Add It Up: FaaS ≠ Serverless
We can not stop people from claiming that serverless and FaaS are synonyms, but there is data to show that they are not the same thing.
Serverless architecture is not needed to use Function as a Service (FaaS). In fact, 54 percent of FaaS users without production deployments say their organization does not utilize a serverless architecture, according to a survey conducted in August 2018. In comparison, 96 percent of organizations with FaaS broadly deployed say they use a serverless architecture. Our upcoming Guide to Serverless Technologies covers serverless architecture, technology and computing, and will include the complete study results.
The survey defined FaaS as typically providing event-driven computing where developers run and manage application code with functions that are triggered by events or HTTP requests. Serverless architecture broadly describes an application design that incorporates third-party Backend as a Service (BaaS) services, and/or that includes custom code run in managed environments on a FaaS platform. In many ways, serverless architecture looks similar to other application designs focused on events and microservices.
Using FaaS for isolated use cases or playing with it test environments does not require an organization to rethink the way it writes code or manages infrastructure. But, without re-factoring an application, FaaS can easily increase computing costs when scaled for production use. With many other challenges arising when FaaS moves into production, it is not surprising that almost all organizations with broad deployments are using unique architectures for serverless applications.
It is not unexpected that 95 percent of organizations with serverless architecture are using FaaS to some degree because of the way the terms were defined. Despite attempts to educate the market, we still believe the word “serverless” connotes many different things, especially for the 79 percent of organizations that plan to adopt serverless architecture but have not planned to use FaaS in the next 18 months. For these people, we believe they are enamored with the larger benefits of cloud-native development as opposed to using a shiny new toy like AWS Lambda. In future surveys, we will remove FaaS from the definition and see how FaaS adoption is associated with event-driven and microservices architecture.
Although 63 percent of respondents without plans for serverless architecture plan to start using FaaS, it is hard to determine the likelihood of those plans coming to fruition. Which brings us to a final note of warning to about the speed and scale of FaaS adoption. Bottoms-up exploration by development teams is just the first stage, with architectural considerations needing to be addressed before FaaS moves into production.
Serverless consultant Mike Roberts assisted in the definition of terminology used in the survey.