John just finished the first wave on cloud workload migrations for his company. With a solid 500 applications and related data sets migrated to a public cloud, he now has a good understanding of what the costs are after these applications have moved into production.
However, where John had budgeted $1 million a month for ops costs, all in, the company is now getting dinged for $1.25 million. Where does that $250,000 go each month? And more concerning, how were those costs missed with the original cost estimates? Most important, where will John work now once the CFO and CEO get wind of the overruns?
In my consulting work, I’m seeing the same four missed costs over and over.
Mistake No. 1: Resource overprovisioning
This means that organization using cloud computing resources are using too many resources, including storage, compute, databases, etc. By doing this they are in essence buying a four-car garage even though they only have a motorcycle.
Unless resources are kept in check, organizations using cloud computing will typically overspend. Cloud cost governance tools are useful here, so someone is looking at the bill on a daily basis. Moreover, people need training as to how to use the properly sized cloud resources.
Mistake No. 2: Using free resources that are no longer free
I’m also seeing a number of cloud users signing up for free trials, only to have those trials expire and then a big bill show up. These trials need to be managed as if they are paid-for cloud resources and shut down if they are not to be used after the trial ends. And, for goodness sakes, read the fine print!
Mistake No. 3: Forgetting to shut down resources
I recently did a demo of a public cloud provider where I allocated storage instances, forgetting to shut them down after the demo. I got an unexpected bill for several hundred dollars. Now, multiply that by 10,000, and that is what enterprises are paying each month for resources that are not put back after use.
Mistake No. 4: Neglecting that cloud applications can spike traffic
When I was writing applications for a time-sharing service decades ago, a bug in the application caused the CPU loading to spike, and thus generated a huge bill. The same can happen with cloud systems, where bugs in applications cause spikes in network, CPU, or storage usage and thus generate surprising high bills.
The shame of this is that all of these issues can be eliminated with a bit of planning and some cost governance. Until then, better keep those want ads close by!
