Despite proclamations from Amazon.com Chief Technology Officer Werner Vogels that usage of Amazon Web Services’ Lambda usage has risen dramatically, recent surveys of Kubernetes communities have found relative declines in the adoption of the serverless service. As the number of cloud native organizations continues to grow, they are slightly less likely to include the AWS serverless compute technology in their plans.
Two-thirds of the Cloud Native Computing Foundation’s (CNCF) June 2020 survey of its Kubernetes-centric community is using or considering using serverless technology. But, strip away the evaluations and theoretical plans and only 30% are using serverless in production, a far cry from the 92% of respondents with containers deployed in production.
Overall, 46% use at least one hosted platform (e.g., AWS Lambda, Google Cloud Functions) or installable software (e.g., Knative, OpenFaaS). That is exactly the same level as in the 2019 survey according to our calculations. AWS Lambda adoption dropped from 67% in 2019 to 55% in 2020 as organizations evaluating serverless are somewhat less likely to have it on their roadmaps. Meanwhile, Microsoft’s Azure Functions made significant progress, going from 16% to 24% year over year.
The installable software option implies that an organization will host its own FaaS solution, perhaps so it can deliver functions via on-premises infrastructure and/or a hybrid cloud. Only 17% of the study is using this approach, but Knative is the most popular option, followed by OpenFaaS. Both of these options are oriented towards Kubernetes, but to what extent are non-Kubernetes-related options being adopted? That is a difficult question to answer, especially since this survey was fielded by the organization that manages the Kubernetes community.
StackRox, a container security company, conducted a series of surveys of container user that support these conclusions. In their reports, AWS Lambda adoption stalled, going from 37% in Spring 2019 to 38% in Fall 2020. Meanwhile, Azure Functions rose from 16% to 25% and Google Cloud Functions went from 9% to 15%.
Blurring the Lines Between Containers and Lambdas
Remember FaaS ≠ AWS Lambda. We are only digging into Lambda adoption because AWS’s more publicly available analysis is available because the community of users emerged earlier.
Sumo Logic’s analysis of its AWS customers’ activities found a modest increase in those running Lambdas, going from 36% in 2019 to 39% in 2020. In early 2020, Datadog reported that more than 40% of its AWS customers had adopted Lambda, with that figure almost doubling among AWS customers that also used containers.
Although AWS Fargate runs on top of Lambda, this shouldn’t count when calculating how many organizations use the service because developers aren’t actively writing applications that call functions. As far as we can tell, the analysis from the aforementioned monitoring providers excludes this type of use case, but it is important to note because almost half of the services Amazon (not just AWS) developed in 2020 utilize Lambdas. Growth of Fargate and other AWS services may be a significant reason why the number of AWS Lambdas being invoked has reportedly skyrocketed.
AWS continues to blur the lines between containers and Lambdas with its announcements at this year’s AWS re:invent. AWS Proton will allow platform engineering teams to manage the infrastructure as code necessary to deploy both container-based and serverless applications. In addition, Lambdas can now be deployed as a container image.
Serverless evangelist Ant Stanley believes “containers + serverless is the norm, not the exception.” Containers make legacy applications more efficient, while new applications tend to use FaaS and event-driven architecture. From this perspective, organizations that have pushed towards cloud native application development, which too often narrowly defined as container-based, will inevitably start to write applications that utilize functions. This view makes sense, especially in a world where large swaths of the corporate world have already made a leap towards containers.
Huge Serverless Increases
While AWS is eating its own dog food, early adopters have also scaled up their usage of Lambda. Earlier in 2020, New Relic reported that organizations with AWS Lambda in production had approximately doubled the number of functions they used and how often the functions were invoked each week. The increased number of functions means developers are writing more applications. Since the number of functions invoked is an expression of application activity, the reported doubling indicates applications are scaling up. By these measures, serverless adoption surely rose, and we have no reason to doubt that its growth has slowed.
Exponential growth is impressive, but remember it is coming from a small base of applications and use cases. Still, our analysis of surveys in 2018 and 2020 indicates that early adopters expect to continue to build new applications despite the production-level challenges they face. To measure this type of growth we need to ask about the percentage of applications that utilize event-based architecture and how many applications require a function to be called.
Measuring serverless adoption is an imperfect science. This entire article takes a cloud-native slant, but companies like EDJX think about FaaS in terms of edge computing. From this perspective, developers focused on IoT, XR (alternative and virtual reality) and other emerging use cases should deploy distributed applications utilizing functions that run close to where devices are located. In the future, there is a real alternative execution venue. Major cloud computing providers and containers do no have a monopolist grip on the IT world.
Need for More Research
In terms of surveys, there is an obvious need to look beyond the opinions of backend developers, platform engineers and people with a focus on containers. Additional surveys should recruit developers that work on IoT, mobile, gaming, and less enterprise-centric use cases. The New Stack looks forward to reviewing updated versions of New Relic and Datadog’s serverless reports. Machine-generated data can also be utilized by analyzing the number of package downloads used in popular frameworks. The number of questions asked on StackOverflow can also be tracked. What is your favorite metric?
AWS, New Relic, and the Cloud Native Computing Foundation are sponsors of The New Stack.
Feature image via Pixabay.