Even big cloud providers are embracing edge computing

Edge computing is attracting a lot of attention these days, even from cloud providers that have previously said “everything will go to the cloud.” For example, take AWS Outposts, a new service that AWS announced in November. This service, which should become available next quarter, enables customers to run AWS on-premises, and then connect to the AWS cloud. To accomplish this, AWS ships customers fully managed racks of AWS-designed compute and storage hardware running VMware.

But there’s more. AWS IoT SiteWise, part of a suite of IoT services, collects data from industrial sites through an on-premises gateway. The very next week, Microsoft announced new functionality for its own IoT service, Azure IoT Edge. Clearly, AWS and Microsoft now realize that the cloud alone will not be enough to provide the performance that enterprise IT and the growing number of applications for IoT will require.

The market interest from cloud providers, tech vendors and investors in edge computing is easy to understand: The global edge computing market was worth just shy of $8 billion in 2017 and is expected to surpass $20 billion by 2026, according to market research firm Stratistics MRC.

But that market interest is ultimately driven by a known technical weakness in public cloud-only models: performance. Applications, workloads and data sets that require a nearly instantaneous response function poorly when they are located far away from the users who are working with them. There’s nothing an engineer can do to transmit data faster than the speed of light, and the distances between large metro areas, multiple endpoints generating data and big cloud data centers are large enough to introduce significant latency, an issue that’s becoming even more acute with the growth of IoT.

As IoT continues to expand, users and applications will be generating more data than ever. But as effective as the cloud is for storage, the concept doesn’t work when it comes to delivering a fast response when there are great distances involved. And the availability of internet bandwidth in places where IoT and other remote devices are apt to be located can be quite performance-constrained.

When you consider how much data IoT will create — connected cars alone are expected to generate and use as much as 40 TB of data per vehicle per day — there are simply not enough pipes and bandwidth to move to and from the cloud with a low level of latency. Autonomous vehicles, for instance, can’t afford to wait seconds to send data to the cloud, process that data and receive an answer. By then, it may be too late to avoid a hazard or take a better route to the destination.