As the lead story of SiliconANGLE’s Cloud Special Report makes apparent, serverless computing is the next big disruption the cloud is bringing to computing, freeing software developers from the hassle of managing hardware, software and networking technologies.

Essentially, serverless technology executes software functions automatically based upon an event triggered by a human or by a program, allowing services to scale up much faster and saving a lot of money in the process. It may prove to be the prime way software gets built in the future — not just by startups but by arge enterprises such as CapitalOne, Hearst Corp. and the Financial Industry Regulatory Authority.

One company that has led the way to serverless is Amazon Web Services Inc., whose Lambda serverless platform kicked off a lot of interest when it was launched four years ago. In a conversation with SiliconANGLE last week, Tim Wagner (pictured), general manager of AWS Lambda and related serverless services, shed light on what serverless computing means, what AWS offers in serverless, the surprising uptake by large traditional companies and what’s coming next (a supercomputer in the cloud?). This is an edited version of the interview:

Q: Partly because of the confusing name, it’s not entirely clear what “serverless” means. How does Amazon view it?

A: We think of serverless, especially with regard to Lambda, as three things. First and foremost is that there’s no infrastructure to manage. That’s where, really, all the things in our serverless portfolio kind of look similar, and we’ve been at this game for a long time. S3 is essentially serverless blob storage, going back a decade or more. Things like Fargate are more recent entries. We very much now look at it as, “How can we build out that portfolio at that higher level of abstraction?”

The second piece of that definition, moving more into the compute category, is that Lambda is functional and event-based. Those make it different, for example, from models of where you bring your own operating system image, like Fargate or other kinds of container-based systems. That’s the difference in programming model and approach.

Q: How so?

A: One of them is, really, about being able to lift and shift or repurpose existing server-based software, but with Lambda we’re intentionally trying to break that model. We’ll take those requests, we’ll handle that event ingestion, we can actually provide a lot of value by managing that for you, and then you get to focus on the business logic. That doesn’t very much change, in the way that we think about the application programming piece of it. Obviously, the code that people bring, what they write in those functions looks very conventional, you don’t have to learn a new framework, or a new language, obviously, but we are changing the ways that you get access to it.

Q: And the third piece?

A: The third piece is about unbounded scaling. By having the economies of scale, we can offer lots of customers the ability to burst through capacities without having to provision for that. That’s good for customers, that saves them money, that saves them time and energy. Customers like Vevo, who have an 80 times peak-to-average ratio, for example, don’t have to spend for that overload in order to be able to be safe for it.

Q: Which of those factors is most important to most customers?

A: There are certainly two pieces that we heard from developers and companies of all sizes: “reduce operational costs” and “faster time to market.” You can take a look at companies like Agero, for example, which does the roadside assistance call infrastructure. They were able to bring an app to market way faster than if they had normally built it out, through a serverless approach. We definitely see that faster time to market, and for companies like iRobot and Bustle, who have gone all-in on serverless, it’s dramatically reducing their operational overload. That whole piece of the pie that used to be, people who had to be dedicated to the care and feeding of the server fleet, they can now spend their time on, hopefully, something far more differentiating, which is making sure the application itself is working well, whether that application is a robot vacuum, or a fantastic web content experience for Bustle.

Q: What about even larger enterprise customers?

A: The thing that often attracts the enterprises, especially the big high-volume, high-revenue players, tends to be cost-savings. Hearst is a great example. They took this old-school, fixed-hour Hadoop [big data analysis] job, and they rewrote it as a serverless analytics pipeline, so they can get the result very quickly, in under a minute. They also got a 95 percent cost savings when they did that. It wasn’t just dramatically faster, it was dramatically cheaper as well. That’s not atypical.

We have this little slogan we use with Lambda, “Never pay for idle,” and that has a real technical meaning. In a serverless approach like this, you only run, and therefore you only get billed, when you have work to do, and that’s very different than the conventional model of standing up servers, deploying stuff to them, and hoping that it works, and then having to try to scale them.

Q: How do companies generally move to serverless?

A: There’s an adoption pattern we often see, an initial DevOps adoption, maybe people will begin running serverless cron [scheduled task] jobs, and then it will move into more back-office situation, and then eventually the more mission-critical systems.

Q: Is serverless going to become the main way to create applications and run them in the cloud, or is it just another tool in the arsenal?

A: If you have traded any stocks, or had any stocks traded in your behalf, FINRA processes those stock trades at the end of the closing day using Lambda, so there’s a big chance here that the trade you made was evaluated and validated by FINRA using Lambda. Thomson Reuters does four thousand transactions every second with it, Fannie Mae runs its 20 million mortgage calculations through there.

So these are not ancillary, some line-of-business, or over-in-the-marketing-department kinds of pieces. These are mission-critical software that is now tied at the hip to a serverless architecture. We think of it as one of the major ways that people are going to use compute, and Lambda is certainly one of the fastest growing businesses within AWS.

Q: It’s not yet for every application, though.

A: There are certainly, today, pieces that hold some companies back. Sometimes, that’s the usual sort of fear or doubts about a newer technology. Other times, it’s the point-in-time piece, like running the longer running times or, last year, we doubled the memory so that people could bring larger, more complex workloads on Lambda. You should certainly imagine that those things that look like a general challenge will get solved over time.

Just like assembly language got replaced with high-level languages over time, I think you should expect that infrastructure is the assembly language of the cloud, and we will see people coding in these higher-level patterns, for the same reasons that they prefer to code in Java, or Python, instead of machine language.

Q: An example?

A: I saw high-school students [at] a hackathon had written an Alexa skill. These high-school students had possibly never heard of EC2, certainly didn’t know an availability zone, or a network configuration, but they’re able to get an advanced, human language interaction and resulting code, written and running, at scale, in the cloud, using things like Lambda and Alexa.

The serverless services like Lambda are an increasingly large percentage of where people begin to use the platform. We’re having that effect of actually expanding the demographic, not just giving optimizations to people who already use the platform.

Q: What about enterprises that aren’t cloud-first?

A: Thinking about how to help customers migrate and repurpose existing pieces is one of the things that we’re most excited to go after. Unfortunately, this kind of “meeting customers where they are” is a multidimensional problem. One piece of it is making sure that they have the tools necessary, [such as] the ability to put it into a pipeline really easily.

Honestly, most of our big customers, they are all hybrid at some level. They all have servers, obviously, and they all have serverless, and often the two are playing together, in terms of a solution. I do think you will also see startups, and others, and likely us continuing to invest in helping customers with migration tools, with ways to harvest their existing code. The one that we’re probably the most focused on right now is, make serverless play really, really nicely in that IT sandbox.

Q: It must be difficult for companies with working, mission-critical systems to shift over even with the apparent benefits.

A: One of the things that we’re seeing, and one of the reasons that enterprise is the fastest-growing segment for Lambda today is that some of those companies, or those pieces of them that have been a little bit slower, are choosing to skip some of the intervening technologies, and swap straight to serverless. They see it as a way to outsource a lot of the IT challenge.

Q: Any examples you’d point to?

A: Where a lot of the revenue comes from … is big, event-driven, fast pipelines of data that is sitting in enterprises, analytics data, companies like Hearst doing analytics processing. Fannie Mae is a great public-sector example where, they do these Monte Carlo simulations, and it was a big on-premises exercise, and they were starting to outgrow it.

To move that process straight to Lambda, skipping every technology in between, now … what used to take four hours takes an hour and a half, they can extend capacity without going and purchasing new hardware, and waiting for that to land. Especially in the public sector, you wouldn’t necessarily have thought, “Wow, they’re going to be one of the first and one of the largest adopters of the serverless approach,” and yet, here they are.

Q: What’s coming next in serverless in terms of new capabilities?

A: You’ll see, probably, more things getting built after this level of abstraction, where the infrastructure is hidden from the end-user. We’re just at the part of the year where we do our business planning, and without going too much into the secrets of the roadmap, I think I can probably safely say that you will see us spending a lot of energy in continuing to enhance the programming model for Lambda.

We’ve got some of these things we’ve talked about, like big data and machine learning, that is clearly in our sights, and we want to make sure that, whether you’re writing an enterprise app that regulates data, or an application for a mobile or web user, or whether you’re treating Lambda like a supercomputer the way folks at Fannie Mae do, or whether you’re doing lots of analytics processing, and you need to make sense of that data for your business purposes — all of those things should be able to be done, efficiently, easily and simply, on a serverless stack with Lambda.

Q: Any other limitations you need to overcome?

A: We don’t think that there’s any inherent technical limitation that will keep any of the major workloads from being used on Lambda. Today, obviously, if you have something that takes five and a half minutes, and we only let you run for five, that can be a limitation. You should expect that those are very much point-in-time challenges that we’ll get past. Many of those are things that we’re already hard at work on.

We definitely believe that for what we think of as business logic, mainstream computation, serverless will be the easiest, the simplest and, for many customers, the preferred way of doing that in the future.

Photo: Serverlessconf/YouTube