Why your cloudops staff is quitting


It’s Monday morning and you have another letter of resignation on your desk. This time from a woman who was doing performance monitoring and cloud-system tuning. Last week it was a database operations administrator, and two more from the cloudops team quit the week before.

What happened? The cloud was supposed to make things easier. Are you underpaying or overworking, or is there something else that is harder to fix?

The fact of the matter is that cloud operations teams are going to be abused during the next one to six years. There is very little understanding as to how the jobs will morph, and we grossly underestimated how complicated and challenging the cloudops roles would be.

It all starts upstream. You would expect companies to have a master plan to make things less complex and easier to operate. The fact is most are sprinting from one cloud migration to another, layering complex systems on top of complex systems, using whatever cloud technology seems cool at the time.

This results in a mix of containers, serverless computing, machine learning, big data using columnar databases, big data using object databases, eight brands and flavors of security, topped off with inconsistent governance models and tools. Guess where all this complexity goes to live a long life? Cloud operations. People are figuring out they are being set up for failure and are jumping ship before things get worse.

You have a couple of choices here:

First, you can toss money at the problem. Hire more people who have a large inventory of skills (container operations, serverless operations, AI operations). Moreover, give them all the tools they need. Unfortunately, CFOs don’t like blank checks. Cloudops, like other areas of the company, has to maintain and live within a budget. The odds are overwhelming that nobody budgeted for the forthcoming operational complexity.

Second, you can get a lot better at architecture and planning. Build for common services, including security, devops, and data. Standardize on heterogeneous cloud services that are consistent across application and data domains. Thus, what flows to cloudops will be consistent and less complex, in line with existing skill sets and tools.

The idea is to move towards less complexity, not more, and stem the tide of cloudops personnel leaving the building. Until that happens, I can’t blame them.