When computing began, computers that were far too expensive for most companies were shared through timeshare services. Processing was centralized using multi-user systems.
Then mini computers, PCs and LANs were added and we moved the processing to PC workstations and smaller computer platforms. We have seen the decentralization of computing. Now, years later, we are centralizing processing again on hyperscalers for public clouds, but this time with a multi-tenant approach. Dizziness?
These days we are again thinking about decentralization with the rise of edge computing. We've already talked about edge here, and my conclusion remains that there are reasons to use edge computing to reduce latency and store data locally.
The pandemic has brought employees and processing into a highly distributed model and not of choice. Edge computing is in the foreground because it should be used next to cloud computing – and not instead of the cloud. Let us clarify a few things.
There are some edge computing models. First, data is processed directly on an IoT device, such as a thermostat or an autonomous vehicle. Let's call this "device-oriented". Second, some computing platforms are used by services that are geographically distributed and used by multiple clients, typically workstations. Let's call this "edge server oriented".
The second model is the most interesting for companies that are rethinking the distribution of computers after a pandemic. It's also the latest use of the edge computing model and is available in two different flavors: the use of proprietary edge devices sold by public cloud providers, and the use of private servers located in small, geographically disbursed data centers There are office buildings and even houses.
When transitioning to these new Edge models, most companies skip a few considerations, including:
safety. Edge architectures add complexity because data needs to be backed up both on the client workstation and in the cloud. Some add an intermediate server that also requires security. Instead of focusing on backing up the data in a single public cloud, we've backed up the information on multiple systems that store data.
data volume. If you add distributed computing platforms with lower performance, the data volume can overwhelm them. A public cloud storage system with integrated automated database scaling can process almost any volume of data that is thrown at it. This does not apply to edge servers or client workstations.
This does not mean that edge computing cannot be a focus of your transition to post-pandemic cloud computing. I say that you need to understand the problems you are facing.
Copyright © 2020 IDG Communications, Inc.