The serverless revolution isn’t coming soon, it’s already here!
Our recent Serverless NYC 2018 show was packed due to a wide range of great speakers. After a full day of serverless insights, our big takeaway is that serverless looks primed to become the fastest growing new application architecture to be adopted for both new and existing development, primarily because it lets developers focus on building and running auto-scaling applications without worrying about managing servers, as IT operations are all taken care of behind the scenes.
The great lineup of vendor presenters included key serverless solution providers (such as Google, IBM, Microsoft, and Iguazio) who directly create and deliver serverless solutions to the market. While this is an emerging and competitive market, our key speakers agreed that committed developers already have all they need to successfully build large-scale serverless applications today.
To prove that point, several real-world enterprise serverless champions presented how they are deploying significant production applications completely built out of serverless functions. In fact, more than one presenter explained how they’ve already moved their entire IT architecture into a serverless cloud and are now benefiting from vastly increased development agility, TCO savings and elastic scalability.
Ben Kehoe of iRobot explained that once it went 100% serverless – no VMs, machine instances or containers – iRobot significantly lowered its IT costs, eliminated operational burdens to now bring new solutions to market faster. Kehoe spends most of his time improving business value instead of dealing with infrastructure firefighting. Watch Ben's presentation here, or read account of it by Sara Castellanos in the Wall Street Journal.
We heard from Ben and other speakers that not only is serverless design the likely future for all scalable applications, but it may just be that containers and DevOps have already had their day in the sun. While it’s true that less than half of enterprises have yet to deploy production container solutions, the sense of the show was that having to even manage Kubernetes simply shifts the legacy IT infrastructure burden up to the application server level. Unlike server virtualization or containerization, serverless can truly change the fundamental relationship between cost, service and operations for the better, especially if IT can get rid of, or let go of, all infrastructure management (i.e. “no servers to operate at all”).
Serverless is Serving More and More
Of course, we heard that making this paradigm shift is still challenging. Our keynote speaker, Kelsey Hightower from Google, related some of his hard-earned wisdom as an early serverless adopter. Despite presenting much practical advice and many warnings about some serverless speed bumps to watch out for, his enthusiasm for the serverless approach was infectious.
Kelsey pointed out that serverless design in general is not a new idea, but one that has been evolving for decades – even demoing a 70’s era superdeamon (i.e. xinetd) and talking through some of the intermediate computing approaches. Still, today’s serverless FaaS (“Functions as a Service”) offerings change the game, promising a nirvana of riches to application developers (and absolution from the responsibility of operations of any kind).
As a pragmatist, Kelsey said that while developers do it all for data, he’s “not sure it’s 100% true that we don’t have to focus on infrastructure at all.” After all, we will still have to look inside any black boxes and will always need ways to troubleshoot, trace and debug our code. But the good news from his perspective is that “the future has a lot of problems to solve!”
What Are You Buying When You Aren’t Buying A Server?
Another highlight of the show was our panel about open source serverless. One challenge is that while there are several FaaS solutions available, there as yet are no fully adopted serverless standards. While each serverless provider can implement their own internal IP and bring various vendor-specific features to market, many developers would still like to see both more standardization and an open source foundation, akin to the big data market.
As a counterpoint highlighting why open source is important, IBM’s Dave Grove outlined how IBM sponsors Apache OpenWhisk to offer consistent serverless environments, both in its public cloud and as an open source solution for development and hybrid/private cloud building.
Yaron Haviv, Iguazio’s CTO, also discussed Nuclio, our open source serverless solution purpose-built to support real-time and data-driven applications. Iguazio built Nuclio to serve as a platform for real-time machine learning and AI, natively integrating with data scientist-friendly Jupyter notebooks which can then run consistently and seamlessly across edge and multi-cloud. The entire panel is available here.
There were a great many other presentations during the day, including inspiring stories from serverless adopters such as Tyler Love from Bustle, Jason Katzer and Angela Razzell from Capital One, Orit Nissan-Messing from Iguazio and Gwen Shapria from Confluent, to name a few. Below are some of their recordings. Mark Boyd posted about Gwen's presentation on The New Stack and there's a great post about the entire conference on Jeremy Daly's blog.
IT organizations are looking to lift themselves out of the weeds of infrastructure management and focus 100% instead on providing business value. Serverless will be the easiest infrastructure to operate and here at Iguazio we are proud to be at its forefront. We'll see you again next year!