A Practical Guide to Building an Edge Powered IoT Architecture

The internet of things (IoT) promises to revolutionize use cases such as connected cars, smart cities and industrial IoT (IIoT) for manufacturing and edge analytics. However, data management can present a major hurdle when building the digital infrastructure to support applications like these. In this blog, we will share the approach we used to develop a digital edge reference IoT architecture based on our experience deploying IBX SmartView^TM, Equinix’s data center infrastructure management (DCIM) software as a service (SaaS). IBX SmartView provides real-time access to environmental and operating information relevant to our customers’ Equinix International Business Exchange™ (IBX^®) data center footprint on Platform Equinix^®.

The scale and distributed architecture of IBX SmartView

To monitor and analyze data center device IoT data at Equinix, we developed a DCIM software platform that powers the Equinix IBX SmartView software service. The DCIM system has data center edge (DCIM edge) software components deployed at more than 131 Equinix IBX data center locations. Each DCIM edge collects and processes data from hundreds or thousands of heterogeneous devices supporting each data center. Devices include electrical and mechanical infrastructure assets, environmental sensors, building management systems (BMSs), power meters and more. Every day, the Equinix DCIM IoT system processes a high velocity of hundreds of million messages and manages a high volume of hundreds of terabytes (TB) of data.

The heart of the architecture is distributed IoT workload processing between all DCIM edges and the centralized core DCIM platform. In this architecture, we separated two types of IoT use cases between the edge and the core:

• Edge IoT scenarios that require high-performance, low-latency data access, scalability, data filtering (noise reduction), device specific analytics and multi-cloud data routing.
• Core IoT use cases that require IoT data aggregation into data-lake, data-science and analytics on aggregated data and business applications.

Reference architecture for distributed IoT processing

The following diagram illustrates the reference architecture we developed for modern, edge-powered IoT applications that need to operate at scale based on our learnings from the IBX SmartView DCIM deployment. This type of architecture can easily be applied to other IoT use cases mentioned above such as connected cars.

Edge-Powered Multicloud IoT Reference Architecture

The steps we took to build this reference architecture are as follows:

1. Edge components: The device-centric capabilities of IoT, such as data collection, device state management, filtering, noise reduction, store-forward, and real-time analytics, need to be deployed across edge nodes. The edge software components can be built using open source Java frameworks such as Akka Spring Boot, Go frameworks such as EdgeXFoundary for optimized resource deployments, and Redis as a store and forwarded cache. For workload orchestration and management, use Kubernetes combined with cloud-based data streaming, data lake, and IoT platform as a service (PaaS).
2. Network traffic optimization: Leverage Equinix Fabric^TM for split and fork data flow between edge and multicloud points to ensure private secure low-latency network connectivity. We recommend noise filtering to reduce network traffic between the IoT edge and core. The device timeseries events can be filtered out based on techniques, such as Change of Value (COV) threshold filtering, sampling, outlier detection and/or error frequency, to reduce event flow and network traffic to core cloud services.
3. Distributed analytics: Latency sensitive, high performance analytics and machine learning workloads can run at the edge, while cloud services should be used to build the core data lake and aggregated data analytics. Cloud native services, as well as open-source managed services, can be harnessed to implement the data lake and analytics.
4. IoT security: Enable security through Https, Mutual TLS, Oauth2 and JWT tokens to ensure security in connectivity and data at motion. Leverage encryption to enable security for data at rest. Also cater to integrate IAM for RBAC and ABAC.
5. Distributed multi-tier data storage: Apply distributed data stores at the edge or cloud based on data processing latency, query latency and data volume.

Benefits of deploying edge-powered IoT architecture on Platform Equinix

Building this edge architecture on Platform Equinix provided the following advantages:

• Distributed data center edge locations enable low-latency IoT workloads. Businesses can use Equinix colocation services or Equinix Metal^TM to deploy their IoT edge software components. To explore all the edge services available on Platform Equinix, check out Equinix Edge services and Equinix Marketplace.
• Use cases that require data orchestration between various cloud services, including data routing, split and fork mechanisms, or multicloud can easily be addressed via Equinix Fabric software-defined interconnection.
• IoT data center edge-to-cloud traffic can be reduced via noise filtering, change of value, sampling etc. In our IBX SmartView deployment, we observed a reduction of more than 30% of traffic to the core based on IoT change of value-based filtering rules.
• Fast and low-latency response through edge data processing and edge analytics.
• Edge-based security avoids direct device exposure to cloud services.
• Public cloud capabilities such as IoT PaaS, data science and core analytics are easily accessible from Platform Equinix.

To learn how to build a globally scalable IoT solution by leveraging a common digital platform, download the IoT Digital Edge Playbook and IoT Blueprint.

You can also learn more about IBX SmartView and register for a free trial.