 Live from San Diego, California, it's theCUBE. Covering KubeCon and CloudNativeCon, brought to you by Red Hat, the CloudNative Computing Foundation and its ecosystem partners. Welcome back, 12,000 here in attendance for KubeCon, CloudNativeCon 2019 in San Diego. I am Stu Miniman, my co-host for this afternoon is John Troyer and happy to welcome to the program recently out of stealth two gentlemen from Chronosphere. See to my right is Martin Mao, who is the co-founder and CEO and his co-founder, Rob Scalington is also the CTO. We've stated on theCUBE actually, you understand where this conference is, where co-founder and CTO is the most prominent title that we seem to get on here because that's the type of geeks we love on the program and in this community. First of all, congratulations on the launch and thanks so much for joining us. No worries. All right, when I've got the founders on, I'm going to start with the why. Tell us kind of the problem statement, where you were coming from and what led to the creation of Chronosphere. For sure, for sure. So with Chronosphere, we found a actual gap in the monitoring market in a very crowded monitoring market. We found a gap and the gap exists when companies with very large, complex technology stacks or large enterprises move on to CloudNative Technology in Kubernetes. So with this migration, what we found was there's actually a lot more monitoring data being produced because there's a lot more pieces now. We're moving from monoliths, micro-services, we're moving from physical machines to VMs to containers and pods, and that generates a lot more things that you need to monitor and track and not only a lot more things, but you're generally monitoring the relationship between these things. So as the number of things increases, the number of relationships exponentially increases. So yeah, that's the sort of problem we're solving is like monitoring all of these things at large scale and we couldn't find anything that could even store all of these things. So that's the sort of problem. All right, so what is the background of the team that made you in the position to work on this problem? Yeah, great question. I mean, me and Martin go back quite a few years. I appreciate his wedding only very recently actually. And yeah, we basically worked together at several different companies. You know, I think both of us are entrepreneurial at heart. I'll let Martin talk a little bit more about the last few years. Yeah, so like a few years ago we started working at Uber and at Uber we went through this migration to our native Kubernetes and through that migration that's when we sort of had to solve the problem ourselves and we solved the problem at Uber with an open source project called M3. That's really where this whole thing started and Chronosv is sort of building on top of M3 now and providing a product on top of the open source platform that we created. Can you talk a little bit about the business? I noticed that there are many ways of approaching open source in 2019. You know, open core and but also as a service. So can you talk a little bit about how you've approached your business model? Yeah, for sure. So we're very much in the position of building more in the camp of as a service, right? Because you know, a lot of companies do do open core and they sort of go into the enterprise support model. We sort of didn't want to go down that route. And also with our open source product, it's not really an end to end solution in itself. Like you use open source M3 but you still need to plug it together with other things yourself. So what we really wanted to do was to give customers an end to end solution that was built on top of the great technologies we built with M3. But really it solves the problem sort of end to end. We do that best as a service, so. Rob, maybe you can help explain M3 a little bit for us as to how that fits in the landscape but what it works with and the like. Yeah, of course, yeah. Yeah, it's basically at its heart a metrics platform that is built on at first the lower layer M3DB, which is a distributed time series database. And then on top of that, we have basically an aggregation platform that is actually aggregating a lot of the samples and metrics that we're collecting. So we can really do some transformations on the data as it comes in before it's stored in the database itself. And this lets us do a lot of smart processing of what signals actually matter, what signals don't matter, storing them in a way that can be accessed much faster than like other typical systems that don't really do any aggregation before it gets stored. And then we have of course like a query engine that works with this distributed set of data. And so it's really a database that was designed from day one to be a metric store. It's not built on Cassandra, it doesn't use ROXDB at the lower layers. It literally every part of it was built for this purpose. Can you talk a little bit about dimensionality and cardinality? As I look at this observability monitoring space, I see a lot of current discussion about that and frankly a little bit of fighting. And I'm not always, I can kind of see but I think why it's important but kind of what are some of the reasons and what do people do by having and what is it actually? Let's go with that. Yeah, for sure. So with this hot topic of like high cardinality, high dimensionality is what I was talking about earlier where as you move into a cloud native world, you're now monitoring things at like a pod level. So it's like instead of tracking things on like a per host level, you know tracking things on a per pod level now. And that just adds- You're tracking more things per pod. More things per pod and like every pod, you know that these are ephemeral pods now so I don't even live for February long. So you end up having more pieces of data and they kept you on for a shorter period of time and now you need a system that can store all of these pieces of data because you want to see them uniquely. So you want to monitor each individual pod to see exactly what is run at the finest levels, right? So you actually need technologies that can store a lot more data than you could before. And you know, adding to that, there's a lot more people running with like mobile applications that use, you know that are running in markets all around the world using different cell providers and different backend services. You may deploy your backend services multiple times a week or even a day. And if you want to tag, you know, the metadata on and slice and dice by that metadata with your business and with your applications and your system, that requires, you know, adding yet another dimension on your data which adds to the cardinality. Every time you add a dimension, you know, that just multiplies the cardinality of your existing data set or monitoring data. And it quickly adds up a lot, right? So. Martin, maybe since you're just out of stealth, give us some of the speeds and feeds. You know, product GA is globally available. Series A funding, you know, who's behind that. Yeah, for sure. Some of those pieces. Yeah, so we just kind of still two weeks ago, we closed our Series A a few months ago actually. It was led by Greylock. We raised 11 million dollars and our partner at Greylock is Jerry and we like him very much. And you know. Hi, Jerry. Good to hear from you. And you know, the state of the company and we are currently in private beta right now. So with our hosted platform, we are onboarding customers into a private offering right now. And early next year, we'll sort of open that up to a more public beta. Yeah. And the way folks would use this, you'd be using Prometheus or Graphite or something and you'd be, so you'd be, you'd have tracing, you have logs, you have other things and you would be plugging all of them into your service. Is that? Yeah, it's a great question. So you mentioned two of the technologies. So if you're using Prometheus or Graphite, like the Graphite metrics, both of those can be pushed into the M3 system for sure. We actually just announced a tracing integration this week at KubeCon actually. Rob gave a talk about that integration earlier this week at KubeCon. We haven't moved into logs yet because the way we look at the problem is not from like a sort of a providing and one stop shop for all observability solutions. We actually look at it from a use case perspective, right? So the use case we're looking at is like real time monitoring and remediation. So tracing is part of that story. It's a critical part of that story now to add additional context when you get alerted based on your metrics. But we haven't quite moved into logging yet. And we don't really want to solve any of these problems without knowing it'll work at scale. Like the fundamental reason we even built the open source project in the first place was we were dealing with cardinality in the tens of billions of unique time series. And so we don't want to just kind of like roll into every single feature under the sun. We really want to solve it once correctly and be able to systematically roll that out to enterprises at scale. I mean without talking too much about Uber and any Uber secrets, I mean it seems like the game has changed with that kind of a scale of you could not have done. You can't run Uber if you're tracking all those cars like literally without some sort of a tracing like high cardinality sort of a system, right? Because you're literally tracking cars all over the world, people all over the world, routes all over the world. Exactly, exactly. We're uniquely positioned. We had the requirements to solve it at such a scale and that's how we had to build this technology to solve it for that unique situation because technologies ahead of time do not really have this use case to solve. So that's why we couldn't find anything out in the market to solve it at that scale. That's why we sort of had to build our own to uniquely solve it for this use case. And yeah, and I would add to that that typically engineers at larger organizations tend to want to organize everything very nicely and split it up and really control how they're monitoring their data. But we've noticed actually, definitely over the last few years, more and more people are open to letting people just start collecting random data that is relevant to their systems that they're building as they're rolling it out even as they're experimenting with it. And systems today that are built from scratch to deal with to be as efficient as possible with very unstructured data is becoming wildly popular because that's how developers want to develop software. They don't want to have to slice and dice it neatly and package it up and pass it on to others to run. They want to basically slice and dice however they want to and dynamically as they scale up with that. I've always enjoyed every SQL schema I've had to do. Or change. All right. How have you found the show? How's the reception been, give us a little bit of the vibe of the show and how it's been going for you? Yeah, it's been fantastic for us actually. So we just came in itself, so the name is still quite new. But yeah, we've had a bunch of folks that have built the whole day. They've been giving a demo on the product. So a lot of companies are getting excited about it. I think we're solving it at a scale and that really resonates with a lot of the people here at the show. We're still solving it at a scale, but we're solving it at a scale that's also in a cost efficient way as well. So that's really been received quite well so far. Rob, you gave some sessions. What kind of feedback are you getting from people? Is the problem statement that we talked about at the beginning, resonating with people that you talked to? I mean, I was really pleased to hear that after my session today that a lot of people that came up to me and said, you've never really seen metrics being linked to traces, the way that we're doing it. In fact, that's the first time they've ever seen a demo that can do what we're kind of trying to upstream. We're actually upstreaming a lot of those changes in the open source world as well at the same time. And so, what we've found, especially in a lot of the companies today that are pushing everything forward with development wise and how they're running operations is that they're using a lot of features in open source. And then those features are battle tested in open source. Generally it becomes abstract to the point where it works for a very large amount of people. But then when they need to scale it up, that's when it becomes difficult. So I think that a lot of people have been very positive with basically us being able to also push forward the feature part of open source. Back upstream into the M3 project. Yeah, and also into Prometheus. So I'm an open metrics contributor and that's essentially an exposition format that's built on the Prometheus exposition format. So it's going to become a standard way of exchanging metrics from one system to another. And that's going to basically commoditize and democratize the exchange of metrics and make a lot more systems interoperable with each other, which we fundamentally believe with as well, of course. We're developing an open source and we believe that these systems need to play nicely together so we can build, have building blocks that large companies and organizations can all share and build better things on top of. All right, so looking to go to public beta early 2020 is what we said. When we come back in 2020, what are some of the key KPIs and metrics that you'll be looking at to be successful in your first year out of stealth? Yeah, it's a great question. So some of the KPIs we're looking at doing is coming out of public beta, making that available to a large range of companies because right now we're sort of onboarding companies sort of one or two at a time. So yeah, seeing how many companies adopt the product and also we're again adding more features over time for that particular use case of like, monitoring your technology stack in your business in real time. So it'll be a lot more features coming down the pipeline and a lot more customer adoption hopefully along with that. And I will do also say, our host of platform is really about offering like deep isolation between our tenants as well. So basically when we, in the next few months come up, we want to make sure that it works basically like clockwork then and everyone can, we can roll out and scale that highly isolated platform for tens and hundreds of organizations and thousands eventually. So and doing that at scale is hard. So I think yeah, we'll see how we're doing with that. For sure. Rob, Martin, congratulations on coming out of stealth. Look forward to hearing more and thank you so much for joining us. Right, thank you so much. For John Troyer, I'm Stu Miniman. We'll be back getting towards the end of three days of all the wall covers here at QCon, cloud native pod. Thanks for watching this video.