We define edge computing as the ability to place some amount of processing and data near the sources of the data as well as near the systems or humans that need quick access to the processing.
It’s a simple idea, and certainly nothing new. However, the popularity of edge computing continues to gather steam as we move more systems to centralized public clouds and modernize related applications and data stores.
As a result of this migration, we now recognize that not all modernized applications and data stores should only exist in a central location. Thus, the ‘new’ option of moving them to the realm of edge computing, specifically, to the edge of public clouds.
Much of the initial confusion with edge computing came from erroneous messaging from the tech press (and even from some companies) that edge computing was a replacement for cloud computing and other notions that were incorrect at their core. Yes, there are questions that need to be answered when any new hyped technology concepts hit the technology zeitgeist. However, once when we understood the concepts of edge and cloud computing in context of each other, the patterns of synergy began to emerge. Hopefully the confusion will continue to subsidize.
What drove the concept of edge computing was the rise of IoT and other technologies that are distributed to be optimized for the systems and humans that leverage them.
For example, it doesn’t make sense for a self-driving automobile to send all data and requests for data processing over a cellular network to some centralized system in a public cloud. The only way self-driving cars will work is if they can maintain data and processing at the edge, meaning, in the car. This allows the data and processing to occur with little or no network latency, providing fast enough reactions that you don’t hit a tree.
However, edge is not just for devices anymore. Edge clouds are now an option for those who want to have a small cloud instance in their data center. This allows local processing and data storage with much less latency than if the data and processing requests were sent one thousand miles away to a public cloud server that is shared with hundreds of other tenants.
The idea is to keep some but not all public cloud services on the edge clouds while still supporting a symbiotic relationship with edge clouds and their public cloud overlords. They can work together as needed for storage and processing, sharing data and processing tasks. System developers have the option to deploy data and applications on the edge cloud, the public cloud, or within applications and data sets that are divided between the two.
Microsoft’s Stack and AWS’s Outpost are the best examples of edge clouds. However, other smaller cloud providers have exploited the desire for some enterprises to leverage edge clouds as well. The larger cloud players often look at edge clouds as a path to their public clouds, which typically have more services and benefits. However, some companies will opt for edge cloud over public clouds ongoing.
Beyond edge clouds and edge devices we have:
- Edge sensors, where data is usually consumed through a triggering event. For example, your smart thermostat might send a text to your phone when it’s time to change your filter.
- Edge branches maintain their own set of compute and data storage functions. For example, banks may leverage this model to support a remote branch that independently uses local systems that don’t require constant interaction with centralized systems or public clouds.
- Edge enterprises, like edge branches, allow independent systems to exist within parts of a larger geographically distributed enterprise. For example, an edge enterprise system can support a European office with special data and processing that’s specific to the local office’s country and location.
- Edge datacenters are smaller datacenters that exist to support a geographic region. This edge segment saw strong growth since the pandemic started, with employees working from areas that needed a closer datacenter to support their remote activities.
Right now, the confusion seems to center around how cloud and edge computing are supposed to coexist. Reality check: Edge computing typically means the edge of public clouds. You can partition the processing and data between the public clouds and edge-based systems in such a way that the each carries out a role that each does best.
Consider our self-driving car example above. The device within the car should do some immediate processing, such as figuring out that you’re headed for a tree and take immediate evasive actions so you’re not killed. Also important but less critical, the edge device may share massive amounts of engine data with systems on back-end public clouds servers that can proactively determine maintenance issues. They may leverage more process-intensive services such as AI and deep data analytics to find and match patterns to ensure that you’re not stranded on the side of the road with a mechanical breakdown.
The idea is that each tier—the edge tier and the cloud tier—carry out their own sets of functions that are proper for each tier. The cloud takes on tasks that require large amounts of storage, processing, and even spiralized services such as AI, analytics, and pattern matching. The edge device does tasks that don’t require excessive processing and data storage but need to provide an immediate response with limited or no latency. Together, the edge and cloud systems form a single unified system with edge and cloud components that are purpose built to be hosted on an edge or cloud platform.
The Edge on Public Cloud
The public cloud providers saw this coming a mile away. All major cloud providers offer edge development and deployment services, including those that use container, serverless, and other technologies developed for clouds as well as those developed for edge computing.
Public cloud providers can manage deployments to edge-based systems, and even maintain digital twins for edge-based devices and systems. This allows you to maintain versions of applications and data for testing and deployment that run on most types of edge systems.
Public cloud-based edge development and deployment systems can even handle versioning, configuration management, and other functions related to dealing with a massive amount of distributed edge-based systems. This supports most of the edge computing models listed above such as enterprise, device, edge cloud, and edge datacenters.
Yes, edge computing is many different things, but most paths lead back to public cloud computing. The edge needs to be at the edge of something. In most cases, it’s at the edge of a public cloud. Edge computing and public clouds exist with synergy and codependence. This is the de facto model going forward.