Has serverless computing run out of gas?

Serverless computing offers the promise of applications executed as event-driven compute cycles, whereby your code runs only when triggered by a request.

Subscribers to this paradigm can save money by paying only for compute time actually used instead of maintaining a virtual or physical server. Users also are spared from having to manage infrastructure.

AWS Lambda, Microsoft Azure Functions, Google Cloud Functions, and other serverless compute offerings have attracted a lot of attention in recent years. But the serverless revolution may be showing signs of stalling, with some shops not exactly falling all over themselves to adopt the paradigm.

In a report on cloud adoption published this spring, O’Reilly Media found that, among participants surveyed who had not adopted serverless, most—about 62 per cent—had no plans to so so. “It certainly feels like serverless has lost its steam, based on the conversations we’re having,” said Mike Loukides, vice president of content strategy at O’Reilly Media.

Loukides noted that serverless has fallen short of growth expectations. “Some of that is due to those technical issues that won’t go away,” he said. “Designing systems that can tolerate lots of latency is a big architectural challenge. But it’s pretty clear that a fair number are making it work.” Whether they have dealt with the architectural issues, ignored them, or have a use case where it doesn’t matter is an interesting question, Loukides said.

Peder Ulander, product market lead for enterprise and developer audience at AWS, which offers the AWS Lambda serverless platform, disputes the notion of any kind of lull in the serverless realm.

“We’ve been very happy with regard to the growth of our serverless business and seeing it becoming more and more strategic,” Ulander said. He added that serverless reduces overall costs and improves developer productivity. Hundreds of thousands of AWS customers are using AWS Lambda.

Inquiries about serverless have increased this year at Gartner, analyst Arun Chandrasekaran said. There is a class of applications that serverless is extremely good at, particularly event-driven architectures, with serverless adding a substantial advantage, he noted. Highly ephemeral applications that live for a few minutes or a few seconds are also suitable for serverless.

But Chandrasekaran cautioned that new technologies like serverless often go through a hype cycle before customers realise what the technology actually can do. “In many cases, customers may be trying to use the technology beyond what it’s meant for,” he said.

Andrew Davidson, vice president of cloud products at MongoDB, concurred that serverless was going through a hype cycle but stressed it was still in its early days.

“There’s plenty of people who haven’t even started using serverless,” Davidson said. Serverless has worked for some situations but not for others, he said. “It can be used to augment specific, targeted use cases, providing ancillary application services,” he said. MongoDB has a serverless service native to its MongoDB Atlas database service to extend the value of the database.

Gartner’s Chandrasekaran believes challenges in the serverless ecosystem today include application debugging and testing, monitoring, and security. More maturity and tools are needed for this, he said.

Serverless ecosystem vendors need to provide the same tooling in edge environments as in public cloud environments. AWS’s Ulander, meanwhile, expects that as serverless advances, there will be more integrations around containers or Kubernetes.