A big year for Canadian clouds

15 years ago, I was busy running cloud computing content for Interop and Cloudconnect. The Enterprise Cloud Summit at Interop was arguably the first big cloud event—and we had the heads of the major cloud platforms (Amazon, Google, Microsoft and Red Hat, at the time) on stage for the conference. It was the right time: “cloud computing” was on the front page of every tech magazine.

For that matter, fifteen years ago, we had tech magazines.

Three cloud flavours

To recap, then—as now—clouds come in three distinct types:

• Software as a Service is simply a web application. It’s software functionality, delivered via a web interface. There’s nothing to install; and you migrate by moving content and permissions to the cloud.
• Platform as a Service is a place to build applications where you don’t see the underlyiing components. Sometimes called serverless, you migrate by moving your code.
• Infrastructure as a Service is access to virtual building blocks — servers, storage devices, and so on. You migrate by moving machine images.

This handy taxonomy, which I grabbed from a slide deck I made 12 years ago, explains things a bit:

Of course, plenty has also changed in that time. We're much more comfortable with "serverless" computing; cloud providers offer dozens of services from API-based machine learning to content acceleration to diverse storage and realtime analytical tools.

This means that there are myriad variations on this simple classification. Some platforms, like Airtable or Google Sheets, have powerful programming languages built into them that blur the line between SaaS and PaaS. And many on-demand functions, from image processing to search lookups to content delivery networks, are simply services.

Behind this all is one fundamental idea: Own the base, rent the spike. Computer workloads are bursty, particularly those that are seasonal (like tax filings) or deal with data (like building a machine learning model from unlabelled data sources.) Making compute resources shareable makes them more cost-effective; making them easy to use reduces the time to create a new app or run a prototype.

I've written a bit more about this on the FWD50 website, including some of the challenges clouds face, from lock-in to unbounded costs to compiance and repatriation. Ultimately, drawing the right line between what you control, and what you use in the cloud, is critical if you’re to retain data sovereignty, effective cost management, and the ability to move workloads around. This is complicated stuff you need to get right.

 Grab your ticket here to join us at FWD50 and chart your best move to the cloud.