What is CloudOps?

I was asked this question recently. Here’s my answer.

CloudOps?? - DALL-E

CloudOps sounds a bit like DevOps and in a way it is. In the olden days before the Internet, we looked after servers on the premises of the business. On Prem servers. There was a team, looking after the ‘tin’, infrastructure team.

Ops

This started to evolve into an ops team with the advent of virtualisation. It was around 2005 when to fulfil my application installation project, I would order a server, but this time it was a server running on VMWare, rather than a physical racked server. The advantage was this took 1 day to fulfil, whereas a physical server order would take about two weeks.

The Ops team would be performing the work, to deliver the server to me in a state where I could log onto it remotely. When they started delivering virtual machines, there was automation to help them. Templates, scripts and so on to reduce the delivery time further.

Cloud

Then the cloud was getting a bit more serious. For my project I could order a server that was running on the cloud, rather than in the On-Prem Data Centre. The team providing this server, was largely the same team as before, only now they were no longer called Ops, but CloudOps.

DevOps

This separation of server delivery and application delivery is an Agile and DevOps anti-pattern. Working together, by using the self service capabilities of the cloud provider directly, a DevOps team can order a server in seconds, having it online with an image the team have constructed and running the application in minutes. This enables rapid scaling of server resources, if required.

The speed of delivery gives another benefit. When you have to wait for a day for the delivery of a server and the replication of the paperwork that goes with it, it is not good use of time to destroy and recreate on a daily or even weekly basis.

If you know you can create it within minutes, then you aren’t afraid to destroy the server. This means the bill payer may only get the bill for a server that has run just long enough to provide value to the developers, the company and ultimately the customers.

Per hour, per minute or even per second billing is no use if it takes a long or manual process to create the server. To take advantage of measured service you need automation in place. For the one off cost of writing the automation, for servers, usually a combination of Terraform and Ansible, you can save lots on an ongoing basis by only running the servers when they are required.

Want to know more? Come ask me questions. You can find me on Slack.

If you found this useful, subscribe to updates and look out for my YouTube channel content.