Why an Edge-Cloud for a 5G Urban Network?

In October 2020 we started working on our first Smart City deployment using Cell-Stack. We partnered with Vivacity Labs and Transport for Greater Manchester to build a private 5G network that will reduce the installation costs of future connected infrastructure such as the Smart Junctions sensors, in the city of Salford (Greater Manchester), and at the same time provide a platform for innovation where other Smart City applications can be tested.

Before Weaver Labs was involved in this project, Transport for Greater Manchester (TfGM) and Vivacity had been working together deploying sensors to improve the traffic signal management systems. Vivacity’s Smart Junctions product uses reinforcement learning to adaptively control the traffic system. It has already demonstrated a 23% reduction in journey time across a single junction and can adapt to different policy objectives, such as reduction of vehicle congestion, prioritisation of pedestrians and cyclists, reduction of pollution, etc.

As with any Smart City application, Smart Junctions needs a network. Cities need dedicated connectivity to provide resilient future city management services. In this case the network must respond fast: the system must send input data to the cloud and receive actions as a result of some algorithm computation — the network must be reliable and send short streams of data in a very short period of time. When deploying these sensors, there are different options to connect them to the Internet:

1. Public 4G or WiFi Networks: connect the sensors using the available coverage that exists in the cities. It has been noted in past deployments that the network becomes less reliable when we all get connected at the same time — so public 4G or WiFi it’s not adequate for having a mission critical application such as controlling a traffic light system.
2. Use wired connectivity to connect the sensors to an Internet connection in the controller cabinet (i.e., that little box in the junction where all the equipment is kept): this solution works well but it’s costly to build, and can’t be used for anything else (because it’s a cable). It also limits the public sector’s capability to innovate and trial any other future solution.

With the team at Vivacity and TfGM we decided to put our efforts together and build a private 5G network that can be used for the Smart Junctions application, but has the possibility to offer connectivity for other applications in the future.

What is a private 5G Network? It’s simply a 5G Network, same as the one we use with our phones, but it runs over a shared spectrum — which is much easier to access and anyone can apply for it. This means that while using a shared spectrum band, anyone can own a 5G Network, similar to WiFi. Thi private 5G network sets the foundations to reduce cost and create wider opportunities to innovate.

The Challenge: a cheap network that can grow fast

We’re creating a “Connected Corridor” to cover a total of 10 junctions that run along 1.5Kms in the area of Salford, Greater Manchester, UK.

However, getting this Connected Corridor wasn’t an easy design task, we had very important requirements we needed to comply with. The reason? Because we wanted to prove that a Smart City network can be scalable and affordable:

1. This Network had to be strictly separated from the consumer traffic (aka, people like you and me streaming youtube using our phones), to make sure it’s available when the sensors need it.
2. This Network had to be affordable and easy to maintain, presenting similar deployment and maintenance costs as that of a WiFi network. With this, we aim at reducing the gap of adoption by using a commoditised approach to connectivity.
3. We wanted to leverage the existing public infrastructure, such as the existing fibre connectivity, the traffic control cabinets to house equipment (i.e., servers), and the traffic lights to mount the small cell antennas. With this design principle we show that existing street assets can be used very easily to scale networks in Smart Cities, and there is no need to rent expensive data centres or large computing platforms. The network can grow by adding more low-cost nodes in the street cabinets such as these ones.
4. This Network has to be able to serve more than one application, with separation of services and that enables a commercial route for public sector owned infrastructure, creating a sustainable business model for local authorities investing in infrastructure.

An Edge-Based 5G Urban Network

In order to successfully deliver a 5G Private Network that complies with the requirements we had, there was little or no alternative than to consider a cloud-native edge network. That’s a lot of key-words together, but essentially what it means is that we have low-compute power servers distributed along the Connected Corridor which are controlled via a Cloud environment.

We’ll elaborate a bit more on the How, and the Why:

1. To be a cost-effective network (i.e., low operational and capital cost) it has to be a Cloud-based network running on commercial off-the shelf computing hardware (COTS). This means we should be able to run the full 5G software out of one server and minimise hardware costs. Why Cloud? Because it will be easy to manage using software based tools, cutting operational costs.
2. The nodes running the 5G software must be placed in the traffic control cabinets, making the network design distributed in nature. This is a very challenging design as every cabinet needs to have access to: fibre connectivity, extra space to install a server, extra power supply and the equipment should support temperatures of up to 70 degrees Celsius.
3. The network must be able to grow by adding more nodes that are easily connected to the existing network. This means the Cloud we create must facilitate the creation of hyper-scalable networks, following a plug and play model. This is achieved with the use of the peer-to-peer network created by Cell-Stack.

Also, we had some hard-core requirements when it comes to the Radio and the Software that supports the 5G Logic:

1. Radios must support the shared access licence bands allocated by Ofcom. Being an outdoor 5G Private Network, the most appropriate band in the UK is the n77b, which there are not too many phones available in the market — in all honesty, we’re still dealing with that problem.
2. The size and weight of the small cell radio must meet the structural requirements of a traffic light, where an extension pole will be used to extend the length of the traffic light. This limits the weight of the radio to up to 9kgs and the size to something similar to a shoebox.
3. The 5G Network software we choose must follow a micro-service architecture and API design such as OpenRAN, whereby disaggregation of components and integration with other radios is feasible.
4. The software provider can be closed source, however it must provide open APIs for network automation, control and service orchestration to comply with the plug and play principles of network scaling and low operational costs.

After much work with Telet Research, our system integrator and Accelleran and Attocore, our chosen software providers, the resulting network design is the following:

We will get into more technical details on how the network is running in a Cloud infrastructure while also being compliant with the very demanding requirements of the Radio: let’s say that the lowest part of the 5G stack doesn’t like virtualisation, therefore makes it very challenging to run this in a cloud. However, we can proudly say, we have the very first OpenRAN Smart City deployment, running on an Edge Cloud.

How does Cell-Stack glue altogether?

Cell-Stack, is a peer-to-peer (P2P) software managing all this infrastructure. Cell-Network is a grouping of infrastructure communicating through Cell-Stack reference software forming a distributed peer-to-peer telecommunications network with no central coordination.

Each junction is a Cell Node in the Network, creating a P2P overlay on top of the 5G Infrastructure. Cell-Stack is in charge of establishing, managing and monitoring the connectivity across the different Cell-Nodes (i.e., peers), using WireMQ an asynchronous messaging system that wraps the protocols underneath and creates channels of communications across all the peers in the network. At a high-level, our architecture looks like this:

Wrapping up

The work we are doing in Manchester is a great representation of the opportunities and future challenges we will face as an industry to streamline connectivity deployments for Smart City applications. The infrastructure can be 5G, WiFi, or any IoT technology — but the challenges remain the same. Networks should be able to scale fast and be cheap to attend the needs of the applications bringing innovation constantly into the market.

Throughout the development of this project and working with partners in this trial, we have refined the design of Cell-Stack while always keeping the same principle: to be able to integrate and aggregate infrastructure so that networks can be consumed easily and scale fast.