Part 2 in the 3-part Edge Computing series
The Internet of Things (IoT) is growing. By how much? IDC predicts there will be 30 billion connected things worldwide by 2020. After 2020? That’s anybody’s guess—but one clear indication that this growth will be big is the move to a new Internet addressing system: IPv6.
The problem is that the current system, IPv4, allows for only 4 billion addresses or so, which requires some devices to share addresses. With more and more sensors embedded in more and more things—each requiring an IP address—this state of affairs is unsustainable.
IPv6 solves this problem by bumping up the universe of available addresses to a number that’s hard to comprehend —something like 340,000,000,000,000,000,000,000,000,000,000,000,000 (or 340 trillion trillion trillion).
But what about the data?
I can’t say that I expect 340 trillion trillion trillion IoT devices out there anytime soon. But as the IoT grows, the amount of data generated by proliferating sensors embedded in connected things will grow as well. And for organizations deploying IoT devices to move all this data back and forth via the cloud is simply untenable.
Hence the idea of edge processing. Edge processing, as I explained in a previous blog, is the idea of processing data on the “edge” where IoT devices are deployed—rather than sending all sensor-generated data back to mission central over the cloud. Without edge processing, I don’t think the IoT could be a reality.
But even if we were to revamp the planet’s Internet infrastructure, would you still find value in all that data? In fact, much of the data produced by sensors is not particularly useful. So instead of doing a rip and replace of the Internet, why not just process data at the edge and use an IoT gateway to run the analytics on site, sending back only what’s useful to mission central?
The four pillars
It is such practical concerns that make edge processing an appealing approach for real-world IoT deployments. But how do you move forward?
In a recent white paper, SAP explores some of the primary concerns, categorizing them according to the 4 P’s of intelligent edge processing: presence, performance, power, and protection. The paper examines these four pillars and focuses on better ways to cleanse, filter, and enrich the growing volumes of sensor data. Let’s a take a quick look.
Intelligent edge processing requires your systems to be present at the creation, as it were—on the edge, where the action take place. Using machine learning and smart algorithms on the edge, you can generate insight and take action without human intervention. This is good, because running in a more autonomous fashion is an imperative for the digital economy.
As an example, the paper dives into automated reordering and receiving using warehouse shelves equipped with sensors. A different example, though, is automated work orders triggered by analysis of events. This is interesting because the automated action—creation of a work order—requires a follow-on action involving humans, like putting a technician on site, let’s say. In this way, many organizations will use edge processing in conjunction with human beings doing things. It all depends on the scenario that works best in context.
Intelligent edge processing can improve performance for IoT scenarios by solving the problem of overwhelming traditional data-storage technologies. Take the example of processing in manufacturing where the goal is to approximate a standard set by the “golden batch” for all subsequent manufacturing runs. Combining operational technology with information technology, you can process the complex events that happen on the edge, and bring new batches closer into compliance with the golden batch. This helps improve manufacturing performance, from the perspectives of both speed and quality.
Intelligent edge processing gives you the power to execute processes where they take place—without the latency of data transfer in the cloud. Take, for example, a remote mining operation with limited connectivity. Whatever processes occur on site—say, the ordering of replacement parts for mining equipment—can still be carried out with edge processing. Workers can record the order, and replacements can either be made where parts are locally available or put on hold until the part arrives. In either case, the need for the part is recorded, and the information can be synced opportunistically when a connection becomes available.
Intelligent edge processing can help deliver the security needed for IoT deployments. By their very nature, such deployments emphasize openness and are designed to work with other networks—many of which may not be under your control. With intelligent edge processing, you can track the unique identities of sensors in your network, encrypt any data sent out, and run the necessary checks on data coming in. On-site processing in this fashion, in fact, is required—because managing such security via the cloud would not only introduce data latency into the equation, but could also open up holes to be exploited by malicious actors.
So, yes, the IoT is growing—and along with it, the volumes of data companies are required to manage. This volume of data cannot be managed entirely via the cloud. Edge processing is a solution to this problem. Take a look at the “4 P’s” paper here: “Excellence at the Edge: Achieving Business Outcomes in a Connected World.” And stay tuned for my final blog in this series: “Edge Computing and the New Decentralization: The Rhyming of IT History.”