Harnessing The Power Of Compute At The Edge For IoT

Harnessing The Power Of Compute At The Edge For IoT

This article was originally posted on Forbes. Check it out here!

Today’s internet of things (IoT) landscape is expanding wildly – everything from connected in-home dishwashers to machinery in automobile plants and point-of-sale mobile devices in retail stores – and it’s changing the way individuals and businesses operate. Each of these connected devices has a unique performance requirement that doesn’t allow for buffering and latency. IoT today runs from the cloud, but the cloud is still subject to network connectivity issues and potential performance degradation. The cloud was never designed to handle the bandwidth overloads that IoT creates. Plus, often the data generated by those devices is useless if it doesn’t get analyzed within a fraction of a second, right there at the edge.

As an early adopter of the SD-WAN shift, I have seen firsthand how enterprises today are becoming more reliant on compute power and the development of better technology-enabled devices. With the integration of intelligence into every device we touch, every connection and compute requirement is mission critical. The data we create in our IoT world happens at the edge. Today it is sent across a network into a data center, processed and sent back to the edge to be delivered to the user. But that non-real-time method of data flow is no longer cutting it for businesses. Imagine a manufacturing plant that has hundreds of automated machines that trigger hundreds of data points across hundreds of tasks — there’s not enough bandwidth to support the constant passing of data to the data center and back to the edge to give true real-time performance. Most organizations today don’t have the ability to process big data at the edge because they lack the computing power and resources needed for machine learning.

Consider a retail shop as an example. Today’s intelligent world of IoT utilizes beacons in everyday shopping experiences. The beacons ping devices that are in range, which generate data. They run analytics to identify the best-fit promotion or advertisement and deliver it back to the shopper in real time. With consumers’ average attention span decreasing by 88% per year, any lapse in data processing or delay to the consumer has the potential for revenue loss. That doesn’t leave much room for sending data miles away to processing at a data center, just to lose your customer in real time.

I believe it’s critical for companies to intelligently integrate and plan for the future of the integrated network. IDC predicts that by 2020, “IT spend on edge infrastructure will reach up to 18% of the total spend on IoT infrastructure, driven by deployments of converged IT/OT systems that reduce the time to value of data collected from their connected devices.” This means no longer focusing on individual functions at the edge, but consolidating and giving way toward the internet of things. Routers, switches and edge functions are just the beginning for network edges. Soon to follow, I believe, will be physical compute power sitting at the edge, complete with virtual machines on physical hardware stacks that give the processing and compute power to the device right in your store, office or field location.

Preparing for this new way of edge computing can be daunting. The more automation and integrations we include in our network, the more we feel we’re giving up control. But this doesn’t have to be the case. As businesses consider alternatives to the status quo, there are a few places to begin. I’ve seen the SD-WAN space become more widely adopted; IDC is forecasting it to be an $8 billion industry by 2021. I believe service providers looking to offer customers the option to deploy the cloud within their networks will move toward other network function virtualization (NFV) platforms.

There are steps you can take on your own to prepare for the change. First, look into your existing carrier options. You want to find out whether there are new or more efficient connectivity options that allow for better performance without the high price. Also consider whether you currently have physical or virtual devices at the edge right now. Before moving to more virtual platforms, consider looking at how much time is left for you to find the full value of that hardware before eliminating it.

I believe the next evolution of the network is one that doesn’t require a data center, doesn’t require remote infrastructure and doesn’t rely on a one-to-one architecture. Sounds like something from the 1990s, right? Back when we had server closets in our offices and ran all our applications locally. But this is a next phase of the cloud. Data center connectivity becomes a thing of the past because the compute power sits right where you need it the most — at your network edge.

Understanding the next evolution of the cloud means understanding that living in a data center doesn’t become the be-all and end-all. Businesses require real-time data processing and results with a fraction of the response time that they’re getting now. Moving compute to the physical edge means the edge network becomes that much stronger, that much more flexible and that much more stable.