An “edge cloud” refers to a middle ground between cloud computing and the edge of a network. This is often not a precise definition because the area between the edge and the cloud is ambiguous. While a cloud embodies core centralized compute and storage resources available via the internet, the edge of a network, sometimes more specifically described as the device edge, is a term that refers to the furthest physical and/or logical aspect from the core—for example, where there are IoT devices.
The edge can be described in other terms, by the burdens it places on a network. Specifically, the edge is an endpoint whereby latency, bandwidth, and network congestion become significant enough to warrant placing compute and storage resources closer to the data producing device rather than rely on backhauling data to the core for processing. Imagine several regional data center substations replacing a single central data center, commonly known as a distributed system.
In this way, the edge cloud is formed (the cloud core in the above example somewhat disappears by edging outward towards the device edge via multiple stations), and instead of a single brain with many very long latent connections, these edge devices now transmit data to the most available station (read closest or least congested).
Edge cloud is a common enterprise practice today, and may simply go by another name, redundant distributed systems. Companies that want to maintain a high level of available services, like streaming, cannot rely on a single central data center to meet demands, and so distribute their cloud content in strategic locations.
Yet, because it serves a purpose similar to the device edge, the edge cloud is different from redundant content delivery systems. The device edge challenges networks by introducing latency, bandwidth, and network congestion issues. Before IoT and the proliferation of devices, data was still generated in relatively low amounts. Today, these devices and other demands have overburdened the internet as a whole. Data transiting the internet is now exponentially more voluminous every year leading to the device edge challenges mentioned. The net effect is slower Internet infrastructure, and greater inefficiency for businesses that rely on the Internet for backhauling their data.
One solution is to avoid the data backhaul. For edge devices, placing storage and compute on the device, or local network, can eliminate the need to “fetch” compute power. For example, a small server in a smart home could collect and process device data, eventually relaying a summary back to the cloud. However, homes with their own servers may be overkill.
But, in smart cities, where the edge is more diffused, edge clouds find a use case. This diffusion has earned edge clouds another name, “fog”. In a fog, the physical edge falls off and the logical edge remains, for smart cities the logical edge may be the city limits, but edge devices can be everywhere. In these cases, maintaining focus on the purpose of edge computing is necessary: moving compute and storage resources closer to where the data is created offering local processing so as to reduce latency, bandwidth, and congestion issues.
Edge cloud architecture refers to the combination of core and remote device hardware, and their configurations, that produce a distributed system. Because this definition allows for a wide range of imaginative architectures, the best approach is to understand that architecture serves the computing tasks, and different architectures must be designed for different uses.
In general, edge environments consist of many smaller devices, and these devices are specialized for a particular task. The core cloud on the other hand aims for consistency, uniformity, and avoids specialization in favor of economies of scale. This is the main reason data is backhauled to core resources, where it is cheaper to analyze data. Edge cloud architects must evaluate the costs of returning data to a data center, processing it, storing, and responding to requests, against running those tasks at the device local. The question to ask is where is it best for processing different workloads. The right fit typically means pushing time-sensitive tasks to the edge, and retaining heavy processes in the core.
Laboratory settings provide a clear example. In testing millions of biological samples, machine vision and edge computing are used to cut down on the latency of a central server and accelerate testing. Machine vision devices are outfitted with firmware and processors that can evaluate the validity of test tubes at the camera, and reject invalids, like poorly capped tubes and more. The alternative of sending imagery to a central server was the first approach but demonstrated latency that slowed testing. In the real world case of pandemic, testing samples faster proved to be a significant use case for edge computing, as the need to test millions of samples rapidly became in demand.
Real-world cases are only one driver for edge cloud architectures. The rapid adoption of microservices architecture also promotes the relationship between cloud and edge. Microservices are stripped down apps that perform a very limited set of tasks, and are designed to run in very small and portable environments called containers which can be placed nearly anywhere. Because of the granularity this provides, and because containers can run virtually anywhere, developers can be very judicious in asking the question, where should certain workloads run, in the cloud, or at the edge.
This is further empowered as the infrastructure edge, such as where 5G meets mobile devices, continues to expand and capabilities improve. Which means that connectivity will improve. But this does not mean a backslide to cloud only, instead this may promote even more complex workloads at the edge, as well as an expansion of the edge to newer, sophisticated devices.
The edge is a colloquial term that refers to the devices or infrastructure away from an organization's core cloud resources, respectively, device edge and infrastructure edge. While the edge can be a physical location, the simplest model, the edge can also be demarcated by logical separations. Logical separation may look like a warehouse within a compound of warehouses. Many IoT devices may be deployed throughout each warehouse, and each warehouse may be equipped with a server to process data, sending summaries to the cloud (a data center potentially in a different region).
Cell towers are a case of infrastructure edge, providing the connectivity for edge devices, say for autonomous vehicles. A self-driving truck may make thousands of calculations in the moment, but also rely on fast 5G connectivity to understand other circumstances, like where other vehicles are, the road conditions, and how its payload fits into the larger supply chain in real-time. While these are all important, latency seriously curtails the actual driving of the vehicle, and is the reason compute and storage are placed on the vehicle itself. Backhauling data used in driving to the core cloud is simply an unwanted latency risk that can lead to accidents and the loss of life..
Edge to core cloud pipeline refers to the data journey through data generation at the edge to AI analytics in the cloud. This end-to-end pipeline is segmented into 4 stages.
From this model pipeline, infrastructure not only represents a bi-directionallity of data, but also a tiered system of analytics, placing the most immediate needs at the edge, and the most intensive needs in cloud cores where compute and storage are more cost effective and real-time is less of a concern.
The edge plays a security role in the total network. While the edge does open new points of attack, it offers companies a way to limit backend exposure. Placing controls closer to those edge areas enhances visibility into requests. At this point, applying security policies have the effect of creating a drawbridge between the edge and cloud core.
Cloud edge security prioritizes these kinds of gateway policies, and important security fundamentals such as data encryption. But the edge demands modern security practices, such as automated authentication of requests. One way to achieve edge security is through software-defined technologies. Secure Access Service Edge (SASE) is used to deliver these security controls and policies as a cloud computing service directly at the edge rather than at the data center. SASE rests on the concepts of digital identities, zero-trust, and real-time context—always verifying access and authentication rather than opening the door to “trusted” entities. For instance, if all security checks pass, but a request comes from a contextually ambiguous IP-address, then the request could be flagged, prevented, and later followed up on.
Implementing cloud-to-edge strategies are challenged because there has not yet emerged a systematic standard for cloud-to-edge computing. Many strategies in use today are pioneers and custom. Despite this, there are a few fundamentals to consider when devising implementing plans.
The main benefit of edge cloud is to reduce latency, bandwidth, and network congestion issues inherent in the continued proliferation of devices, and the subsequent volumes and velocity of data generated into today’s cloud world. This is achieved through a variety of architectures. The model is to place compute and storage resources where data is generated, namely on those data generating devices, like IoT devices.
In the case of edge clouds, the aim is to place compute and storage resources very close to those devices, but this does not have to mean more data centers. For example, Micro Modular Data Centers (MMDC) are self-contained compute, storage, and networking resources that can be placed in close proximity to edge devices. This is highly applicable in disaster situations where deployment is impromptu and immediate. Another application is in areas where the internet is not available or poor connectivity curtails bandwidth.
Other benefits besides overcoming bandwidth, latency and congestion challenges: autonomy, data sovereignty, edge data security.