Cloud

Serverless architecture is the future, but we're not getting rid of containers just yet

Containers are billed as the future, yet serverless already seems set to surpass them in importance. Here's how organizations will juggle them both.

"Using containers today is like using CloudStack or OpenStack in 2012," argues Cloud Opinion, they "cannot be your long-term tech strategy."

Can this possibly be true? Just as Kubernetes and the container revolution rumbles to a thundering crescendo, could it be that serverless offerings like AWS Lambda have already moved past them?

Definitely. Maybe.

Regardless, the future takes time. Serverless looks certain to dominate the future—it's simply too convenient for developers to overlook—yet containers may well give enterprises an easier way to transition out of their VM-laced past. We're likely to see a whole lot of serverless functions running side-by-side with container-driven apps and even some OpenStack thrown in. Oh, my.

It's the future!

Remember OpenStack? It was supposed to be a magical open source elixir for turning old-school enterprise data centers into new-school "private clouds." It didn't work. Instead of OpenStack displacing AWS, the dream of many a legacy tech vendor, AWS mowed it down, bringing along an army of developers who sought the true promise of the cloud: Convenience.

And yet...

Listen in on Red Hat's latest earnings call and you'll see plenty of OpenStack adoption. While significant traction for OpenStack has come from the telecom industry, as big vendors like AT&T spend 8-figure amounts with Red Hat to try to reshape their aging infrastructure, 60% of Red Hat's largest OpenStack deals in its third quarter came from finance, media and entertainment, and other non-telecom markets.

SEE: Job description: Cloud engineer (Tech Pro Research)

Ironically, the pace of adoption for OpenStack might actually be picking up. As Red Hat CEO Jim Whitehurst articulated: "[W]e're starting to see real growth on the enterprise side." Why? Because if you're stuck with on-premises workloads, for whatever reason (many bad, some good), "[Y]ou're either kind of stuck in the old, expensive world of VMware or you're moving to OpenStack if you want scale out infrastructure."

We may contemptuously sneer at OpenStack as yesterday's tech for yesterday's enterprises but it still represents an upgrade on what most enterprises have, and an easier upgrade path than jumping straight from mainframes to AWS Lambda. It is, in short, a pragmatic choice, even if it's not an elegant one.

Which brings us back to containers.

Cloud convenience...realized

The primary driver behind cloud computing has been convenience: Developers haven't wanted to wait on IT to provision servers, line up software licenses, etc., so they've pulled out a credit card and jumped in. As cloud adoption has grown, this basic premise has largely remained true. Serverless takes the promise of convenience to the nth degree, removing the need for developers to futz with infrastructure at all, and instead focus on writing their apps.

SEE: Serverless but not stress-free: enterprise computing moves outside the enterprise (ZDNet)

Cloud Opinion is therefore correct to argue:

If you thought Cloud was inevitable, the drivers behind serverless are even more compelling. Migration to Cloud from [datacenter] was a harder choice than migrating from a VM-based app to serverless app. Developers will start experimenting with things like [AWS] Lambda and will end up liking the potential of it.

Some, like Expedia, like it a lot. As Expedia vice president Subbu Allamaraju told me, "Our teams at Expedia did over 2.3 billion Lambda calls/month" back in December 2016, a number that has mushroomed since then. That was early in the serverless shift. Many more companies (and their developers) have uncovered the promise of serverless since then.

Indeed, Allamaraju went on to blog that enterprises should scrap their private data centers, powered by OpenStack or anything else: "Unless you've at least 200,000 servers in multiple locations, or you're in specific technology industries like communications, networking, media delivery, power, etc, you shouldn't be in the data center and private cloud business." Yet many companies are in that private cloud business, and most are now discovering just how liberating containers can be. Even so, Cloud Opinion is right to argue that containers are "Exciting, but [with] limited future potential." Why? Because they force enterprises to spend too much time in the weeds of their infrastructure, rather than on building the differentiated apps that their customers need.

And yet...

The future takes time

Enterprises will continue to use OpenStack, and they'll spend years with Kubernetes managing their containers. They'll also increasingly invest in serverless. They'll do all of the above, and probably still muck about in green screens and mainframes. Why? Because the enterprise is messy, slow, and complicated, and it's never as easy as "stop everything and do [insert containers, serverless, or whatever the currently hyped alternative is]."

In short, the future is serverless, but it will take enterprises a long time to get there. In the meantime, there's money to be made selling that future (AWS, Microsoft Azure, Google Cloud), money to be made selling the bridge to the future (Red Hat) and, sadly, there's even money to be made selling the past (names withheld to protect the not-so-innocent legacy tech vendors).

Also read

Servers in cloud data
Image: iStockphoto/Trifonenko

About Matt Asay

Matt Asay is a veteran technology columnist who has written for CNET, ReadWrite, and other tech media. Asay has also held a variety of executive roles with leading mobile and big data software companies.

Editor's Picks

Free Newsletters, In your Inbox