphones, not billions of IoT endpoints. The projected massive influx of IoT devices will create immense strain on the network and is likely to push it to breaking point.
If we want to avoid cellular network meltdown and help carriers benefit fully from the IoT boom we need to reduce the pressure on the central core. To do this, we need to rethink how mobile networks are designed. A complete overhaul is not only economically untenable, however, but also unthinkable from a techno centric point of view. My approach is that carriers should take advantage of Mobile Edge Computing to enhance and extend their LTE networks.
Mobility moves to the edge
Today’s LTE networks are based on a star topology with a centralised EPC (Evolved Packet Core) with the average base station handling signalling for up to 1,500 devices on average. As future IoT applications emerge, a base station might have to manage several hundreds of thousands of devices, each one sending signals and transmitting data. Handling authentication, signalling and traffic for billions of devices will overload the centralised EPC, resulting in degradation of service quality for Smartphones and triggering large investment requirements for carriers struggling to handle the load.
The Internet offers some interesting lessons for the carrier community, however. It is a perfect example of a decentralised mesh architecture combined with a star-of-stars topology, which is why the growth to billions of endpoints was handled so elegantly. By applying this model to the mobile telecoms world, we can prevent the central core networks from drowning in the flood of data and signals emanating from billions of IoT devices.
Telcos have already started out on their journey to bring more internet technology to the mobile networks with new network technologies like NFV (Network Functions Virtualisation) and SDN (Software Defined Networks) igniting network architecture redesign. With NFV and SDN, carriers no longer need to use expensive dedicated hardware for separate service functions, but can virtualise services using software running on standard computing hardware. As a result, cellular networks will become more flexible and cost-effective.
I expect to see the first significant deployments of NFV/SDN at the edge of the mobile network. Mobile Edge Computing (MEC) is a new concept, which provides an IT service environment and cloud-computing capabilities at the edge of the network, close to or even right at the base station. Based on virtualisation, MEC promotes decentralised core networks providing connectivity and computing at the edge, running applications and even the EPC closer to the devices.
The Industrial Internet of Things
With a decentralized EPC running at the edge of the network, messages between local endpoints can be handled locally, avoiding backhaul traffic to a centralized EPC and radically reducing network latency. This is particularly key when considering the Industrial Internet of Things (IIoT/Industry 4.0), where ultra-fast communication between devices is paramount. Take, for example, a factory where robots, sensors and actuators are connected over a wireless network rather than with cables; controlling a robot over LTE is impossible because the centralised mobile network architecture could have a network latency of up to 100 milliseconds, which means that the robot will simply not react quickly enough to information provided by external sensors. It would be much more effective and timely to have a lightweight distributed EPC running on small cells installed at a factory, short-circuiting communication between robots and external sensors and providing ultra-fast message delivery with latency as low as 1 millisecond.
Hedging bets in a new heterogenous world
And we may have to look even further than simply optimising the network architecture: no single protocol for IoT applications will prevail. We are already seeing several parallel efforts with specialised and optimised non-3GPP protocols, for example LoRa and SigFox. High bandwidth, low latency IoT applications such as live video streaming from a remote camera will require LTE, whilst LoRa and SigFox are better suited for lower bandwidth applications – such as smart water meters. If operators want to play a central role in IoT, they will need to open up to the new heterogeneous world of IoT and embrace true openness.
In summary, I believe that offloading signalling to the edge and keeping local traffic local will become key to the Industrial Internet of Things as well as IoT in general. The solution is Mobile Edge Computing (MEC) with a distributed EPC as a key component, preventing central network and backhaul overload and at the same time dramatically reducing network latency.