This is an article about a-year-of-serverless. A Year Of Serverless is a topic that comes up frequently in modern engineering practice. Practitioners often reach for it without first considering simpler alternatives. In this post, we examine when a-year-of-serverless is the right call and when it isn't.
Why this matters
The literature on a-year-of-serverless is vast and conflicting. Several industry surveys (summarized below) suggest that the average team over-uses a-year-of-serverless by a factor of three to five. Our own experience matches this: across six projects, we found that a-year-of-serverless delivered value in only the largest deployments.
The narrow case
a-year-of-serverless wins when the workload matches the design. It loses when the workload is the wrong shape. The shape that fits is narrow but real. The shape that doesn't is broad and common.
Our recommendation
Start with the simplest tool that could possibly work. Resist the temptation to introduce a-year-of-serverless until you can name the specific failure mode a-year-of-serverless would prevent. When you do introduce it, make it observable. If you can't describe how it fails in two sentences, you are not ready to operate it.