As enterprises begin to embrace the idea of edge computing, all evaluations should begin with use cases. After all, the only reason to deploy resources on or near a remote location, as opposed to in a data center or in the cloud, is to meet stringent, use case-specific requirements.
The drivers that dominate typical edge computing use cases vary by industry. Typically, use cases that point IT toward an edge deployment revolve around one of three things:
- a need for low-latency responsiveness;
- a need for analysis of data sets that are too large to move to the data center or cloud fast enough or cheaply enough; and
- a need to replace traditional existing branch infrastructure to keep data local -- for compliance reasons -- or to provide resilient services, like for point-of-sale systems.
In manufacturing, the most common edge computing use case is responsiveness. Modern, data-driven factory automation often requires latencies lower than 10 ms. This impedes most site-data center-site round trips and forces at least some data collection, analysis and machine control into edge resources.
IT staff need to work closely with operational technology (OT) staff to make sure they're in sync with how they provision, deploy and manage the system. Ideally, IT and OT could provision a pool of local resources to be shared across workloads.
This article is part of
Sometimes, though, edge computing options are specific to the robots or other equipment on the line, and they come with dedicated hardware. In such cases, IT needs to research and test the behaviors of the system and implement whichever compensating controls are needed to best fit the edge systems into the network and security architectures. After that, IT must integrate incident and change management cycles on the edge infrastructure into overall incident response and change management policies and plans.
With retail operations -- including food service chains and some kinds of financial services -- typical edge computing use cases revolve around the driver of resilience and localization. No retailer or insurance company wants business to grind to a halt whenever it has WAN or internet service problems. Further, an increasing number of jurisdictions around the world require some data about customers in a certain jurisdiction to remain within that jurisdiction.
As a result, enterprises are interested in edge computing as an evolved version of traditional branch computing, trading a wild variation in resources and management for the consistency, automation and central management that are key to the idea of edge computing. Here, the focus is on shared infrastructure; use of common tools for management, monitoring and automation; and smooth integration of disparate resource pools into shared architectures, processes and management teams.
IT employees should select tools and resources to minimize differences with central data centers. They should also ensure they explicitly address remote facilities in all policies, processes and relevant job descriptions, as well as in product evaluations for new data center tools.
Medical and aviation
The edge computing use case of data mass can pop up in many verticals. Some forms of medical testing, for example, can generate data masses so large that some on-site analysis is critical to consolidate it to manageable -- transmittable -- sizes.
Some nonmedical instrumentation also generate mountains of data. A jet engine, for example, may generate terabytes of performance and safety data during each hour of a flight, necessitating massive in-plane storage and, for safety, some on-plane analysis of that data. Again, makers of the equipment may ship the product with edge systems to serve it, but they may also ship it with tools built to run in industry standard infrastructure.
IT teams need to focus heavily on the redundancy engineering of any infrastructure they deploy, whether supplied by the tool provider or not. They should also revamp processes, as well as management and monitoring tools, to support continuous availability and decentralized resource pools.
Whatever the initial driver for edge computing, IT's main goal has to be making sure it isn't just a replay of old legacy branch IT. Instead, it should truly be a centrally managed service that runs on distributed resources and meets specific business needs the best way available.