By Sean Feeney
The rate of change in the IT industry is ever-increasing; there have been several technology trends that have come and gone this decade likely before your business even had a chance to experiment with them. Remember Platform-as-a-Service (PaaS)? Has your Big Data become just data? Have you even started running production workloads on Kubernetes yet?
2019 will be no exception. It’s shaping up to be the year that CIOs begin asking, “Why isn’t this solution serverless?” following in the footsteps of organizations like Fender, who just announced its new Fender Digital services are 100 percent serverless.
Whether you’re new to custom software development or have been deploying microservices for years, the value proposition of serverless, aka Function-as-a-Service (FaaS), is compelling. Pay only when a particular function is used, or scale with zero effort. Spend all of your time developing your business and your code, finally realizing the infrastructure-free dream that PaaS once promised.
But buyer beware — serverless could be the next mainframe, if some commentators are to be believed. While your code could generally be run anywhere and the Serverless Framework is now supported by all three major cloud service providers (CSPs), the ecosystem required to debug such a workload in production is not equal between CSPs. If you couple with additional provider-specific services like Aurora Serverless, which you should for maximum value, you’ve even further locked yourself into Amazon, for instance. The good news here is that unlike IT vendors of yore, CSPs are only known for lowering costs, not hiking them.
The reality is that serverless — Lambda, Azure Functions, Cloud Functions, or the Serverless Framework — should be a part of your toolbox. Like most new technologies, it offers key benefits for specific architectures. This includes a low per-service implementation effort, an attractive operating cost model, and an opportunity for extreme parallelism that pairs well with fan-out architectures. But it isn’t any more of a silver bullet than the last major trend of containerization.
Fender Digital had an advantage that you might not: all of their apps were 100 percent greenfield. Existing applications can’t just be stuffed into a FaaS, like many companies are currently trying to get away with in their Docker and Kubernetes migrations. They need to be re-architected and refactored. It may be just as cost effective, in terms of labor, to rewrite the app entirely.
Doing a cost-benefit analysis like this, and then implementing accordingly, is where Nerdery’s Platform Modernization can help. Let us help you review all of the critical factors and build the business case for serverless before you ask your teams to support another platform.
Published on 02.07.19