4 Types of Edge Computing - Broadly Categorized

Edge computing is considered next frontier in distributed computing. However not everyone knows what it is or the fact that there are multiple types of Edges!

Simply put Edge computing enables connectivity between plethora of devices (usually called User Equipment UE in 5G terms) and core of telecom datacenter by moving the computing closer to end user.

The key enabler being the latency required for round trip communication between the UE and compute servers requiring a highly distributed architecture.

One easy way to understand various types of edge compute is based on its proximity to the end device and round-trip latency to core of the datacenter. Latency being the primary factor, Edge compute could be broadly categorized into following:

• Internet of Things Edge
• On Premise Edge
• Access Edge
• Network Edge

Internet of Things Edge:

Latency expectations at this Edge is usually less than 1 millisecond.

This covers almost any device that has connectivity to private or public network such as internet. The device can be a smart device that capabilities to process data such as mobile phones or a simple sensor that relays information about its surroundings.

Examples include, but not limited to, retail kiosks, cameras, factory sensors, connected cars, drones, connected streetlamps, smart parking meters, remote surgery equipment, etc.

On Premise Edge:

Latency expectations at this Edge is usually less than 5 milliseconds.

One needs a way to aggregate data from many devices at the IoT Edge in order to store, process, analyze and respond to relevant requests using the data. On premise Edge provide compute resources at enterprise premise there by helping localize data processing and reduces the time of data processing. The devices at this Edge usually connect to Network Edge or data center for further requests.

Businesses such as large enterprises, industrial manufacturing floors, large retail operations, etc., benefit from these types of deployments to have ability to process the data close to point of its origin while still having proprietary rights of owning the necessary hardware.

Universal Customer Premise Equipment (u-CPE) is an example where it provides combination of firewall, WAN optimizer, router within a single equipment that is deployed On Premise.

The networking requirements are usually satisfied utilizing Software Defined Wide Area Network (SD-WAN) that allows enterprises to utilize combination of MPLS or LTE or broadband internet services.

5G has enabled Small Cells to be established On Premise as well to operate in licensed and unlicensed spectrum.

Access Edge

The latency expectation at this Edge is usually less than 10–40 milliseconds.

The once traditional Radio Access Network (RAN) that was available as a fixed function device has now been disaggregated to run as set of virtual functions in software using off the shelf servers. RAN is the crucial point that connects wireless devices to the core the telecom operator’s network.

Concepts like virtual RAN (vRAN) and industry initiatives such as Open-RAN and O-RAN have enabled interfaces to manage these virtualized deployments similar to managing any other edge device. Move towards cloud native instantiation of RAN function is easing the life cycle management of broad set of RAN deployments using Continuous Integration (CI)/Continuous Deployment (CD) constructs leveraging DevOps models.

The set of infrastructures necessary to deploy and manage software RAN functions could be broadly termed as Access Edge.

Network Edge

The latency expectation at this Edge is usually less than 10–40 milliseconds as well.

Data from across multiple IoT edges, On-premises edge and Access edge, catering to the specific region, needs to be aggregated before connecting to the centralized data center that can span across vast set of regions.

This type of deployments could be broadly called “Network Edge” or near edge in relation to the core data center of the service provider.

Next Generation Central Office (NGCO) architecture aims to address requirements of Network Edge. Another example in this type of Edge is Wireline Fixed Access Edge that provides broadband services such as IPTV, VoIP, internet services, etc.

With the advent of Fixed Mobile Convergence, the demarcation between various device types at this Edge is thinning due to architectural convergence.

Summary

Edge computing is going mainstream with various deployment models. The 4 types shared above provide the space or domain at which one would operate and draw out interfaces and interactions to be able to scale across Edge types. There are lot more terms used than above (various types of edge compute) but then when it comes to round trip latency, most likely they fall under above.

Do you see patterns or Edge types that don't fall under above? Feel free to comment & leave a note.