In a previous post, I discussed why there currently is a big focus on edge computing in the IT industry. In this post, I'm looking at types of computing, and which qualifies as edge computing.
What exactly edge computing is, seems to be vague and varies depending on who you ask. This post focuses on a definition VMware recently started using and breaks down computing into four parts, Cloud, Compute Edge, Device Edge, and Sensor, and everything outside the Cloud is considered Edge Computing.
Cloud: Cloud primarily refers to massive data centers operated by cloud providers such as AWS, Azure, and GCP, but would also include VMware Cloud on AWS and other cloud or hosting providers. The main characteristics of the cloud are that it's centralized and operate at scale. The benefit is that you have very high infrastructure availability, access to a lot of services and pretty much infinite amount of resources. The drawback is that as these are centralized, network connectivity to sensors or devices at the edge can’t be guaranteed and latency will be higher. Network traffic to or from the cloud most likely also incurs a cost.
Compute Edge: This is also referred to as a micro DC and is a small datacenter consisting of anything from a few up to several racks of servers. These are generally located close or next to IoT devices and might also be necessary for local compliance reasons. The key is that these data centers have cooling etc. and contain standard rack-mounted servers. You can have quite a lot of resources located in these data centers, but not the same range of services and capacity as in the cloud. A benefit is that latency to devices at the edge should be less compared to the cloud, and network bandwidth might be higher and more reliable.
Device Edge: This consists of one or a few small servers and is also called a nano DC. It would only consist of one or a few small servers and have minimal compute capacity. Servers in this category would probably not be rack mounted and would have to be capable of running without cooling. They are often found in locations usually not associated with a data center, such as in factories, wind turbines or cars and can be rugged to handle extreme conditions. The benefit is that they could be located right next to IoT sensors and latency, bandwidth or connectivity issues would be minimal. The drawback is such small devices would only be able to provide minimum capacity and services.
Sensor: IoT sensors are devices that either collect data or control something in the world, such as a security camera, counter, light bulb, etc. They would generally not contain any compute capacity in themselves and would instead communicate with the device edge, compute edge or cloud, depending on the bandwidth, latency or connectivity requirements.
The idea is that future applications, interacting with devices at the edge, would be distributed across these different computing layers. Some parts of the application would need to have direct access to edge sensors and would have to be in the Device Edge, while other parts would need access to more complex services and more compute capacity located in Compute Edge or the Cloud. Components in different locations would work together to make up an application.