Handling API calls in the cloud with a service like AWS Lambda is one of the latest ways enterprises try to save on resources. In theory, you can save a ton of money. But be careful. Costs for API calls can mount up, sometimes leaving you with a cloud computing bill ten times the size you expected. David Linthicum over at SearchCloudComputing explains more.
David begins by saying that enterprises often compare features for cloud providers when choosing a public cloud-based API but rarely costs. But when you break it down, the costs differ between providers and those costs can mount so high it can hit the bottom line. He says companies should develop strategies to estimate and manage the costs associated with API calls.
Enterprises often don’t worry about costs for API calls because they don’t seem to differ much between providers. AWS Lambda and Microsoft’s Azure Functions, for example, both offer a million function calls for free each month followed by a charge of $0.2 per million calls after that. If you have one API call per function call and you make two million function calls, that’s $0.20 you have to pay. Azure Functions is similar but gets slightly more complicated. Microsoft also bills you on functions for observed resource consumption measured in gigabytes per second (GBps). To get an idea of what you’ll pay, you have to work out the average memory size in GB and multiply it by the time it takes to run the function. The first 400,000 GBps are free, but then you have to pay $0.000016 per GBps.
What this all means is that if you make 5m API calls or more each month, your bill could go thru the roof. Many enterprises use a wait-and-see approach when dealing with their cloud bills. With this approach, however, an big enterprise could end up with cloud bills of $100,000 or more per month, having expected the figure to be a tenth of that.
David recommends getting cost governance tools that can monitor API calls so you can avoid end-of-month billing horrors. These tools monitor service usage and can set limits to API costs. They are even cleverer than that. You can receive recommendations to achieve greater cost efficiency. For example, the tools can suggest buying up a bunch of API calls ahead of time when they’re offered at a large discount, or making calls during off-peak hours where possible.
David concludes by saying that as serverless computing becomes ever more popular, keeping track of the associated prices will become fundamental for enterprises.