By Alex Ellis, Berndt Jung and Mark Peek
Is it right to say that serverless technologies are overhyped? It probably is. But that’s a good thing. Now’s the time to invest in them for the long haul.
If you care about the success of serverless technologies, recent moves by the Cloud Native Computing Foundation (CNCF) have been encouraging. Earlier this year, the CNCF Technical Oversight Committee’s serverless working group released a white paper on the subject, along with a landscape survey detailing the major tools, frameworks and platforms in the space. These efforts both reflect and reinforce growing industry interest in the field.
The CNCF effort also spurred interest in—and new efforts toward—greater harmonization across different Functions-as-a-Service (FaaS) implementations and services. The initial work has focused on events and led to the advent of CloudEvents as a way of standardizing how events are transited from one service to another.
But does all of this attention mean that the serverless sector is in danger of overheating? After all, just a few years ago, the landscape encompassed barely a half dozen projects that were fairly stagnant. Now it’s closer to 60, and most are growing fast. At KubeCon, Chad Arimura suggested serverless might be approaching the first optimistic peak of Gartner’s classic Hype Cycle. Certainly, serverless now has tremendous visibility and interest. And it would be right to expect it to drop down into the Trough of Disillusionment before long. Does that mean it’s time to pull back? We don’t think so.
We’ve been interested in the serverless space at VMware for a few years now. We built our own internal FaaS implementation which found a home in the Office of the CTO’s xLabs group. Given the work being done in open source, we then decided to embrace and extend this work with other open source communities. Last January, the team launched an open source project called Dispatch, which offers an enterprise-quality framework for deploying and managing serverless style applications.
At the same time, VMware’s Open Source Program Office has been supporting OpenFaaS, which lets you package anything as a serverless function. Dispatch now uses OpenFaas as its default, but it can interoperate with other FaaS implementations, such as Project Riff, and is now adding kubeless support.
Many customers we talk with are aware of the hype around serverless and are waiting for things to settle before they fully engage, or are using a FaaS for one specific use case (sometimes without even realizing that the app rests on serverless technology) and not looking to go much further for now. But we’re looking ahead to the next phase after the hype has died down. That’s when we think serverless will start moving up Gartner’s Slope of Enlightenment. It will come when users can both control their own FaaS in a public cloud and operationalize it on private clouds.
The path there requires being able to move any event out of any cloud so that it can be consumed easily on any platform, including on premise environments. At the moment, this is so difficult to do that many people aren’t even trying. But once it becomes possible, it will unleash the true value and potential that serverless technologies possess.
To that end, VMware is actively engaged in supporting the CNCF serverless work group as it builds out harmonization, along with its future efforts around function signatures and solutions for chaining functions together. We’re advancing OpenFaaS to free more user data from being locked into any one platform so that companies can run functions in the way they want and in the languages they prefer. We’re also building out Dispatch as an open source mechanism by which enterprise users can interact with existing FaaS solutions with the scalability, reliability and security that they need.
Here’s the bottom line: we want our customers to be able to run any workload they need to run wherever they want to run it. Our work in serverless aims to keep the industry moving toward that goal.