Meeting with Luke Kanies, the CEO of IT automation software maker Puppet Labs, is becoming a regular event for the two of us -- our latest meeting was our third this year.
During this time, the status of DevOps -- which is closely linked to automation -- has continued to evolve. As experienced IT people know, all new technology is hyped. On its launch, the technology will be praised beyond measure by everybody but after the hype comes the reality of making it work.
DevOps is no exception in this and, like other technology developments before it, it is surrounded by so much hype it is somewhere near what analyst Gartner describes as "the peak of inflated expectations". Kanies thinks it is even beyond that.
At our latest meetup, Kanies discussed Puppet Labs' latest product release, the Dell-EMC merger, coming trends and more.
Q: I understand you have a new release of your core product coming up?
Kanies: The release, which is coming out in December, is the first release which has cross-focus orchestration. When I got up at Puppet Con a month ago, I said that I know two things about this product: one is that it is useful for you right now -- everybody can use it to solve real problems right now. The second is that you are going to hate how much more we could have done.
The point is that I believe that the release is useful right now but what it also does is open up a set of opportunities for us. So as this migrates to being a core part of the platform, as it migrates to becoming something that everyone can work with, it becomes the point of action and control across the data centre and all platforms and services -- and so it is becomes so much more useful.
Q: Since we last spoke, a lot has been happening in the market, such as the proposed merger between Dell and EMC. What's your take on it?
Every large, enterprise vendor in the market is undergoing massive change. If you look at the classic disruptive model, Christiansen and stuff, then people will say, 'Oh well, this is being disrupted by that' and so forth. But now, in most cases that you look at, there is no one person or company coming in and taking over.
Prices are going down because buying patterns are changing. This theory we once had of how the enterprise world works just doesn't apply anymore. People cling to the idea that this period of confusion is going to end soon. I don't think that. I think this period of confusion [in the market] is going to last for a really long time -- like decades.
I don't think things are going to settle down for a very long time.
Q: This is bad and good news for companies, isn't it? How do you convince people that in this period of confusion, you are a safe haven to go to?
Well, first you need to have customers who you can use as examples for your products. For that you need to start with people who are comfortable with risk. Then when they are using your stuff successfully, then you can go to others and say, look, not only did they find our stuff interesting but it was useful and reliable as well.
So that is the path that most of us are on. This is why the Geoffrey Moore 'Crossing the Chasm' story works really well. You get the early adopters, they use it successfully and you use those to get credibility to get you into the mainstream.
Q: So what do you see as being the big issues right now?
I think one of the big areas of change at the moment is the wide scale adoption of containers and really of container/schedulers. Just in the last six months the conversation has shifted from the world of containers to the world of schedulers.
Now that is interesting because it is super-obvious that these are important -- it's really obvious that they are going to have some role in the future of IT. If you look at Kubernetes, that is the fourth generation of Google's internal scheduler, so it's been running their internal infrastructure for a long time.
Now what I like to do is to compare the movements of Docker to the movements of OpenStack. Now OpenStack was, in many ways, a disaster. OpenStack was this theory and then a huge number of companies piled on and started funding technology into it -- and while it's not nascent, it did not get as big, as fast as many people thought it would.
Now if you look at Docker, that is the fastest adoption of any open source software, ever. Now the thing is, it doesn't represent a huge change and what people have come to realise is that things that involve a huge change will not get adopted that quickly. In a way, the size of the change implies the cost of adoption.
Kubernetes is a massive change and it is hard to change. People really have to redo all their applications to get them to work in the scheduler.
Now when I talk to customers and ask them, 'How is your structure built?' Those structures, those hierarchies don't work the same in a world that relies on these schedulers.
So if you compare Kubernetes to OpenStack, there are two big differences that I see. The first is that Kubernetes is not based on a theory, it is based on 15 years of production infrastructure. The second, and I think most interesting thing, is that Google is driving this very hard. And Google is not trying to make money out of it either. Google cares about its leadership position, it cares about the Google cloud, it cares about the way people use its infrastructure to build their own infrastructures, but it is not, indirectly, trying to build a business.
That's why I think that Kubernetes will have a lot more impact than things like OpenStack had. Who Google is going to impact and how is going to be very interesting.
Q: Now the early adopters of your technology are pretty well there now?
Yes, well we are pretty well mainstream now and we have been seeing a shift in our user base, for really quite a while now. In fact, over just the last year, we have been seeing quite a dramatic shift.
Recently we got a call from one of our customers complaining about our enterprise product. Basically the customer was saying, 'You don't have an enterprise product'. We thought that was really strange because the customer had been using our enterprise product for some time. When we pushed him to find out what he meant, it seemed that it was because we didn't have a technical account manager, and we didn't have the support hours that he wanted and so on.
So this had nothing to do with what the customer thought about the product, but was about the support systems around the product. It was about customer expectations whereas a year ago, the way we supported the product was perfectly adequate.
Q: How do you ramp up to deal with those people?
For one thing you have to put in the support teams around the world so you can have around-the-clock support without asking people to stay up all night. It's not rocket-science, you just learn from the customers. 'What was it you said that we absolutely had to have? Who does that well?' and so forth.
Q: Where do you think we are on DevOps? It's only been a short time since we last spoke but have we moved on appreciably?
Everybody is definitely excited about doing whatever it is that they are calling DevOps. I think we are definitely about to move into what Gartner calls "the trough of disillusionment".
Not everyone knows what DevOps is but everyone has a vague sense of what they think it is. Everybody is doing DevOps and it is getting near mainstream but the word isn't mainstream.
But to be fair, there isn't a very clear idea of what DevOps is -- there isn't a defined model. There are many kinds of certifications but none of them are realistic. There was an event I attended recently where every attendee got a DevOps certification. People are springing up offering their own DevOps certifications. It's become a joke.
So I think we are at a point where people have a decent sense of what it can bring to the organization -- they are just not clear on how to do it.
We have gone from the stage where it's great because you and your five friends can do DevOps, to 'how do we get 1,500 sysadmins in Germany to do it?'. Now we have teams spread around the world and they don't all speak the same language and they need to implement DevOps and really it has not really evolved enough to handle those things yet.
Read more about Puppet Labs, Docker and DevOps