alphaspirit - Fotolia
Edge computing focuses on putting compute and storage resources outside the enterprise data center, in or near branches. But edge computing, propelled by IoT and poised to grow with the spread of 5G services, is not traditional decentralized computing. CIOs need to understand what drives an edge computing strategy in order to avoid recreating the problems with branch IT that data center consolidation was designed to fix.
Edge computing is distinguished from legacy distributed computing by motivation as well as by technology, architecture and management philosophy.
Legacy decentralized computing arose mainly because of ownership and control issues -- the business lines in the branch wanted IT resources they controlled -- and because of the need to minimize use of WAN network links.
An edge computing strategy, by contrast, assumes centralized delivery as the norm and expects adequate bandwidth to deliver services; it puts compute and storage resources in enterprise branches or in co-location facilities near them, solely in order to meet the functional needs of specific use cases.
Latency and bandwidth
Indeed, edge computing is coming to the fore now because we are moving into an era of tasks that require very short response times or that need to digest data sets so large it's more practical to do at least some analysis locally -- or both. IoT drives most of these use cases. IoT applications including real-time control of high-speed manufacturing lines or automatic control of driverless vehicles (for example, in a warehouse) can require both the sub-millisecond response time and the need for massive data analysis.
But an edge computing strategy is about more than just sticking servers and storage in or near a branch. It focuses on creating a centrally managed but physically distributed infrastructure that supports hands-free/lights-out operations.
Cloud-style infrastructure enables both the fully centralized management and the lights-out operation -- so edge computing assumes it. Compute and storage resources in the form of converged or hyper-converged infrastructure require the minimum amount of manual intervention to put them in production. In a private cloud, most configuration and all utilization of such resources is centrally managed. Extending the cloud paradigm to edge compute models makes it as simple as possible for IT to configure each site to order and then add more resources (if needed); the plug-and-play components require a minimum of skilled staff time and effort, because they are automatically discovered and put in service. Indeed, specialized support for enterprise edge computing is driving development of new products in this space (e.g., products that provide components in smaller form factors, or consume less power or generate less heat, to allow for placement in small cabinets or in wiring closets).
Rethinking branch networking with edge
IT leaders already dealing with use cases that can justify edge computing functionally may also be able to help justify an edge computing strategy financially by using it as a means to change the economics of their branch networking.
Once a robust platform for localized compute is in place, IT can use it to offload work from the WAN (no longer shipping out large data streams to be worked on elsewhere) and to reduce the performance demands on the WAN, because critical response time is handled with edge resources. This might allow IT teams to reduce their site connectivity spending, especially on multiprotocol label switching services. And edge infrastructure can support the functions of an internet and security hub for a region as well, offloading work from other branches in addition to central data centers and the security core and internet links there.
Edge computing is only now emerging as an area of broad enterprise interest, as the use cases that can make it necessary converge with the availability of technologies that make it a practical, sustainable option. Most enterprises do not yet have a use case for edge computing, but more enterprises will each year driven by successive waves of IoT and analytics projects.