FIRST YOU HAD A SERVER running one operating system, then it ran several virtual operating systems, and then the hardware moved to the cloud. Now even that virtual server in the cloud has disappeared with serverless computing.
“Serverless computing focuses on the app itself, after abstracting the hardware and now even the operating system, and you just have your code,” says Kong Yang, “head geek” (yes, that’s his title) of SolarWinds Worldwide LLC, in Austin, Texas. “Serverless computing is your code, period. Everything else is abstracted and handled by your provider.”
Craig Lowery, a research director at Gartner, goes even further. “The servers are there but are controlled by the provider. I can’t see them and can’t get to them even if I want to,” he says.
Why this next step? Cost reduction for one thing. “You can save 90 percent over costs of a virtual machine,” says Bernard Golden, a cloud consultant who is also author of Amazon Web Services for Dummies. “That’s cutting 90 percent of cloud costs, not on-premise costs.”
Serverless computing takes “modular programming” about as far as it can go, too, which makes for easy collaboration, according to Jon Hathaway, founder and CEO of HATech LLC, a provider of DevOps and cloud migration services based in Las Vegas. “If one team prefers coding in Python and the other prefers .NET, they both can develop in their strongest technology stack,” he says. Functions will execute when called, no matter the language.
“Serverless isn’t just another VM or compute model,” Hathaway adds. “To realize the cost benefits, applications must be rearchitected. This means serverless suits new applications more than migrating legacy apps.”
Lowery recently did a case study of five companies testing out serverless computing. “The first time you try, it really fries your brain,” he found. “It forced these companies to think more correctly about designing microservices and how to think in a cloud-services way.”
Channel pros who don’t code but do resell for Microsoft will notice something beyond servers missing with serverless computing: licensing revenue from virtual machines running Windows Server. “Microsoft still regards that revenue as a core product,” says Lowery. “But they are positioned well with Azure Compute and are turning the corner on consumption-based pricing.”
Since every cloud provider has some unique features, Yang warns that vendor lock-in can be a real issue. “When your code is optimized for AWS, you can’t just move it over to Azure,” he says. “They have similar features but the constructs are not identical. In addition, some enterprises don’t have the personnel or talent to code and chain functions together.” There can also be issues monitoring and troubleshooting apps and functions, Yang adds.
Golden points to IoT as a good fit for serverless applications. “Serverless functions are often used for functionality after being triggered by some event, like a thermostat hitting 75 degrees,” he says. “That kicks off an event message to a serverless function that turns on the air conditioning.” Rather than paying for a server to host an app that watches thermostats, developers can make that code a serverless function that doesn’t cost anything until it’s triggered and used.
That issue brings Yang to a new market for resellers. “You can offer professional services,” he says. “More savvy channel partners will have their own in-house teams of experts with expertise in certain functions, like specific verticals. If not, look to partner with a software group that does have the needed expertise.”
In the short term, serverless computing won’t affect resellers any more or less than the public cloud, says Lowery, who predicts that it will keep growing like the public cloud too, for one simple reason.
“You only pay for each millisecond your program runs,” he observes. “What customer doesn’t love that?”