It's the latest buzzword in IT, and often comes up in the same conversations as 5G and the internet of things (IoT). It's edge computing, and if you find it confusing, you're not alone. Here's our take on what it is, why it's important, and what makes it different from just running a PC on your desk.
A lot of our data today sits in large centralised data centres, especially with the rise of the cloud, but it's generated elsewhere - in vehicles, on factory production lines, and in offices, by small connected devices with increasing computing power, ranging from vibration sensors to IP cameras. This is also increasingly where it's consumed.
Edge computing moves that data back to where it's made, at what we call the edge of the network, and processes it there. That makes data processing faster because you don't need to send it on a round trip to a data centre.
This is important in cases where you need to make fast decisions. When fitted to an expensive industrial turbine, our vibration sensor could indicate that the machine is about to fail. Turning it off in milliseconds rather than minutes could save thousands. Software on our IP camera could use AI to spot one person tailgating another into a secure area, alerting nearby security staff to act quickly.
Edge computing is also useful in critical situations where network connectivity is patchy. A computing transporting perishable food might need to monitor temperature and humidity, alerting truck drivers if conditions exceed certain thresholds. They can't afford for alarms not to sound just because a cellular connection stopped working, which makes local data processing important.
Edge computing models vary depending on your use case. Some users might embed data processing capabilities directly into the IoT devices themselves. Others might use a local gateway module to extract and aggregate data from legacy industrial equipment, performing local processing to support immediate decisions but sending everything periodically back to headquarters for longer-term analysis.
To complicate matters further, there are also edge data centres. These are small data processing facilities that sit closer to the edge. In some cases, this means a smaller footprint regional data centre that serves a local community with low latency applications, versus a larger facility located on cheap land in a rural area. Some might even sit near to customers in a prefabricated shipping container-style box packed with computing and storage power. Companies like Vapor IO in the US are busy deploying these boxes to provide low-latency services to businesses and their customers.
These facilities can factor into an edge computing strategy, but it's important to distinguish between an edge computing facility and a small data centre that you're using more for cost or physical accessibility reasons. The more that a facility relies on serving local users with latency-sensitive traffic, the more likely they are to fit the description of an edge data centre. It's partly about location, but also about application and intent.
Whichever model you choose along the edge computing continuum, you'll need to consider important factors including:
- Maintainability: What kinds of maintenance will edge devices need and how often? Who will do it?
- Security: How sensitive is the local data being processed? What physical and logical protection will it need?
- Data architecture: How will data be processed at the edge? What subset of it will be sent back to headquarters (if any), and when?
Edge computing is a new proposition for most companies. Tread gingerly when exploring it and focus on discrete pilot projects to test your assumptions and hopefully generate some quick wins.