Serverless deployment technology won't put IT ops out of a job, but it will significantly change what that job requires.
Serverless computing -- the practice of writing individual application functions for automated execution and scaling on a cloud back end -- sure sounds like it could send IT pros into early retirement. But a smarter bet says serverless will be just another weapon that sits alongside containers and VMs in the app-deployment arsenal.
"I don't necessarily see it as something that's going to be the dominant architecture and take over the world," said Hardip Singh, mobile development manager at The Washington Post. The news organization has deployed the Lambda serverless computing product from Amazon Web Services (AWS), but it's not suitable for everything, Singh said.
That doesn't mean serverless deployment is something IT ops can downplay, as it will still be about 25% of what they have to deal with in the future, Singh said.
Sys admins should become experts on cloud services, including serverless deployment, so they know what service to recommend developers use for different apps, Singh said.
"If I was a sys admin, I would focus my training on becoming an AWS cloud expert," he said. "Then, I can go into any company and they can present me with their problem statement, and I can guide them toward the correct solution."
Serverless deployment ops opportunities
Serverless deployment offerings will undoubtedly change rapidly. But, for now, services such as AWS Lambda and IBM OpenWhisk require some setup configuration for security. Upon deploying a serverless function or application, the administrator must enter Amazon's Identity and Access Management service or Apache OpenWhisk identity credentials to give the services access to back-end databases or front-end web servers.
"You need IT ops in some pure serverless cases to set it up," said Joe Emison, CIO for Xceligent Inc., a commercial real estate data services provider in Blue Springs, Mo.
Monitoring is another area of opportunity for IT ops in serverless deployment scenarios.
"Ops teams focus on very important issues that are required for serverless to be successful, which are: How do you test and monitor and manage the health of the serverless infrastructure?" said Greg Arnette, CTO of Sonian, an email archival and analytics cloud service provider in Waltham, Mass.
Sonian has AWS Lambda in its development and test environment now, and it plans to put it into production this spring.
"It's going to require the evaluation of new tools, scripts and ways of understanding, because you're no longer logging in to a virtual machine box with a command prompt as a way of getting access to the system," Arnette said.
With monitoring responsibilities comes the need for IT ops to troubleshoot configuration issues that may arise in serverless deployment environments.
This is no easy task, as microservices proliferate. And, in some cases, DevOps pros will be called upon to invent ways to discover serverless deployment metrics.
Aater SulemanCEO, Flux7
"Hopefully, we see a lot of growth this year in tools to help keep track of this ever-increasing number of moving parts that are getting tinier and tinier," said Michael Bishop, CTO of financial technology startup Alpha Vertex Inc., in New York. "It's crazy to attempt to do logging and trace through those kinds of functions -- it's like looking at a swarm of bees and [working] out what's going on."
But with monitoring and analysis come new ways of adding value to one's organization, said Aater Suleman, CEO of Flux7, an Amazon consulting partner in Austin, Texas. He cited the example of a user who was able to calculate the exact cost of each customer transaction by analyzing serverless deployment data.
"It's always been black magic how you distribute infrastructure costs, even in the data center among applications. And [this client is] talking about a granularity that has never been possible before," Suleman said. "It's an advantage for serverless from a business perspective."
On-premises serverless deployment: The new PaaS?
At some large enterprises, IT ops pros may even be asked to offer an internally managed serverless environment for their developer team members. Experts see this scenario unfolding in a similar manner to how large organizations adopted platform as a service: as a way to offer developers self-service access to infrastructure.
Here, as with monitoring, IT ops pros have the chance try something new.
"The value is in having a function respond to any event in your infrastructure or your application space, then trigger other actions," said one senior vice president with a global financial services company who declined to be identified. The executive wants to construct an internal serverless deployment environment.
Container platforms, such as Kubernetes and Mesos, have built-in event structures that would be a natural fit for open source AWS Lambda alternatives, such as Apache OpenWhisk, the senior vice president said. But this will take IT operations expertise to build.
The potential benefits for internal serverless deployment environments are significant, said Travis Reeder, co-founder and CTO of Iron.io, a San Francisco company that offers IronFunctions, an open source serverless deployment project that can be used as an alternative to Lambda on private and public infrastructure.
"With serverless, you can have hundreds of apps or APIs able to deliver or respond pretty quickly, but they only execute when they're running, so they can share resources a lot better," Reeder said.
The differences between IaaS, PaaS and serverless
Cloud providers compete on serverless playing field
Does no servers equal NoOps?
Orgs find a lot they don't like about Lambda