How KNIME used Replicated to Modernize their On-Prem Products

In this video and transcript, Replicated's CEO Grant Miller interviews Jim Falgout, VP of US Operations at KNIME. Read or watch below to learn firsthand how KNIME uses Replicated in their distribution workflow and the success they've found selling into customer-managed environments.

Grant: I'm really excited for this. Jim Falgout is the VP of U. S. Operations at KNIME. Jim has been instrumental in the evaluation and the implementation and overall strategy behind using Replicated. I think his perspective here is going to be really interesting. So I'm very, very excited to have a chance to chat with him. The 1st question is just to help the audience know a little bit more about KNIME. So can you tell us about your business?

Jim: We're really a complete platform for data science. All the way from data discovery, transforming data, getting it ready to build models, doing visual visualizations, all that's kind of on the create side. And then we also work on the production side. So, how do I deploy models or deploy data engineering workflows? How do I monitor them and then also consume and interact? If you work with KNIME you can build these data applications, which are kind of web applications that you can access, and that allows you to push out data science all the way to consumers within your company who are not data scientists, but are specialists in their area and can actually use this. 

So, three big pieces of our overall software. 

We have the KNIME Analytics Platform. This is open source. You can download it. It's free to download, free to use and this is where you build what we call workflows. So, it's kind of a designer. You can build out workflows. You can run them. You can do whatever you want to do. The platform is very extensive. It's built of over 4,000 nodes, lots of connectivity. You can do a lot with just the Analytics Platform.  But when you're ready to start interacting with teams, you want to do some collaboration, you want to do some automation, then that's where our KNIME Hub comes into play.

Right now with KNIME Hub we have Community Hub, which is at hub.knime.com. It's software as a service, so you can go there, you can share with friends. We're adding automation here by the end of the year. So you'll be able to run workflows through schedules, things like that.

And then Business Hub is what we're going to talk more about today, and here's where Replicated comes into play. This is where we do on-premise installs and cloud installs. This is for multiple teams within a larger business. We'll cover that a bit today.

Just a quick view of what the analytics platform looks like. It's very visual. It's a drag and drop kind of interface. It's very low code. You can still do code. We have integrations with Python and R, which are very common in the data science arena. Not just building data science, but also putting into production and using one single environment so you don't have to use a bunch of different products to accomplish. This is showing how we can impact data, and scale the impact of data across your company. It starts with that analytics platform. People build really cool stuff with it, other people see it, they kind of want to jump in with it. Also, you can move into Community Hub at that point with teams or into Business Hub. And then, with Business Hub, we have full deployment and monitoring available - ways of building data applications that hundreds of people can use. You can deploy [workflows] as REST API's so you can integrate them with just about anything.

Grant: Thanks Jim. I know that KNIME Hub now runs on Kubernetes, but you maybe had a KNIME server before that, and so you were modernizing your stack.  Can you talk about the sort of evolution of those changes, perceived risk benefits, and how you mitigated all that and what the journey was like?

Jim: Oh, sure. Yeah. And it's been quite a long journey, right? So it's probably been over two years since we started moving towards Kubernetes. The previous product was a single node application and it was great. It helped our customers a lot. 

It was more of a department level kind of server. It could only handle so many users at a time, and so we were getting a lot of push from our customers to move towards running multiple servers. And they were like, “Hey, why can't we just have one thing that all of our thousands of users go to” and so that's when we decided, “okay, we need to move our platform” and we decided on Kubernetes because we didn't want to pick one cloud platform and just go all in on that platform and Kubernetes seemed like, you know… it's cloud native, right? So it seemed like that was what pushed us towards Kubernetes. It was risky, right? We had to move our development team over to microservices. Our operations team now had to learn about Kubernetes. We also use ArgoCD for deploying our Community Hub, so a lot to learn there. And our customer care, our sales team, they had a lot of worry about complexity. They heard Kubernetes and they just, “oh, I've heard horrible things about Kubernetes”. And, so it was a lot of working against that sort of preconceived notion.

Ways that we really worked on alleviating those risks was (and this is ongoing, right? Because we're still kind of in the middle of it, but really focusing on training and communication) finding key people in the sales team. We found a couple of really deep techies that were really interested in diving in and we moved them into more of a Solutions Architect kind of role. They've been really helpful in working with customers and alleviating a lot of fear around this new technology, so to speak. So I think that really helped. That's some of the ways that we try to work around those risks. 

Grant: Yeah, and I guess for Replicated I think you mentioned like it was a monolithic server install. And then did you have a cloud platform at the same time? Or how did it work? 

Jim: No cloud platform at that time. And that's kind of where KNIME Hub came from. We saw this need of our customers to do more collaboration and collaboration with people who didn't even use servers, more of that automation kind of platform, they just had a lot of data scientists or people who were interested in data science and they used KNIME Analytics Platform on their desktop. And they just want to take it to the next level that they want to start building communities and things like that. And that's where we built the Community Hub.

And from Community Hub we were like, “okay, we need to upgrade our overall architecture”. So that's where we came up with taking what we'd done with Community Hub, which was based on Kubernetes, in the cloud. But also making that available to our customers. The first time we showed off Community Hub all of our more enterprise level customers said we love it, but we want it in our network behind our firewall just for our people. So, that's kind of where that started. 

Grant: Great. And so it sounds like you got some experience with Kubernetes before you started distributing to customers by introducing it into your hosted Community Hub, probably helped your team feel a bit more comfortable with it, and then you were looking for a way to distribute that to your customers who wanted a self hosted version, one that they can control.

Jim: Exactly.

Grant: And that's sort of when you found Replicated. Is that right? 

Jim: Right, as we were making that journey from Community Hub to Business Hub, which, like you said, we'd have to install in customers’ data centers. That's where we started looking for solutions for that. And actually, we found Replicated before we even released Business Hub. And so, once we started working with Replicated, we kind of built it in from the beginning. So, our first customers that had [Business] Hub were Replicated installs. 

Grant: Great. What were your expectations of these different customers who maybe had a variety of Kubernetes experience expectations? How did you prepare for that? 

Jim: First off, we have a big disparity in customer deployments. We have really small customers who are maybe a small five person team all the way up to enterprise customers who have hundreds, if not thousands of people, using our software.

Replicated install has been great. It's really helped get that going very quickly. We can do proof of concepts and take an hour to do an install and have it all up and running. There's a little bit of expectation with Kubernetes that this is going to take days to install. It's going to be super complex. So taking all that complexity away was huge.

We had to try to juggle between those two, which is quite a demand. For the smaller customers embedded kURL, single node, Replicated install has been great. It's really helped get that going very quickly. We can do proof of concepts and take an hour to do an install and have it all up and running.

There's a little bit of expectation with Kubernetes that this is going to take days to install. It's going to be super complex. So taking all that complexity away was huge and great seeing some of the things you guys were talking about today. That helps us even more.

For larger customers I think we were a little naive when we first started building our deployments and thinking, “Oh, we have to provide ingress. We have to provide OIDC. We use keycloak - we have that embedded within our stack. But when you walk into some of these enterprises, as someone earlier was alluding to, it's all over the place, right?

Some are pretty open to anything. Some of them are really locked down. And so we had to find our way into doing that. And ArgoCD came up a few times with, “why don't you just hand us your helm charts and let us do the install. Let us manage that for you”. So seeing that again today with the demo with GitOps was really interesting.

So it's kind of been all over the place and we're still working through some of that complexity.  One thing that we found though… ingress was probably the first one we ran into because we deploy our own ingress. And a customer was like, “Hey, we have that already. Don't deploy your own”. So, we had to go back and we found that in several different places. Having the flexibility of the config within Replicated was super valuable. We could do an install and they could just click “don't deploy ingress”, and then we would give them a bunch of options that helped us integrate with their ingress.

Replicated really helped with that. We really appreciate that flexibility. 

Grant: Yeah - ingress, storage. Some of these add-ons can add up some complexity to these deployments in terms of different clusters and different requirements. Just be honest - when you're looking at the vendor, particularly looking at a vendor that's going to be helping you with something so instrumental in a lot of your largest customer environments,  there must have been some hesitations, right? There must have been some objections. What objections did you hear from your team at large when you were thinking about kind of adopting Replicated and then how did you overcome those?

Jim: Yeah, good question. From the leadership team, there was really a lot of, “it's an installer. Why do we have to pay for an installer?” And so it was just really, once we went through the proof of concept with the sales team at Replicated, we learned so much more about it and also helped us form a story of why this is important and where the value comes from within Replicated beyond installs.

We started doing installs and we got good feedback from the customer care team and things like support bundles and one click updates, all of those features, really helped and I think the vendor portal was important in that also, and that we could see a little bit of what's going on with customers.

That really helped them. And I think over time there was still some skepticism, but as we started doing installs and we got good feedback from the customer care team and things like support bundles and one click updates, all of those features, really helped and I think the vendor portal was important in that also, and that we could see a little bit of what's going on with customers. I think someone was demoing some of that today. That data is really powerful, right? Because we can see who's using it, Who's not. We had a few times where we were surprised with customers churning with KNIME server, right?

And we were like, “wow, we thought you were using it. We thought you were happy with it.” So for us, this is really valuable. They'll see that and be proactive with our customers. It brings value to them also. 

Once we got into those first few installs and I got feedback that really, I think, set aside a lot of the skepticism that the leadership team had.

Grant: Okay. So overcoming skepticism with the leadership team. Really about showing value, particularly beyond the installer, I think is a common theme, right? Because installation can be fairly commoditized. Are there other ways that you think about the value that you get from Replicated? 

Jim: Yeah, I think we talked a lot about the Replicated software. You've seen a lot of demos today on the kind of things that you guys are doing, which is really great. 

So I think a lot of the value, at least from my perspective, that I see is not just Replicated software, but the people that you have, how their breadth of knowledge that they have in the Kubernetes world, is really amazing.

So I think a lot of the value, at least from my perspective, that I see is not just Replicated software, but the people that you have, how their breadth of knowledge that they have in the Kubernetes world, is really amazing.

That initial support we got through doing a proof of concept with the sales team was really valuable. We didn't want to do it -  we kind of were pushing back a little bit, but we were really happy we went through that process because actually we learned a lot about the software. And also, someone who's working with [Kubernetes] day after day, they just have these little hints and things that you can do. We just learned a lot going through that process beyond what the software does, but also I think that people really made a difference. 

And the support team, we've been really happy with the support team. You know, all software has bugs. We understand that - we make stuff for ourselves. But it's been really great working with the team. And it's really nice having Aaron as our sales rep. He's here in Austin. We've met and we've met with you [Grant] also. And that's really helped getting that kind of personal touch. 

Grant: Yeah. It has been great to visit your offices, which are not too far from my home. So, pretty cool particularly in the post COVID world.

Jim: Yeah, that's unusual. 

Grant: Yeah, exactly. Yeah, I think it's interesting to think about that value you get and talk about the vendor portal and being able to kind of get some telemetry and understanding.  Do you think there's also value that your end customers, sort of folks that are running Business Hub, get because you're using Replicated? 

Jim: Yeah, definitely.  I mean, everyone thinks an install should go smoothly. But when it doesn't  it's a problem, right? So, a smooth install is huge. That's helped us. And smooth upgrades.

Being able to have a customer have insight into what's in the pipeline as far as upgrades, and being able to see those and manage those themselves has been really valuable. It was a pretty rocky road before, with our previous product, and doing installs. So having that automated and being able to have customers manage that process themselves saves us a lot of time, but also provides them a lot of value. 

They'll create a support bundle and upload it before they even kind of get in touch with us. So they see the value in that.  Like we talked about before, customers who are maybe not that familiar with Kubernetes, having the kURL install embedded, those small to medium customers.

For them, that took away a lot of the scariness of dealing with Kubernetes.

We talked about support bundles, being able to support this, without having to jump on a call every time. Customers kind of learned that, right? They'll create a support bundle and upload it before they even kind of get in touch with us. So they see the value in that.  Like we talked about before, customers who are maybe not that familiar with Kubernetes, having the kURL install embedded, those small to medium customers.

For them, that took away a lot of the scariness of dealing with Kubernetes.  Seeing you guys even move closer to not even knowing that Kubernetes is there is really positive. I think that helped alleviate a lot of their anxiety over this super complex thing that maybe is not as complex as they thought it was going to be.

Grant: Yeah, that makes a lot of sense. In terms of the go to market side of the business and your field teams - what have you been able to do to enable those field teams to really deliver a Kubernetes based application and support that and get customers comfortable with it versus this more traditional monolithic application that's based in Java that people probably felt a little more comfortable with? How did you bring the whole team along? 

Jim: We kind of mentioned that a little bit earlier. A lot of that was communication and training. The cloud ops team that kind of came up to speed with Replicated and learned that. And then we started doing training classes, or maybe workshops with them and have them walk them through the install. And then once they walk through that first install and could see that, “okay, Kubernetes got installed, our application got installed. This took less than an hour. Everything's up and running.” That's kind of cool. Right? 

That helped them see that this was doable, and it wasn't this great mountain that they needed to climb. But yeah, I think that’s just communication and training getting them over that initial fear. 

And, like I said before, we built up this solution architect group and continue to build that up. And they're taking on that role now and able to do that with other people within the sales team. 

Grant: That's great. I know one thing that I remember us talking about in a meeting a while back was that a lot of software vendors list their applications in the cloud marketplaces, right? So maybe the AWS marketplaces, AMI have different ways to list. I know we were talking through some of the licensing challenges of using Replicated in those environments. And this, I don't actually know the answer. Did we ever figure out how to actually enable you to use the sort of Replicated delivered version of the application in those cloud environments with our licensing?

Jim: Yeah, we had a lot of discussions around it. There's nothing built into Replicated today  that we're doing. So, what we're doing now is we are going to use Replicated for those AMI builds.  Especially the KOTS pieces is really valuable. And before we had our server product on AWS and Azure marketplaces for years, but doing updates was really hard. So they basically had to go to our website and figure out how to do an install and a lot of them really like the “I just click a button and marketplace and I have an instance. I don't have to install it. I know you guys follow best practices”, but then we threw upgrades in their lap and they had to log into the machine and do all this complicated stuff. So, yeah. KOTS helps get away from that. And, so we definitely wanted to use Replicates for that part. As far as licensing, you start up an AMI in marketplace, you're running in the customer's account. And so we don't, we don't have a lot of control over what that looks like.

So we'll try to reach out to Replicated. We'll create a customer for them if we don't think one exists already, we'll be able to download the license. If we can't do all that, then we're just going to use an embedded license. And then working with their sales team, we just kind of came up with a way of reconciling that every month.

It's going to be a little bit manual to begin with, but the information that we get from the vendor portal will help with that. And of course, we can use APIs for that and try to automate all that as much as possible. 

Grant: Yeah, sounds like a good area for us to maybe look at investing in a little bit more.

Jim: Yeah, definitely. 

Grant: Yeah. What else would you like to see? Other areas that you're that you think that we can do, we can improve. 

There's so many different distributions of Kubernetes, all these different versions, and so being able to just spin up… we have customers using AKS, EKS, some starting to look at GKE. So, being able to spin those things up and do some testing with our application just gives us a lot more confidence that it's going to work in those environments.

Jim: Well, I would, I would have said something like Compatibility Matrix, but you guys are working on it already. So we're super excited about that. And thanks for the demo today. That was really fun to see. There's so many different distributions of Kubernetes, all these different versions, and so being able to just spin up… we have customers using AKS, EKS, some starting to look at GKE. So, being able to spin those things up and do some testing with our application just gives us a lot more confidence that it's going to work in those environments. So that's huge. I appreciate that. 

When we were talking to customers about it [The Compatibility Matrix], I think it can actually introduce a step function change in the reliability and stability of their applications.

Grant: Thank you. Yeah, we're really excited about that. I'm equally excited that it's been great to see how quickly your team has taken that up. And we think there's a lot of value it can bring. You know, when we were talking to customers about it, I think it can actually introduce a step function change in the reliability and stability of their applications.

Because, you know, what we've noticed is a lot of the best vendors are doing some testing, but no one's really kind of digging that deep into the full matrix of different combinations. And you talk about versions and distros, but also there's all the different ingress combinations. So we want to get into those pieces as well to really give you confidence that your end customers are going to be successful in running and upgrading your application. So, I'm glad you pointed that out. 

We have had some of our larger customers say, “Hey, just give us your helm charts.” But we don't want to lose the goodness of Replicated.

Jim: We're very excited about it… there's no big, huge features that we look at and go, “you know, we see a big hole here”. Some of the things that we talked about with cloud marketplaces would be great. But, I love to see some of the innovation you're doing. I'm glad I joined to hear about the helm install. That definitely kicked off some ideas with us because we have had some of our larger customers say, “Hey, just give us your helm charts.” But we don't want to lose the goodness of Replicated. And so definitely glad to see that happening.

Grant: Yeah, that's exactly it. Today, I'd say 5-10 percent of end customers are saying, “Hey, we'd actually prefer a helm chart”. And so we just want to make that possible and say  “Hey, it's the same release. You can just install with helm. You can use KOTS. You can use this embedded installer. Doesn't really matter. You can use GitOps, whatever else it might be.” So, that's our goal is to be able to release, support and serve that full spectrum of end customer requirements. 

So, Jim, thank you. I really appreciate you being here and sharing your experience with our community and with our team. It's amazing to hear the success that your business is having, and we're really proud to be able to be a partner and to help you help you grow even more.

So,  thank you for the opportunity. 

Jim: Oh, thank you. I appreciate it.