Over the last decade, I’ve attended a lot of conferences dedicated to topics around the smart city and studied many cities to determine how they can best realize their smart city goals. In meetings with urban planners and city managers, I’ve seen an evolution from pursuing point solutions, such as smart parking or smart lighting, to a realization that they need a more holistic solution. This inevitably leads to a question about the network that connects everything to everything else: what is the best network fabric for the smart city?
Part of what is driving municipal managers is the cost of replicating the computing and networking resources needed to support each point solution. To overcome this inefficient approach, a programmable and automated cloud infrastructure is needed, forsaking the traditional proprietary and closed “bare metal” approach. Then there is the need to integrate the data between applications securely. This is especially true when cities get into deploying IoT sensors everywhere. Many of these devices can produce data that can be shared between many different entities to satisfy various use cases. But to do that efficiently and securely with their existing WAN network is not scalable and operationally cost-prohibitive.
The problem with traditional WAN architectures is that they assume the things they are connecting are relatively fixed. Once a connection is “nailed up” (the language is revealing), it is assumed that it is for a very long time. Security usually requires securing that one, static connection, once only with a traditional IP/MPLS VPN.
Today’s smart city applications require something much more dynamic for two reasons. First, the numbers of IoT sensors that need to be deployed will operationally overwhelm the traditional “nailing up” approach. Second, the ideal smart city cloud, through which IoT sensors communicate, is not just run from a single data center location. Today’s cloud-based applications are becoming modularized and are delivered as a set of logical constructs known as microservices. Cloud operations occur at various locations including the edge of the network, in government branches, health centers and public clouds. This allows for auto-scaling and dynamic workload placement, adapting to the ebb and flow of city life.
There are many factors driving this distribution of cloud functions, a key one being low latency for optimized performance. Imagine a smart road or highway with roadside sensors picking up audio, video, temperature and other data with which it controls smart lighting, signaling and signage, all the while coordinating with autonomous vehicles. It is far more efficient for the network to process the huge amounts of data locally than centrally, and it ensures that the response times (latency) are fast enough to actually improve over drivers’ reaction times.
Now imagine that the weather suddenly deteriorates and it is rush hour. The computing resources required to handle the sudden increase in load mean that the local resources have to scale up to handle the local incident.
This dynamic scaling of resources is perfectly do-able with distributed clouds because the processing is done by virtual machines that can be created instantly. But it means that the location in the network where things happen can shift almost instantly. How do the sensors know where to send their data for processing at any given time?
In a traditional WAN, this is an insoluble problem to address in these time scales. But in data centers, it is old hat. It is the main reason why software-defined networks (SDN) were first deployed for data centers. The SDN controller oversees all of the SDN nodes and connections on the data center network. If an ongoing process requires virtual machines (VMs) to be spun up to help handle the load, the SDN controller is instantly notified by the cloud management system. It then adds them to the forwarding tables and updates the underlying connectivity so that each new VM is reachable for the corresponding applications. The SDN nodes don’t need to know about VM changes, the SDN controller is on the job and instructs the SDN node forwarding plane what to do. The traditional WAN lacks this kind of self-adaptive capability and elegance as both the control and forwarding planes are combined in each node.
Most municipal governments are not in a position, however, to completely refashion their existing data centers and WANs as SDNs. Fortunately, there is a simple and elegant solution: an SDN can be deployed as an overlay onto any physical network. This is called SD-WAN, and it is the ideal network fabric for the smart city.
An SD-WAN is surprisingly simple to create. Not much has to be changed about the existing WAN — SD-WAN is designed to be WAN-transport agnostic. SD-WAN gateways are simply inserted at every connection point across the city: for instance, a gateway for each data center function, whether at the edge of the network or centrally, and another gateway connecting IoT sensors installed in the city infrastructure. These gateways “talk” to the SD-WAN controller (which is also coordinating with the SDN controllers in each data center).
The SD-WAN controller communicates with all gateways to instruct them how to build the overlay networking paths across the physical WAN network. When WAN connectivity is not available, it can resort to other means such as commercial cellular data service or the internet. In addition, a policy engine in the controller distributes policies so that the gateway can ensure applications receive the performance and security they require. The security policies can be incredibly granular and specific to the application. They can also be created instantaneously as VMs are spun up and down with the ebb and flow of the city.
The SDN management platform is equipped with an open, northbound application programming interface (API) that enables, for instance, an IoT application manager to specifically manage the operation of that part of the network, and only that part of the network, connecting the IoT sensors. As the city installs new sensors, the API can even enable operational staff to automate new connection setup, significantly speeding up deployment.
As the world’s aging population becomes increasingly urbanized and sustainability issues demand dramatically improved efficiencies, the smart city needs a high performance and dynamic network to power the next-generation of hyper-connected smart city applications. SD-WAN is the network fabric that smart cities need to reach their full potential for their citizens.
Want to know more?
Read the white paper for an in-depth view of how to build a high-performance, cloud-aware network for a smarter, safer and more sustainable city.
Visit our smart city page to learn how our solutions can help turn your city into platform for creating innovative digital services.
SCEWC 2019 offers a unique opportunity to explore how technologies, such as 5G, IOT and machine learning, can help you build an intelligent, integrated platform for true city innovation. So be sure to stop by Nokia Booth B249 and talk to our experts about your community’s specific needs — and visit our speaker’s presentation and demos. You’ll learn about our innovative ‘city as a platform’ approach to scale your city digital development.
Share your thoughts on this topic by joining the Twitter discussion with @nokiaindustries using #smartcity #cloud #IoT #SDN #SD-WAN