What is edge computing?

- So what is edge computing anyway? Well, it's an architecture in essence where we're putting the compute and the storage in the data resources as close to the source of the data as possible. Therefore making it more effective and more efficient. It also includes the notion of a centralized processor. In other words, centralized compute, centralized storage, centralized databases. And while that can be a physical computer that sits in a traditional data center. Typically it's going to be a public cloud provider. So what are these Edge devices? Well, it's any device that supports storage of data or information and networking and compute. In other words, the ability to process the information. And, it may support application development. It may support different security systems and it's self-contained. And its place in a location where it'll do best for the system. In other words, reducing latency, increasing performance, things like that. Now here's an example of an Edge device, Raspberry Pi. This is something that hobbyists typically buy. It's a small inexpensive computer. Typically they're less than $50 US. And they allow you to in essence build systems on this device. And it really is a true computer. In other words, it has storage. It has a central processing unit. It has networking and it can run any number of operating systems depending on its use case. So, this would be something that would be a good candidate for being an Edge device, because it has all the attributes of an Edge device. And we can build applications upon it. So the basic Edge architecture is this. We have data input, which is typically gathered through sensors. We have the edge computers themselves which allow you to do storage compute. They may have things such as machine learning on them. And then we have the cloud based central computer. In other words, the ability to leverage some sort of centralized system where there's a one to many relationship between the cloud-based central computer which is typically one or a few to very many edge computers. Which means that we're talking to many devices through a single set of management infrastructure that sits on the central computer which is typically housed in a public cloud. So you've got the data sources coming from the sensors. You have the ability to deal with lightweight edge processing at the edge computer. Typically something that is going to be fairly lightweight considering the fact that we're dealing with underpowered processors, low storage devices, low speed networking devices, things like that. And where the heavyweight activity is occurring within the central processing system. Typically the public cloud where the resources are almost unlimited and they're able to deal with terabytes or petabytes of data at an instance. In essence deal with the information coming from multiple thousands of edge computing devices that may be talking to one single instance of a management process that exists in the public cloud. So keep in mind that there is an input and response paradigm here. So in other words, we're dealing with sensors in data but there's a closed loop between the edge computer and the data input. So in other words, the edge computer is able to analyze the information on the fly. And without having to go back to the cloud-based central computer can make decisions. Such as, if it's on a jet airplane and it notices that the engine is overheating. It can make a decision to shut down that engine automatically before it catches fire without having to transmit the request back to some sort of centralized system. In other words, it's smart into itself. It's able to make autonomous decisions. But there's also an input and response mechanism between the edge computer and the cloud-based central computer. Typically this is asking more holistic questions. In other words, if the jet engine that we were just speaking of, is starting to exhibit behaviors that may be leading up to a failure. We can check against the knowledge base, for example, on the cloud-based central computing. That is able to store petabytes of information thousands of hours of engine behavior over time understanding the net results of that behavior and determine the percentage of the risk that that engine is going to fail. And therefore we get a big data experience, but do so on a very small computer system. In this case the edge based computer. So in other words, it's able to do things on its own. And it's able to in essence, ask bigger questions to the centralized cloud processing system to be better at its job.