What is Edge Computing?

We have all heard the term “edge computing” before. But how is edge computing different from the embedded computing use cases that we have known for decades? One of the most significant differences between embedded computing and edge computing is the size of the dataset that edge computing applications utilize. The typical edge application utilizes a dataset that is hundreds of terabytes in size or larger, with some edge application datasets approaching or exceeding a petabyte.

As an example, the engine management system of a Boeing 787 can generate half of a terabyte of data per flight, and next-generation aircraft are expected to generate orders of magnitude more data than that (think 100 terabytes). A facial recognition systems in an airport might have a database as large as a petabyte, while each self-driving car is expected to generate roughly 4TB of data per day. Typical embedded applications (postal kiosks and vending machines for example) seldom have data sets that exceed a terabyte, and typically are well under 100 gigabytes.

This brings us to the second way in which edge computing differ from embedded applications – the scope of the analytics that occur within the application. Edge computing applications generally involves significant computation loads. This is not only due to the amount of data that edge computing applications create and utilize. For instance, think of the workload involved an artificial intelligence-based facial similarity search. Another example is engine management systems – for commercial aircraft, these systems must not only be able to record the data, they must also be able to process it inflight, so that the aircraft can be serviced when it lands. The applications managing self-driving cars are at least as complex. The combination of these two requirements makes edge computing use cases in a class by themselves. We will discuss some of the difficulties in creating and deploying edge computing solutions in the next blog of our series.