Nobody ever believes me when I say this, but the world of enterprise technology has actually been pretty exciting over the last few years.
The rapid changes in the technology behind nearly everything you do with a computer can be staggering to comprehend, even for people who live and breathe this stuff, and these shifts are not always forgiving. It doesn’t take long to move from being the dazzling new kid on the enterprise tech block to a sunk infrastructure cost your developers can’t wait to get rid of.
That’s not exactly the case when it comes to software automation company Puppet, but it’s also fair to say that the sexier parts of enterprise technology have shifted away from Puppet’s established business, helping companies develop software faster and more reliably for their own self-managed servers. The cloud, containers, and emerging technologies like serverless computing have changed the game for everyone in the software factory tools business, and Puppet is going through a change in product direction led by Omri Gazitt, who joined the company in February 2017.
Puppet, founded in 2005, has more than 500 employees and has raised more than $107 million debt and equity financing. Puppet cut 3 percent of its workforce last month as part of what the company called a “realignment of our workforce will allow us to innovate more rapidly on behalf of our customers.”
Gazitt and I recently sat down at Puppet’s offices in downtown Portland for a wide-ranging discussion on the state of Puppet’s product strategy and this world of enterprise tech at large. The transcript that follows was lightly edited for length and clarity.
GeekWire: So when you came in, what direction where you given? What were you asked to do as part of your tenure at this company?
Omri Gazitt: When I first talked to Sanjay (Mirchandani, Puppet CEO), there were a couple of things that he had on his mind. One was with Luke (Kanies, Puppet founder and former CEO) moving on from the company — staying on the board, staying involved, but not operationally — there’s more than the formal role of CEO that we have to fill. There’s also the focus on product innovation and strategy, and Sanjay was looking for a partner for that and that’s exactly what I was looking for.
But even more attractive was the fact that Sanjay felt a very strong imperative to be not just relevant, but essential to the world of developers, the world of cloud, the world of containers.
The idea of building an automation platform was a very, very strong one, and I saw a company here at Puppet that had that exact aspiration, to become the automation platform for the industry. If I had to describe how I think about it and how we think about it as a company, we want to go build a comprehensive automation platform. When I say automation, it’s not just about wrapping shell scripts and being able to execute them in a one-and-done fashion. We think about automation as a continuous process.
So you see that with every single product that we put out. The product that we put out just now, Puppet Discovery, it’s a continuous discovery product. The whole point of it is to keep the picture of what IT you have running up-to-date. The current solutions that you look at, traditional solutions, things like CMDBs and so on, they get out of date the moment you’re complete with them. So the notion of continuous is pervasive throughout everything that we do, so continuous configuration automation, continuous deployment. The original Puppet engine is all focused on basically having a model and enforcing it continuously. Continuous integration, continuous delivery. All of those things are very resonant for us.
Where does that move you into, in terms of the competitive landscape? There’s a traditional notion of where Puppet sits in this world, and it sounds like you want to change that a little bit and go after some of these other markets.
Omri Gazitt: So I would say traditionally, we live in what Gartner calls the continuous configuration automation segment, which is a pretty narrow definition of what people use even our traditional bits for. One of the things that we saw quite a bit was that we were being used quite heavily to automate application release, or ARA, application release automation. We showed up actually pretty strongly with Forrester Wave last year, and that was even before we got a first-class product in that space, so to speak, Puppet Pipelines.
I understand Puppet in the traditional IT world, and the role that it played, but as that shifts — and you’ve got multicloud, you’ve got hybrid cloud, you’ve got all these different things — cloud vendors offer a lot of these services. How do you think about that part of it, that you’re walking into territory that is sought after very heavily by some of the biggest companies on the planet?
Omri Gazitt: So one thing that our customers have always turned to us for is a platform-neutral abstraction. But having one framework that essentially abtracts the underlying operating system was incredibly valuable to a certain class of customer. And, fortunately for us, that segment is large. I call that the open hybrid segment, if you will, right?
So if you think about open versus proprietary and you think about public cloud only versus the world is going to be a hybrid place, that quadrant — open hybrid — is a quadrant that we play and actually win in quite a bit of the time. So if you think about, we know certainly other organizations that have made just a 100 percent bet on (Amazon Web Services), and that’s fine. They’re incented to take advantage maximally of that platform. So they’ll use Cloud Formation, for example, for automating the creation of (virtual machines), storage, networks, security groups and so on, all of the different resources that AWS has.
But the kind of customers that we spend a lot of time with — enterprise customers — their environment is just really messy. They have some stuff on premises, some stuff in the cloud, they have traditional applications, containerized applications. They need something to actually help them deal with all that complexity, so they can’t afford to just (tie) themselves to one particular solution. So those customers find our framework very attractive because we have a single platform and a single language that allows them to model all of those things, regardless of whether they’re running on AWS or on Azure or on premises. That’s a very powerful thing.
I’ll add to that, back when we set out to do systems management if you will, we didn’t have to be an OS vendor to do that, right? So we could ride on top of that layer.
In the last 15 months since you’ve come on board, it seems to me that the layer has shifted up to the container level, the orchestration level, which seems to be the more important foundation at the moment. Where does Puppet fit in that world?
Omri Gazitt: Yeah, I have a lot to say about that. I wrote a guest column for InfoWorld a couple of weeks ago called containers are eating the world, an homage to Marc Andreessen’s old article, and we believe it. We see that every day. We see that application development is massively shifting, there’s a tidal wave of shift towards packaging applications as container images. We think of Kubernetes as the new operating system. That analogy’s actually one that I use with the team quite often. Same way that we don’t have to own an operating system in order to effectively be essential towards managing it, we don’t need to own a Kubernetes distro in order to actually be essential to the process of building, deploying applications that are containerized into Kubernetes.
We made a bet on Kubernetes probably, I would say, a year ago. It was a safe bet at that time. Three years ago, maybe a less safe bet. But it’s indisputable that it is the clear multivendor opensource ecosystem that’s going to win, and it’s just as clear that everything at that layer is commodity. Red Hat is a distro company and they’re rounding up all the distros. They want to corner that distro market. It’s not an interesting place for us to play. What we do wanna build is essentially a container and delivery platform on top of that.
So the key recognition is that automation has moved from essentially managing change on running systems to managing change through the build and deploy pipeline, because what comes out of the build and deploy pipeline is an immutable image. So you don’t install agents on that immutable image and then drive change through those agents making change to that running system. That’s no longer the model. Instead, the model is if you want to actually help the automation process, the control point is moved to the CI/CD engine. That’s why we invested in (Seattle startup) Distelli late last year, that’s why we picked them up. We were already strong in ARA, in application release automation, but we wanted to skate to where the puck was going, which is containerized application delivery.
So now we have a new product line for Puppet, called Puppet Pipelines, and we have Puppet Pipelines for Applications, which is the competitive offering that we have for traditional applications, and Puppet Pipelines for Containers. Having both of those is actually very attractive because all of our customers, 100 percent of our customers have traditional applications. A few of them have production containerized applications, and pretty much all of them wanna move in that direction. So having a foot in both of these camps makes you a credible vendor. A lot of times when we win in those scenarios, these customers credit us with being able to meet them where they are, as opposed to require them to go move to a completely different world before we give them any value.
One thing I’m asking everybody that I talk to this year is about multicloud. I’m hearing a little bit more and more about are people who are truly balancing between the Big Three, like production workloads in multiple public clouds. From your visibility into your customers, how many people are really actually doing that in 2018?
Omri Gazitt: We’re seeing an increase in the number of customers do that. So there are three, I would say, generations here. The first generation was AWS. For a while that was a very safe bet, that was the only gig in town. Then generation two was customers that saw Amazon as a competitive threat and were looking for anything other than AWS. It could be on-premises with OpenStack, and I would say Azure has been the primary beneficiary of that. We see a lot of customers that specifically have made strategic investments in Azure because they have a contentious relationship with Amazon the retailer, if you will.
Then, I would say this phase that we’re seeing is around companies that truly want to have a balanced investment across different clouds. We see a lot more interest in Google Cloud Platform than we have in probably the last year or the year before. We see a lot more customers, especially in (Asia), that are deep into the IBM Cloud. You don’t see that almost at all in North America honestly, but IBM has a vast empire and, honestly, the further you get out of Silicon Valley, the more you see vendors like IBM, like Oracle and others who aspire to be in there.
So I’d say regionally, you see a lot more variety than you see here in this country, but even in this country we’re seeing an uptick in the number of customers that actually do have hybrid deployments. They’ll have … often times you’ll see it’s AWS and GCP, or maybe GCP and Azure. I talked about that scenario around the Puppet Container Registry, a lot of the killer app for that is the ability to go take a push, a Docker push, into one of those registries and be able to automate the process of distributing those images close to where they need to run, right? So if I think about Pipelines, it was built to be a multicloud product.
Do you always see a role for on-premises workloads? Do you, at some point, think that maybe it doesn’t make sense to even do that?
Omri Gazitt: I think Steve Jobs put it, when he talked about computers and iPads, he talked about cars and trucks, right? That may not have played out exactly, what he predicted, but the analogy is not terrible either. If you’re a company that gets started today, you’re born in the cloud. You’re not going to ever buy a server, you just won’t.
Well, unless you Dropbox your strategy and you go back at some point, which I think is interesting. That’s maybe a separate conversation.
Omri Gazitt: Yes, and one of the things that drives that transition is just competing with the same company who is your supplier for storage, which is an intolerable situation to be in. So Dropbox is an infrastructure company, the vast majority of companies are not really infrastructure companies. So I think the vast majority of new companies assume that they want that on-demand approach to infrastructure.
I think that there’s a ton of friction still for companies that are established, and that represents pretty much all of our Enterprise customers, to be able to take all of their IT so to speak, and run it in the cloud. When I was at HP, we routinely used to go in, the advisory services team would come in to a company like British Petroleum that was running literally 10,000 applications. Their first few months, they would just basically go through an assessment.
But it turned out that probably 20 percent of the applications weren’t worth touching. There was no ROI in moving them anywhere. It was a much better idea to just leave them where they are. Then there’s probably another 20 percent of applications where it didn’t make sense to have them anymore. There were SaaS offerings that were just available, continuing to run (those) applications made no sense at all, you just switched to something “off-the-shelf” but available as a service. Then there was applications that you could lift and put in cloud (virtual machines), lift them and shift them, and that made sense. Then there were applications that had ROI to actually modernize. That was a pretty small fraction of them, right?
So the notion that everything, every single thing is moving to the cloud and every single thing is moving to containers, I think that’s way overblown. At the same time, if people don’t believe 100 percent that brand-new greenfield applications will be cloud-native and will be packaged as containers, they’re confused.
So we see even a lot of the vendors that initially led with a very strong, “it’s only containers, it’s only cloud, containers or cloud all the way,” certainly in the last two or three years. In order to actually connect with value, with customers and with revenue, they’ve had to walk back that story quite a bit. I recall, I think it was a couple of years ago when Ben (Golub) was still running Docker, he got up in Seattle on stage and said, “Hey, just so you understand, it’s about containerizing your existing monoliths.” People were like, “What?!” But that was a reflection of the fact that Docker just really needed to connect with an audience that had a whole bunch of stuff already, and was struggling to figure out how to embrace the container wave without having to throw out everything they have.
That’s again, why I feel so good about our position with Puppet, is because we have an extremely well-respected brand with our customers, and obviously our biggest challenge is making sure that they understand that we are the vendor to take them into the future. But having that trust that we’ve built with so many of these customers is a huge asset.
It’s interesting from a competitive standpoint, I guess, because you’re starting from this one place and you’re trying to bring your customers into the container world a little bit more. There’s people who started in the container world that are trying to reach out back to (the on-premises) world. You’re going out against companies that are saying, “We get containers, then we’ll figure out the rest of your stuff.” You’re like, “We get the rest of your stuff and we’ll help you figure out containers.” So how is that received and how do you think about it?
Omri Gazitt: Obviously there may be a little bit of a selection bias because our customers love us, but every one of the last 6 to 12 months-worth of customer conversations have gone a little like this.
“Hey, what are you guys doing that’s new?”
“Well, we’re focused on CI/CD (continuous integration/continuous delivery), we’re focused on cloud, we’re focused on containers.”
“Huh, that’s exactly what I need to do, can you please help me? What do you have in terms of stuff that can help me solve problems?”
“Well okay, CI/CD, let’s take you through what we’re doing here.”
“Oh, I need that. I want to buy that. In fact, I was gonna go roll out a competitor using Puppet. I was gonna go automate rolling out a bunch of Jenkins servers. I don’t really need to do that right now because you’re now solving that problem for me as well.”
So we see a lot of ability to go cross over into the adjacencies that we planned on building out, so that’s been working for us very well. With that said, obviously that rising sea floats all boats, and we certainly see the companies that were born with a container-native or container-only capability are obviously growing and thriving as well. So we like our point of view because it’s always good to be able to have a big stable of well over 1,000 customers to start with, as opposed to have to knock on doors and struggle to find your first few.
Puppet I feel like is the perfect sized company to be able to go do that. It’s not such a huge behemoth where the efforts that you’re making with new technology are lost or drowned out in just the reputation of a company as an old guard company. At the same time, it’s got a real business. We have a lot of power that we’re building from, that we can use towards getting ourselves establish in new markets or adjacencies.
I went to this presentation at KubeCon in Texas last December that Pinterest was giving about how they made a big Kubernetes and container bet, basically. It was kind of funny to me that this is Pinterest, this company built entirely on the cloud, and it’s like, “Oh wow, like, we’re out of date already. The industry has shifted so fast that we (already) have to rethink the way we do things.”
One thing that called out very specifically that I was surprised (by) because you never see stuff like that in these presentations, is that they were looking to get rid of a lot of the Puppet stuff that was automating their delivery process. I mean, it only stuck out at me because these things are generally … You’ve both been to a million of these things. Rarely do you see vendors called out in these presentations.
Omri Gazitt: What’s available as a commodity keeps on rising, you have to go adapt to that. A lot of folks that used to automate the deployment of applications using technologies like Puppet, or any of the rest of the technologies in our space, are moving towards basically Kubernetes doing a lot of that stuff for them. So the notion of, for example, lifecycle management of the actual artifact, Kubernetes does rolling updates. The notion of monitoring the health of containerized applications, that’s disruptive to all the monitoring companies that have agents on the thing. Logging, the notion of being able to aggregate logs that come off of all the different container images. That’s disruptive to all the companies that do logging and log management, so Splunk and ElasticSearch so on and so forth, right?
So what do all these companies do? They recognize the new abstraction and they build around it. Puppet’s no different. So for us, the focus on automation has moved from insisting that change has to be managed through an agent that sits on a container image, to wanting to add value in the automation process by helping our customers automate the build-test-deploy cycle. It’s trivial to go create a simple deployment pipeline that says, “I’m gonna register a WebHook on a GitHub repo, on a branch, and automate and essentially build a container image, just do a Docker build on it. Then go deploy it into Kubernetes.” That’s easy. Being able to create a sophisticated pipeline that is able to go manage promotion from dev to staging to production with a manual overrides, with a whole process involved, that’s a whole ‘nother ball of wax.
Where does serverless fit in for Puppet, and does it even? I mean, is it something that you’re thinking about?
Omri Gazitt: We’re seeing customers ask for that. We’re doing experiments with serverless as well. Automating the build-test-deploy cycle for anything, whether it be traditional applications or containerized applications or serverless applications is completely within the purview of what we think of as problems that we wanna help our customers solve. It is still early with serverless. It’s always back to the future, right? I mean, containers have been with us, container technology, for almost 20 years now. It wasn’t until Docker came along that we finally had a couple of essentially usability pieces that enabled the standardization and then the mass adoption of it.
I’ve been involved with Cloud Foundry since pretty early on. I was on the board of Cloud Foundry, and it’s back to the future. The whole vision of being able to CF push an application, essentially just providing the code and allowing the system to basically take the code, reason about its dependencies, composite it with a base image and then do the binding to all of the different services that it wanted to talk to. Whether it was databases or message queues and so on, that’s an old idea. I think the user experience now, especially with the large cloud providers, has gotten good enough that that’s become a very interesting way of building an application.
It seems like with the way the cloud providers are offering serverless services, that they do a lot of that. A lot of that is baked into the service, right, this automation of the testing and deploying.
Omri Gazitt: Well, and again, it’s the same conversation that we had before about whether you want to be multicloud, hybrid or not, right? If you’re going to be 100 percent in a particular public cloud, then there’s a certain amount of incentive to go adopt that entire platform.
I joined Microsoft in 1998, it was just right place, right time, I was one of the first 10 people. Me and Scott Guthrie and Jason Zander and a couple of other people all started on that. It was, for a long time, we were struggling to find the value proposition against J2EE, and that value proposition was, “Look, you run it once, run anywhere, it’s all standard,” but they were all different, right? Websphere was different from WebLogic.
In the end, we built a platform and we basically expected people to just write to .NET. If you were taking the .NET bet, you were going all the way, right? So I think that’s the way that large platform companies think, that’s how Microsoft thinks with Azure. They’re basically gonna try to affinitize you to as many of the services as they can. That’s how Amazon thinks, that’s how Google thinks. (But) there are a certain class of customers that really don’t want to go be affinitized to one platform.
So that notion of the move to multi-cloud, that’s really helpful for us because as that market matures, and as people are trying to go build things that are truly portable and cross-platform if you will, that helps an independent like Puppet have a value proposition that works against everything.
I get that (when it comes to) containers and Kubernetes, in serverless it seems very different because in serverless, at least the way it’s rolling out right now, you are super locked in to whoever you choose. If you’re gonna build stuff on Lambda, you’re all in on Lambda, and you can’t move that to Azure Functions or Google Functions or whatever. You bet on that. So I guess I don’t quite understand the portability question with serverless yet.
Omri Gazitt: It’s more being able to use a single consistent tool no matter what serverless platform you’ve picked. It’s not so much that you can take a particular application that was built for the Google Vision APIs and magically have it work against another platform, it’s about the consistency of the tools and the offering language for automating in the pipelines and things like that.
So it’s not that big of a departure for us to be able to automate the build-test-deploy cycle for a serverless function. In fact, we have 85 percent of the tooling already done, it’s just really the last mile that we really need to change.
(Editor’s note: This post was updated to correct a few transcription errors.)