#122 September 22, 2020

Grafana, with Torkel Ödegaard

Hosts: Craig Box, Adam Glick

Torkel Ödegaard is the creator and project lead of Grafana, and co-founder of Grafana Labs. Learn how Torkel went from modding video games to building a data visualization platform, and co-founding a company that is now offering a complete monitoring service built on Prometheus.

Do you have something cool to share? Some questions? Let us know:

Chatter of the week

News of the week

CRAIG BOX: Hi, and welcome to the Kubernetes Podcast from Google. I'm Craig Box.

ADAM GLICK: And I'm Adam Glick.


CRAIG BOX: I'd like your opinion on something. If you're going to watch a movie about a historic event-- that is a fictional retelling of that event-- and you also have a documentary about that event, which order do you think you should watch those two things in?

ADAM GLICK: I think you should watch the documentary first, so that you understand the situation and then can understand the wonderfulness of the story, but not get confused about what the reality is.

CRAIG BOX: That's interesting, and I say that because I guess it depends on the skill of the filmmaker-- obviously trying to tell a story. And a filmmaker will assume that you don't have any background and tell that story. And as long as they're doing it in a relatively truthful way, then you will end up with a deeper understanding, I find, if you watch the documentary second.

All this is to say, that two years ago in 2018 there were two movies made, which are my recommendations for the week. There was one called "On The Basis Of Sex," which is a telling of the story of the life of Ruth Bader Ginsburg through her Supreme Court cases. And then of course, there was the documentary, which is just titled "RBG"-- the Notorious RBG as she was known-- talking about her life. And they came out in the opposite order. But I am going to recommend that you watch the fictional version first, and then you watch the documentary afterwards.

ADAM GLICK: And I'm like, how do I transition from that?


CRAIG BOX: How's the weather there?

ADAM GLICK: Things here have gotten a lot better. The smoke has cleared up a little bit. So we had about a week of smoke and staying indoors with the air quality. The upside of that being that we got to start the "Picard" series-- realized that I'm a little bit late on this one, as usual as we get around to things. But am really enjoying it as both me and my wife are fans of "Star Trek," and so really interested to see what they do with the Picard character.

CRAIG BOX: I hope they don't give him hair.

ADAM GLICK: They don't pull a Luke Skywalker on this one? Probably not. Should we get to the news?

CRAIG BOX: Let's get to the news.


ADAM GLICK: A kernel vulnerability was disclosed last week, which has the potential to allow a container user to gain root access. If you have the raw capability in your network namespace, you can write small numbers of bytes to places you shouldn't be able to. If you don't need this capability, you can remove it by using Gatekeeper, pod security policy, or manually in your pod spec. Some distributions also disallow it by default. Check with your vendor to see if you need a patch or to mitigate.

CRAIG BOX: One group that didn't have to mitigate was users of gVisor, or the GKE feature built on it, GKE Sandbox. gVisor has long touted its theoretical usefulness in just this scenario. And a write-up by Google engineer, Fabricio Voznika, explains exactly how the defense and depth model works to prevent this attack even if running on an exploitable kernel.

ADAM GLICK: IBM Cloud announced a new serverless product called a Code Engine. Code Engine is a serverless platform for running interactive programs, or batch jobs, which can be provided as a container image or compiled from code using build packs. It's built on top of Kubernetes, Knative, Istio, and Tekton, and is free to use while in beta.

CRAIG BOX: Mirantis has launched Docker Enterprise Container Cloud-- an updated and renamed version of their pre-Docker Enterprise acquisition, Kubernetes as a service product. Container Cloud can run on top of any operating system and any virtualization platform, and integrates with the Docker Universal Control Plane. It's free for three clusters up to 15 nodes, and enterprised priced after that. Mirantis also shared more information on its plans for Docker Swarm, which boiled down to replatforming it on top of Kubernetes. You can learn more about Mirantis and their acquisition of Docker's enterprise business in episode 110 with CEO Adrian Ionel.

ADAM GLICK: The CNCF has welcomed KubeEdge into the incubation phase. KubeEdge is an open source system for extending containerized application orchestration capabilities to hosts at the edge-- or what we used to call computers. It's designed to address three main problems-- network reliability between the cloud and the edge, resource constraints on edge nodes, and scalability challenges of highly distributed edge architectures. This brings the number of projects in incubation at the CNCF to 21.

CRAIG BOX: Data platform company Segment has moved to a git based workflow for delivering apps to their Kubernetes clusters. They've open sourced a tool called kubeapply, which they use to expand, validate, diff, and apply config files. It runs in the command line, or as a GitHub bot, which they use to have humans validate and approve changes using comments. The new process and tooling has improved the developer experience for production changes, and increased Segment's confidence in doing deployments. And they hope it might be similarly useful for you.

ADAM GLICK: Grafana Labs has announced the release of Grafana Metrics Enterprise-- a Prometheus as a service solution promising enterprises an easier, cost effective way to run Prometheus at scale. The product builds on Cortex-- recently promoted to incubation in the CNCF, and of which Grafana Labs is the primary sponsor. Learn more about Grafana in today's interview.

CRAIG BOX: In the biggest news-- from what we are retroactively naming Storage Week-- Kubernetes storage vendor Portworx, with an x, has been acquired by Pure Storage for $370 million. The idea of the acquisition started when Pure realized some of its largest customers were running Portworx software on top of the storage hardware. And indeed, Portworx was ranked at the top of a recent GigaOm Radar report for Kubernetes storage. Congratulations to the Portworx team.

ADAM GLICK: Israeli storage startup, Ionir-- previously known as Reduxio-- has reformed with an $11 million VC investment, and announced a platform for moving data between Kubernetes environments. Using their so-called data teleport technology, Ionir promises to instantly move or copy persistent volumes between Kubernetes clusters in the cloud in under 40 seconds independent of the size of the volume, or the amount of data involved. If you want to find out more, all paths lead to a form where you can request early access.

CRAIG BOX: NetApp launched a new beta cloud volume service for GCP this Storage Week. The existing NetApp cloud service runs on colocated hardware. But this new version is hosted on GKE, and as such has a planned Recovery Time Objective, RTO, of less than 90 seconds, and an uptime target SLA of 3 and 1/2 nines.

ADAM GLICK: Microsoft's entry into Storage Week was a new AKS release, which adds a CSI driver for Azure Files and Azure disks in preview. Red Hat's entry was OpenShift Container Storage 4.5-- with new features around Ceph and OpenShift as well as support for air gapped and proxy installations.

CRAIG BOX: VMware is expanding Tanzu to more environments and more editions. Initially only available on top of the all singing, all dancing, VMware Cloud Foundation hyper-converged stack thing, Tanzu is now available on top of vSphere in an edition which will be known as Tanzu Basic. Advanced and enterprise additions are also coming in the future. Not wanting to miss out on Storage Week, they also announced a vSAN Data Persistence Platform for storage vendor's integration. Control plane Tanzu Mission Control also added security policies and policy insights powered by OPA and Gatekeeper.

ADAM GLICK: The security space continues to see a lot of activity. This week, SentinelOne announced an automated application control engine, a runtime container protection product that utilizes AI-built allow lists to protect running containers in Kubernetes. The product runs both on-prem and in the cloud, and uses a single agent to detect attacks and respond. The product is designed to help DevOps teams protect containers without having to build custom allow lists. SentinelOne hopes to expand this into additional environments in the future.

CRAIG BOX: As summer ends in the northern hemisphere, so too does the Google Summer of Code. Over 1,100 students from 65 countries worked on the program this year, including 16 interns mentored by the CNCF. Their projects are written up in a blog post, and the deep dive has been published into one project. Somtochi Onyekwere, from Nigeria's Federal University of Technology, worked on building operators for Kubernetes cluster add-ons. Her work, mentored by Justin Santa Barbara, is written up on the Kubernetes blog.

ADAM GLICK: Two co-located events will be run before KubeCon-- not in Boston-- this November. Cloud Native Security Day and ServiceMeshCon are both back by popular demand-- both on November 17th, and both at a $20 uplift from your KubeCon ticket. Calls for proposals have opened for both, and they run until October 4th.

CRAIG BOX: Finally, in episode 89, we talked with Marin Jankovski from GitLab about their journey to Kubernetes. Their hosted version, gitlab.com, started migration one year ago. And to celebrate the anniversary, GitLab's John Jarvis has written a year-end review retrospective. He covers lessons learnt along the way, and how they will influence future migrations.

ADAM GLICK: And that's the news.


ADAM GLICK: Torkel Odegaard is the creator of Grafana, as well as a maintainer of the project. He is also the co-founder of Grafana Labs. Welcome to the show, Torkel.

TORKEL ODEGAARD: Thanks, true pleasure to be here.

CRAIG BOX: How did you journey to software start?

TORKEL ODEGAARD: A long time ago, I guess the first experiments was in BASIC with Commodore in the late '80s. But I was a bit too young then. By I think 8th grade or 7th grade, I had bought a book on C++ and that got me really into programming. And throughout high school I'd worked on games with friends, so there was never any question on what I should do at university. I was just-- yeah, I'm going to work on software.

CRAIG BOX: No backup plan?

TORKEL ODEGAARD: No, no real backup plan. Luckily the 2001 crash happened at the same time while I was in school. So got a good education in c omputer science-- much longer than I expected because of the IT crash. But I really fell in love with writing computer games. Quake 3 mods was a big thing as well that I worked on at high school.

CRAIG BOX: Do any of those games still exist? Can we share them with our audience?

TORKEL ODEGAARD: No, I don't think my hobby games exist, but I worked on the Quake 3 mod called "Rocket Arena," which was quite popular.

CRAIG BOX: Oh, it was great fun.

TORKEL ODEGAARD: We took over that from the original "Rocket Arena" author. And then just by accident when I finished university, my first working gig was a Microsoft related stack, so I got in to. I haven't worked on C++ professionally since then. That was the majority of my teenage and high school years.

CRAIG BOX: You didn't want to try and follow gaming into a professional gig?

TORKEL ODEGAARD: I did. I had friends who took that path, but I just kind of got into writing web apps, and B2C, and C2C web shops for Panasonic. And it was just another world of very different types of problems. I mean, when I wrote computer games I thought about design and architecture, but not about maintainability, code readability, testability, and long lived code bases. That became real, a passion as I took on big Microsoft, and big .NET projects that I worked on, and websites.

ADAM GLICK: What was the tech scene like that you were dealing with? You had the gaming scene. Microsoft was very popular in terms of the technologies you're talking about there. What was the tech scene broadly in Sweden?

TORKEL ODEGAARD: I got really into TDD, Test-Driven Development, and extreme programming, which was kind of a niche community at the time.

CRAIG BOX: Not quite as niche as extreme ironing.

TORKEL ODEGAARD: No, but within the .NET, the Microsoft scene, using open source, and using the best tools for the job, that was really kind of a fringe movement that I was a part of in the Stockholm scene, which was really exciting to sort of push extreme programming, open source, and test-driven development.

CRAIG BOX: You worked a lot of gigs as a contractor, and I know that's something that may not be as familiar to the American listeners. But especially in Europe, there are a lot of people who do contracting gigs versus working full time. How would you describe the difference between these two styles?

TORKEL ODEGAARD: It really depends, but the main difference isn't that big. In Sweden, many contractors work many years for the same employer. So it's basically a way in Sweden where employers can work around employment protection laws and have more flexibility around temporarily boosting projects. But if you're familiar with software development, temporarily boosting up software development doesn't really work, because the software still needs to be maintained and usually new things need to be added.

CRAIG BOX: And that whole mythical man-month thing.

TORKEL ODEGAARD: Yeah, but those contractors usually need to stay to maintain and to continue development. So the contracting situation in Sweden is you're basically an employee.

CRAIG BOX: One of the places that you were contracting to was a company called Tradera back in 2009. Would it be fair to have called them the eBay of Sweden?

TORKEL ODEGAARD: Yes, and they were owned by eBay as well.

CRAIG BOX: By acquisition when eBay found that they couldn't enter the market because there was a local competitor?

TORKEL ODEGAARD: Yeah, they acquired Tradera-- I'm not sure when, 2005, or something. But we still used our own stack. We didn't use eBay's auction platform. And that was a super interesting gig. I was sort of an architect, and we had a microservice architecture using queues and using a very complex production environment. And the thing that I worked on at the time that I was super passionate about was continuous delivery, deploying, getting to a software architecture, and also tools that allow you to deploy multiple times a day.

CRAIG BOX: Very forward-thinking for 10 years ago.

ADAM GLICK: About to say, you were ahead of the curve there.

TORKEL ODEGAARD: Well, yeah, it was a big talking point that the time as well, but that got me into monitoring. Microservice architecture and continuous delivery really requires you to have good insight, good observability of the system.

CRAIG BOX: Careful care and feeding.

TORKEL ODEGAARD: Then you're constantly changing in production. So I had a friend from the Operations Department show me Graphite, which is a time-series database, and one of the first really popular time-series databases, and dashboard and tools, a super easy API-- just send metrics. You didn't have the same kind of worry that you have with logs. Logs you have to be really careful about how frequently and how often you write data.

But with these metrics, they were so ephemeral-- or just UDP packets-- they were super easy to instrument the applications. And all of a sudden, the applications became alive in these dashboards that I was building to see the user behavior, the application behavior, the messages on the queue and where the bottlenecks were, and how that deployment changed that. You could be able to see live in the graph the impact of a change.

It changed the way we built software, and the conversation around changes as well. It became more about what metrics can we add, so we know that the change we're working on right now has the impact that we expect.

CRAIG BOX: Now, you mentioned Graphite there. That's a software package that was created at Orbitz by Chris Davis back in 2006. Graphite is a time-series database. Did it come with a visualization? Did it come with a way to present these metrics?

TORKEL ODEGAARD: Yeah, it did come with-- very similar to RRDtool-- graphing capability as well. And so we had PNG images-- very flexible, powerful graph capability. But the way you built a dashboard in the UI was kind of clunky. And the graphs weren't really intractable. And you had to learn Graphite-- the query language-- as well, and the UI wasn't really easy to use for that. It was kind of hard to understand the queries because it's a small text box.

So I was struggling a bit getting my teams to adopt this and learn it. So that's why I had started looking at alternatives to the Graphite default dashboard.

ADAM GLICK: What else was around at that time? Was Kibana an option that was out there? I mean there's been visualization tools for quite a while.

TORKEL ODEGAARD: Exactly. At the time, we were also starting to use the Elasticsearch for logs, and Kibana. And I fell completely in love with that combination-- the power of centralized logging, searching through the logs, and just the ease of use that you could build dashboards in Kibana, and the graph in Kibana was actually intractable. You could zoom in. So it had a bunch of things that I really liked.

So I took Kibana as a starting point when I started working on Grafana. And I really wanted that easy to use dashboard experience, combined with an easy query builder experience. Because that's where many of my teammates struggling, editing, and understanding the graphic queries. Because it's a small text box. It's a long query nested structure.

So that was where I put a lot of effort into making that easy, and also making it look really vibrant, and good and stripped down. I wanted a really elegant UI, so you could put it on a TV and it wouldn't be full of buttons and icons. It would just be beautiful graphs. So that's the initial design philosophies that I put in.

CRAIG BOX: Kibana was the UI for Elasticsearch as you mentioned. Was it possible to connect it to Graphite at the time?

TORKEL ODEGAARD: No, there was a popular feature request for that. But Kibana was all about very squarely focused on only exposing Elasticsearch capabilities and features. So it wasn't really interested in supporting other data sources.

And at the time, I created Grafana with a very clear focus on time-series and graphing, and not on logs. I removed all the Elasticsearch capabilities and features. It's much later that we added back Elasticsearch as a data source, because Elasticsearch gained a lot of metric features. And since then, Grafana has also gone into logs, and there's been a convergence on many fronts.

CRAIG BOX: So on one hand, we've got Graphite. And on the other hand, we've got Kibana. Is it fair to assume that this might be how the name Grafana came about?

TORKEL ODEGAARD: Yeah, definitely. I mean, Grafana, the name, is definitely a homage to Kibana-- Kibana, but with a focus on graphs. That's sort of where the name come from. And it's spelled with an f because I think it looks better, and it's how it graphs are spelt in Swedish and many European languages.

ADAM GLICK: Your goal was simply on the visualization layer here. That you wanted to attach to the time-series databases, like Graphite, that already had the information. This was really about how people consume and use the information?

TORKEL ODEGAARD: Yeah, in the early days, Grafana was just a front-end web application like Kibana 3 was at the time, so you didn't have a back-end component at all. You host Grafana with Apache or nginx. So it was definitely just about talking to a time-series API of some sort and visualizing that data.

CRAIG BOX: Open source software doesn't have a reputation for being pretty or easy on the eye. And Grafana is a bit of an outlier in that regard. You mentioned that you had that as a goal-- to put something on the screen there-- and you have your gaming background. Did you have a lot of art background there as well?

TORKEL ODEGAARD: I don't have an art background but I do have, I think, an aesthetic eye. And I notice that I engage so much more with software both from a user and from a developer standpoint. If it looks good, and if it looks polished, it looks thought through. That makes it so much more fun to program and work on it and also use it.

So I worked on similar things in the past. And I've always enjoyed the UX puzzle of defining something that looks good, that is easy to use, and is space efficient. It's a really interesting challenge, I think.

ADAM GLICK: Was this something that you did as part of your day job, or was this something that inspired you to work at night in order to fix?

TORKEL ODEGAARD: Definitely, it started as a Christmas hackathon. And I was sick at Christmas, and I didn't spend much time with family and just worked on this, and then worked on it evenings and weekends. But once it was released and the immediate feedback and attention it got, was just such a fuel to working on it every spare moment I had. So I remember waking up at 6:00 AM to have 2 hours so I could work on it before my day job.

And the interest in the project was so big and quick that I quickly realized I need to see where this takes me. So after three months, I think-- after the initial release of the project-- I resigned and thought I need to work on this full time, and see where it takes me.

CRAIG BOX: Tradera were using Grafana?

TORKEL ODEGAARD: Yeah. Yeah, so after the Christmas break, I put it up on our TV on the wall. And we started using it. And it was incredible just having that effect, that we started using it in one team, and then all of sudden all the five teams in the company had TVs on the wall with Grafana on it. Friday demos had Grafana in it to showcase the impact of changes.

I think, definitely one part of the Grafana growth in the last couple of years is that how viral it becomes because it's on the wall. So other teams in a company walk by, what is that.

CRAIG BOX: Did they not want to sponsor you working on this?

TORKEL ODEGAARD: They probably would, but not full time-- and definitely not full time, I think. And I also saw the opportunity in building something around it.

ADAM GLICK: Was that when you decided to leave the company? And was your idea at that point to actually turn Grafana into a separate company?

TORKEL ODEGAARD: Yeah, definitely. At the time I was a contractor and consultant, but I had my own company too as a contractor. So I just figured I can work under that, my own company, still. So I was very fortunate to have that opportunity to have some savings in my own company to be able to do this. So I definitely had an ambition for this to be a company.

But here also things kind of surprised me because I asked for sponsorship from the community. And I got some really good sponsorship deals from Squarespace, and Demonhost, and a few other-- no, I actually can't remember. Mediamath, I think, was our early sponsor, and they really allowed me to postpone some kind of commercialization ideas I had because I got these big companies sponsoring the project.

CRAIG BOX: So what is now Grafana Labs started life as a company called raintank. Is that the name of your consulting company?

TORKEL ODEGAARD: No, raintank is an original name of the US company that I co-founded together with Raj Dutt and Anthony Woods in 2015. So I hooked up with them. I had known them before, and they wanted to do something in the monitoring space using open source-- building a SaaS monitoring company based on open source software. So this was still kind of early. I think this was in late 2014 still, and Grafana was still kind of young.

So we didn't really know that Grafana was going to turn out to be this big brand and this big project. And at the time, there was a kind of popularity as well to have the open source company name something different from the project-- to have separation-- to have this commercial entity and then to have this project. I think later, a lot of companies realized that this is stupid. Because we always had to introduce ourselves like, yeah, we are raintank, the company behind Grafana. So we got tired of that and renamed the company to Grafana Labs.

ADAM GLICK: What were your ideas for commercialization when you started the organization?

TORKEL ODEGAARD: So when we started Grafana Labs, our initial idea was mainly around SaaS monitoring platform based on open source software. And later as the Grafana project and brand that became so popular, there was such a big demand for an enterprise that supported on-premise product as well, so that's where we started Grafana Enterprise. That took a long time.

We worked on Grafana purely from an open source perspective for, I think, three years before we even started doing any kind of enterprise differentiation. And the way we started with enterprise differentiation was through a plug-in model. So we developed enterprise plugins for Splunk, and Oracle, and other commercial data sources.

CRAIG BOX: How did the open source project grow over those three years? Was it largely contributions from people who wanted their own integrations, or was this something that you were sponsoring people to work on it?

TORKEL ODEGAARD: I think mainly from my employees and community contributions. So the growth of the project has been incredible, like doubling in the number of users almost every year. So I mean, going from something like 10,000 installations in 2015 to 40,000 to 100,000 to now like 600,000 Grafana servers that we know of that are running Grafana. So we have an anonymous call home feature that many disable, so there's a lot more Grafana installations running out there.

But yeah, the project has been insane in terms of the interest and growth. And the team has grown with it. Most are employed by Grafana Labs, but we have many external contributors as well.

CRAIG BOX: Grafana Labs wanted to become a company with an open source monitoring stack. You have the visualization piece in Grafana. What did you choose to build out the rest of it?

TORKEL ODEGAARD: We always had database pieces as part of this SaaS platform that were also open source. We had a Graphite compatible back-end for our SaaS platform, but we saw Prometheus took off. So back in 2018, we acquired Kausal.

CRAIG BOX: With a K?

TORKEL ODEGAARD: Yeah, with a K. That ran Prometheus as a sort of hosted service. And in doing that we got Tom Wilkie and Dave Kaltschmidt and got a really good footing in the Prometheus ecosystem, which is something we've been investing heavily in. And that's been our base for now-- our new cloud platform as well. So we really got really deep into Prometheus and the cloud-native use cases after 2018.

CRAIG BOX: But I understand you still sponsor people working on the Graphite project?

TORKEL ODEGAARD: We do, but I think that the Graphite project is seeing less and less growth and interest, so our involvement in it is also declining. And we're making our Graphite support part of our Cortex enterprise offering on our back-ends.

ADAM GLICK: So you've expanded from visualization to a monitoring tool, which has both the data source and the visualization layer on it. How far do you see that expanding?

TORKEL ODEGAARD: I see Grafana really as a platform that we intend to build many solutions on top of. So we have the last two years really invested in making Grafana a development platform for different types of solutions. So Grafana at its core-- and the dashboard inside-- is really a toolbox that people need to build a solution themselves in terms of adding the data sources they need, building the dashboards they need.

So what we are trying to do now is really build more out of the box experience and solutions on top of that-- solutions like synthetic monitoring, IRM, and a more turnkey experience for monitoring both on-prem and on the cloud. And we intend to expand that.

CRAIG BOX: We've talked a lot about metrics and monitoring platforms. You've also released a project called Loki which aims to bring that same experience to logs. Tell us a little bit about Loki-- how that came about.

TORKEL ODEGAARD: Loki is really the brainchild of Tom Wilkie from Kausal, and from this cloud-native Prometheus domain. He really had a vision for building something like Prometheus but for logs-- something that had a similar or the same label set as the metrics. So it could communicate in a Kubernetes setting, have the same labels for the jobs and pods, so you could really navigate between metrics and logs because of that shared label set. And also a vision for doing logs in a much more efficient and scalable way than the traditional indexed logs.

CRAIG BOX: Is Grafana still a tool only for visualization and monitoring data, or do you see it as a general purpose data visualization tool?

TORKEL ODEGAARD: I see it very much as a general purpose data visualization tool that we also build more opinionated solutions on top of. For example, the explore feature in Grafana is really tailored to the more DevOps troubleshooting use cases and IT monitoring use cases. But Grafana as a whole is much more open in terms of what the domains and problems you use it for.

ADAM GLICK: You have a Microsoft background, so you're probably familiar with an old tool that I remember I used to use called Crystal Reports for visualizations. And that space has always had players it-- modern times you've got Tableau and other service-based and software-based pieces. Almost all of them are closed source-- sometimes available from cloud providers. Why the focus on open source in this area and going against the grain?

TORKEL ODEGAARD: And that's been, I think, one of the reasons why Grafana has been so widely adopted. There is no huge competition in a very generic domain-agnostic data visualization tool that doesn't restrict what data source you use. You can bring in many, many types of data sources, different types of data, so there isn't that much competition really on the open source side. So Grafana has a little bit of free rein here, which is something we've capitalized on is really this being a single pane of glass, being a platform where you can bring in data from anywhere.

CRAIG BOX: Have you considered expanding the visualization pieces of Grafana to address general data use cases like things like Tableau and Looker?

TORKEL ODEGAARD: Yes, I think definitely we do support SQL data sources, so similar to Looker and Tableau. But we have ways to go there in terms of really making the UI experience for working with BI data and the visualization capability better, but that's definitely on a road map to make that use case even better.

ADAM GLICK: What's the most unique use case that you've seen Grafana used for outside of what you expect from monitoring?

TORKEL ODEGAARD: The one that I've mentioned a bunch of times is monitoring beehives. This was a guy I met at FOSDEM who runs this beehive society that builds sensors that measure the weight and temperature of beehives. And he showed me Grafana dashboards, and he described all the insight that he could get from just looking at a weight sensor of a beehive to sort of see how they were building up nectar, what they were doing during the day. I thought that was kind of incredible.

The other case would be someone visualizing hospital wait queue times in Japan. So they had a dashboard that showcased all the available hospitals and wait times.

CRAIG BOX: Still monitoring though-- you've not seen anyone do say animation display or any other sort of visual hackery?

TORKEL ODEGAARD: Oh. There's a monitoring art plug-in that can draw graphs that become very artistic.

ADAM GLICK: Any plans to do other kinds of visualization for types of data, things like maybe geospatial data or other large data sets, but aren't necessarily kind of logging and monitoring as we traditionally think of in an IT sense?

TORKEL ODEGAARD: The thing that we're interested in exploring is making Grafana dashboards more interactive. So you can have a control surface as well, so buttons that interact with hardware and systems. That I think could be really an evolution of a dashboard. So you're not just passively looking at data, but also have buttons and things that can talk directly to the sensors and the hardware.

ADAM GLICK: I personally look forward to my tricorder, so I hope so. On August 17, you announced that Grafana Labs had closed a $50 million series B funding round. First off, congratulations.


ADAM GLICK: And secondly, with that, how do you convince your investors to invest in a product where you're effectively giving away the source code?

TORKEL ODEGAARD: By showing that you have real revenue, and customers, and growth, that you're a real business, that you're not just building on hype and potential-- so future commercialization ideas. We have a real solid plan for both our on-prem differentiation, and our enterprise offerings, and our cloud offerings. So we have a good and sustainable way to capitalize on the open source project's success. And I think it really makes investors believe in us.

CRAIG BOX: On that note, you've recently released the Grafana Metrics Enterprise Edition. What can I get from that product?

TORKEL ODEGAARD: That is a huge thing for us. Our third product, if you will, that basically turns the thing that powers our cloud, Cortex, and makes it available on-prem with additional security and instance management tools that you normally don't get if you run open source Cortex. And you get support from our awesome Cortex team as well. So that's kind of the foundation, and it's going to expand upon that, and it's going to have features that integrate with Grafana Enterprise as well.

CRAIG BOX: You're still based in Sweden. And you mentioned before that Grafana Labs is a US company. What is done in which location? Did the investors ever ask you to move? Was it pressure to try and become based in one place?

TORKEL ODEGAARD: No, this has been an amazing learning experience, but Grafana really has been a remote first and distributed company from the start. With gravity in US east coast and Europe, but that is slowly evaporating. And we had offices in New York and Stockholm that were our two bases, but we are now, since before COVID, distributed all over Europe and US and everyone working from home. I haven't been in the Stockholm office since March.

But our culture has been a huge deal for kind of recruiting and hiring, especially in this COVID year. But people have been really appreciative of being able to work from home, and being able to live close to family. And having a company where everything is kind of set up to be distributed first has been really incredible.

ADAM GLICK: That sounds great for people. I know as people, especially now, are looking at doing things distributed-- and the open source community has a lot of experience in this. How do you handle the camaraderie aspect of helping to build connections between people in the organization when they aren't sitting around the water cooler so to speak?

TORKEL ODEGAARD: I think that's definitely a big problem. And especially with people that don't have families, there is greater need of social interactions. We have office hours and different gatherings on Zoom. But before COVID, we had many physical get-togetherings as well. And that's been the struggle now this year where we don't even have that. So everyone's kind of locked in their home for months. And that's been our biggest challenge, that we don't have the conferences and off sites that we usually had.

ADAM GLICK: Finally, anyone who follows you on Twitter will see that you're quite a fan of science fiction. What's one book that you think everyone should read?

TORKEL ODEGAARD: I'm a really big science fiction fan. I come from old school things like Asimov and Arthur C. Clarke, but if something more modern I would say Alastair Reynolds's "Chasm City." He's a really amazing author, and that's an amazing book.

CRAIG BOX: Well, I shall be sure to look that up. And thank you very much for joining us today, Torkel.

TORKEL ODEGAARD: Thanks! It was a pleasure to talk to you.

CRAIG BOX: You can find Torkel on Twitter at @torkelo, T-O-R-K-E-L-O. And you can find Grafana Labs and its projects at grafana.com.


CRAIG BOX: Thanks for listening. As always, if you've enjoyed the show, please help us spread the word and tell a friend. If you have any feedback for us, you can find us on Twitter at @kubernetespod, or reach us by email at kubernetespodcast@google.com.

ADAM GLICK: You can also check out our website at kubernetespodcast.com where you'll find transcripts and show notes as well as links to subscribe. Until next time, take care.

CRAIG BOX: See you next week.