This article discusses the features of a decentralized IoT edge computing ecosystem. You will find a breakdown of the components to be designed, from the IoT gateway to the cloud, as well as the typical benefits to address.
We often hear of an ongoing shift towards edge computing as a paradigm complementary to the cloud. Even more, edge computing has established itself as one of the ongoing IoT trends for 2020. According to recent estimates, the global edge computing market has been projected to grow to $9.0 billion by 2024.
As devices get to process increasing volumes of data, there is a need for decentralization. Within the context of the edge computing paradigm, decentralization involves a multitude of local computing devices or “cloudlets” located close to the data source.
The rationale? Moving closer to the data source has the advantage of reducing server load, decreasing network traffic towards the cloud, and minimizing response times towards the end user.
In what follows, we take a closer look at these developments.
How do you decentralize?
Let’s start with centralization. A recent article by Thingforward.io defines centralization as a scenario where all clients of an application connect to a central node. Centralization has been the main paradigm over the past decades.
Massive-scale data centers, also known as “the cloud”, have been created to execute user requests. Here the data flows directly towards these data centers. There, it is processed, clearing the path from raw data to insights.
There are several downsides to this approach. These include a higher degree of dependency on network availability as well as slow data loads and limited data transfer. Streaming data directly to the cloud may also involve potential data privacy issues.
What is decentralization?
In decentralized applications, on the other hand, the clients are not fully dependent on a single node or an endpoint. Decentralization is “a shift from concentrated to distributed modes of production and consumption of goods and services.” Decentralized applications make it possible to spread control, access, or ownership across several nodes within a network. Clients can connect to any of these nodes.
The cloud is thus dispersed into multiple small-scale computing devices called “computation spots”. Because the data is distributed across multiple nodes, it is less likely that any of the individual endpoints can impact the entire system.
Decentralization and IoT
In the IoT example, you have individual devices, i.e. things with sensors, connecting to gateway nodes. Typically, these individual devices are not connected to the internet. They are limited in their field of interaction and are part of constrained networks.
Gateway nodes, have transmission control protocols/internet protocols and can speak to backend services. In this scenario, we have multiple local devices talking to their edge gateways. And then, we have a number of decentralized gateways talking to the cloud.
This is called “edge computing” because you have a multitude of such IoT gateways dispersed at the edge of the IoT network. The individual gateways do not communicate with each other. They serve as gathering points for the content coming from their local IoT devices. This content is collected, pre-processed, and selectively handed over to the cloud for in-depth analyses. In the cloud, the locally produced and pre-aggregated data is combined with data from other gateways, allowing for various Big Data operations.
The IoT edge computing layers
IoT edge architectures, therefore, typically consist of three main layers:
Sensor and actuator layer
This is the level of constrained devices (e.g. sensors). These devices fulfill basic functions such as collecting data and transmitting it to the next layer or reacting to it, e.g. changing a system parameter. They cannot perform computational or storage tasks and only participate in the communication structure as front-end devices with a forwarding function. So they only gather impressions from their surroundings and transmit these impressions in the form of raw sensor data. At this level, the devices have limited battery life and scarce hardware resources.
Gateway layer
At this level, we have IoT devices that are connected to the internet and have capacities similar to those of mini-servers. At the gateway level, you can filter the IoT data, pre-aggregate, perform basic analytics, or deploy machine learning models. The gateways allow for hands-free operations distributed across multiple devices of the sensing layer. In this IoT architecture scenario, the data pre-processed at the gateway level may or may not be forwarded to the cloud. Depending on the use case, some of the data may only be processed locally.
Cloud layer
Some of the content coming from the edge gateways eventually reaches the cloud backends. However, this is only the content you need to perform advanced analytics, train machine learning models (e.g. for distributed AI applications), or utilize in combination with data from other gateways to gain specific insights. As Asha Keddy, Intel’s Vice President of Next Generation and Standards, was quoted saying in a recent article,
“If you keep putting everything back to the cloud, nothing would ever work. … You don’t have enough time for it to go all the way to the core.”
Because of these developments, there is an ongoing trend in cloud computing providers to extend services from the cloud to the intelligent edge.
A recent article on hackernoon.com proposed that “intelligent infrastructures” is a more suitable name for such IoT edge scenarios:
“…intelligent infrastructure will enable software-based intelligence to permeate the physical world, enabling real-time optimization and orchestration of connected ‘things’ … at a system level.”
According to the State of the Edge 2020 report, infrastructures to support edge computing solutions are yet “nascent”. Therefore, companies will have to implement their own custom edge installations until the technology becomes more mature. The projection, however, is that “the cloud will drift to the edge” as edge computing will increasingly become a standard solution.
Towards low latency and quicker response times
Decentralized IoT edge computing improves speed and flexibility and is particularly suited for tasks that require short response times and the pre-processing of large volumes of data. Below is a short breakdown of the typical benefits of decentralization:
Higher speed. To many companies, speed is essential when it comes to the provision of data-driven services. Because data processing takes place locally, edge computing leads to a substantial decrease in latency. In reducing latency and achieving high bandwidth, edge computing brings in the benefit of increased network performance. You also avoid bottlenecks caused by routing the data through local network connections before getting to the cloud.
More security. Because edge computing distributes processing and storage across an array of connected devices (and data centers), it becomes less likely that one IoT security breach will jeopardize the entire network. Thanks to decentralization, it is possible to isolate the affected parts of the network without compromising the whole.
More flexibility. Companies can incorporate new IIoT devices into their edge network without significantly altering their IT infrastructure. Nor does expanding the IoT edge entail building new data centers. Also, you can roll out an IoT application directly on edge devices.
More robustness. In the event of system downtimes or data center failure, the data coming from the IoT edge device can be rerouted. In this way, it remains accessible to the end-users. Network issues become less relevant as the IoT device processes data locally.
IoT development at Record Evolution
At Record Evolution, we have developed an IoT development platform optimized for edge applications. Considering the various IoT deployment challenges posed by heterogeneous IoT landscapes, the platform solution is fully integratable into the existing organizational setting and is seamlessly adjustable to early-stage edge infrastructure implementation design decisions. Find out more in the whitepaper “The IoT Development Cycle: Live Edge Coding and Device Management”.
Get in touch if planning to test these services within the framework of your own industrial IoT initiative or the development of your own IoT solution.
About Record Evolution
We are a data science and IoT team based in Frankfurt, Germany, that helps companies of all sizes innovate at scale. That’s why we’ve developed an easy-to-use industrial IoT platform that enables fast development cycles and allows everyone to benefit from the possibilities of IoT and AI.