Edge analytics improves the way we live

[JoannStarke]

Edge computing is the next wave of digital transformation. It’s the next turn of the crank for your business to be more efficient, enable faster healthcare diagnosis, or quickly identify new patterns or trends that could impact business performance.  

This is not internet of things. Edge computing is about moving the compute to where the data is generated, then processing it to enable real-time corrective action. It’s for use cases that are not latency tolerant and where data gravity is an issue. In fact, the more data gravity you have, the more you need edge computing.

Time for an edge-in strategy

Edge computing exists everywhere the cloud or a data center cannot. It’s everywhere humans, devices, cameras, and sensors are located. This year, 55 billion edge devices, cameras, and sensors will be deployed globally, and all this data is short-lived. For example, you have 500 stores with security cameras watching customers. You notice theft has increased in one of your stores, resulting in the daily review of security tapes. You don’t need to keep all the tapes, just the ones where behavioral anomalies appear. From here you can analyze other store tapes to see if this is a new pattern or trend and adjust with employee training. 

Digital business at the edge needs an edge-in strategy. One that delivers speed and low latency, so you don’t miss opportunities. By creating a small version of a data center at the edge, data streams, MRIs, or manufacturing quality assurance can be analyzed in seconds—eliminating the delay between when the data is born and when it can be processed.

Let’s look at a real business use case   

When HPE builds your server, quality matters. Connectors must be inserted securely and in the right place. Jumper and switch settings must be accurate. Normally, QA was performed by human beings, but humans are prone to not see things consistently and in enough detail.

By replacing human eyes with an array of high-resolution cameras combined with AI for pass/fail decisions, HPE can catch errors in much greater detail. As the finished server rolls by, the cameras instantly capture the product from five different angles. Each camera creates approximately 15 megabytes of data, equating to 75 megabytes for each finished product. Sending 75 megabytes of data to a central processing location took 21 seconds. And that was too long.

Instead, the data is fed into a series of neural networks for pattern recognition. When combined with GPUs and AI, unit pass/fail was made in seconds. This may not seem significant but when you consider thousands of servers per month, the savings allowed HPE to shave seconds off the QA process and lowered missed or mis-diagnosed failures by 33%.

Ready for the next wave of innovation?

At the edge, seconds matter. Analyze data where it is created for real-time insights that allow you to respond quickly to changes that could impact business performance.

Watch this video to learn more about edge-in computing.

Hewlett Packard Enterprise

twitter.com/HPE_Ezmeral
linkedin.com/showcase/hpe-ezmeral
hpe.com/software