Why the Cloud Makes Forecasts Difficult and How FinOps Helps
When organizations move their workloads to the cloud, it can get tricky to predict costs.
The approach many companies take to cloud migration, according to Matt Stellpflug, the guest on this episode of The New Stack Makers, “would be, I’m going to rent a co-location facility. I’m gonna buy, with a big capital outlay, some hardware, network storage, compute, and I’ve got very predictable costs.”
The cost of that initial setup may be predictable. However, the cloud costs are based on usage, and the rate you pay for what you use. “And, of course, what you use is much more granular than it has been previously,” noted Stellpflug, a senior FinOps specialist at ProsperOps, a rate optimization automation platform.
And predicting your costs? It’s complicated.
“You get this mode in which, to get things done, people are always launching new things, and maybe stopping things, and you’ve got this fluctuation. And you’ve got these anomalies. And there’s these other soft costs,” Stellpflug said.
Now, he said, “Not only do I pay for my compute and storage, but I pay for accessing it and transferring it from A to B. There’s all this other nuance to it. That’s really hard to forecast without some kind of a reference workload.”
In this episode, Stellpflug talked to Heather Joslyn, Makers host, about how engineers can implement FinOps to help establish those reference workloads and benchmarks so that it’s easier for their organizations to predict and contain cloud costs.
This conversation about FinOps was sponsored by ProsperOps.
Triage Problems and ‘Do the Low-Hanging Fruit’
Engineers are at the crux of any FinOps initiative, Stellpflug said. They are “the protectors of application availability and uptime, and if anything goes wrong they’re paged.” It makes sense, then, for them to be involved in anything having to do with the integrity of the system overall.
When working with engineers on a FinOps effort, he said, “The key is to work with your engineering team and ask them: What metrics do we care about?”
Stellpflug co-wrote an “Engineer’s Guide to Cloud Cost Optimization” (with parts one, two and three available on TNS). In addition to underscoring the difference between resource optimization and rate optimization (what’s being used, and how much the organization is paying for it), he pointed to some best practices from the guide to help organizations implement FinOps.
Engineers should identify areas of excessive resource consumption and triage based on risk and impact. “You want to do the high-impact, low-risk items first. And of course, high risk, low impact would be last,” Stellpflug said.
“Do the low-hanging fruit. But then indeed, for those high impact, high risk, these are things that need contacts from subject matter experts, so engage them. But understand that this will be a longer engagement than the low-risk types of engagements.”
And when tackling those more complicated problems, he advises, “ you want to maintain momentum. And so as you iterate through this cycle, don’t get hung up on like, well, this is going to take six months, Keep that running but do some easy things as well.”
Listen to the full episode for more on implementing FinOps.