text stringlengths 0 2.02k |
|---|
So then it becomes just a question of principle - do you still really believe in open source? Do you still believe that other people should be able to check your core and make sure that you're not doing anything fishy, or help you find bugs, or that kind of stuff? So it becomes a matter of principle than a matter of bu... |
**Break:** \[01:00:50.09\] |
**Adam Stacoviak:** Coming up in this segment, we're talking with Paul Dix, co-founder and CTO at Influx Data. Paul shares his perspective on running an open source business, how Influx Data is innovating their commercial offering while having a permissive MIT license version of Influx DB. Paul also shares his thoughts... |
**Jerod Santo:** So Paul, tell us your name, tell us your company, and kind of your view of the open source world, and where your opinion is coming from. |
**Paul Dix:** Yeah, so I'm Paul Dix, I'm the CTO and co-founder of Influx Data. We're the makers of InfluxDB, which is an open source time series database. I created it in 2013 and I've been initially running the company, and then a CTO, which I'm still doing to this day. |
My experience over that time has basically been trying to build a business around an open source software project, particularly in infrastructure software, and in databases in particular. Elastic, obviously, I'm very familiar with. I saw it when they were initially becoming a company, I remember the project early on...... |
**Jerod Santo:** Very similar software, very similar business model... It looks like Influx is MIT. Can you tell us your license selection and how your business works around it? |
**Paul Dix:** Yeah, so all of the open source software that we create is MIT-licensed, and our business model is basically -- so we are at this point an open core business. So there's open source InfluxDB, which is MIT-licensed, and people can do what they want with it; it works essentially as a single server. |
We have a fork of the open source project that is closed source and proprietary. If you want high availability or scale clustering of InfluxDB, that is our commercial offering. Essentially, we don't put clustering features into the open source. Everything else is fair game to go into the open source. If it has to do wi... |
So we launched this as a managed service inside of AWS in (I think it was) April of 2016. We launched it as basically like an on-premise software product that people could buy in September of 2016. Our AWS service is still running to this day. Essentially, what that is - it's the closed source software spun up; a custo... |
The on-premise version is essentially you buy the software from us, it's an annual subscription, and then you run and manage it yourself. And that's either in your own data centers, but plenty of people are also doing it inside of AWS, GCP, all that kind of stuff. |
Last year - or I guess late 2019 now - we launched basically version two of our cloud product, and that essentially is a very different kind of thing, because it's not just a database, and it doesn't look anything like the open source software that we create. The API is the same, but the underlying architecture and how... |
So the model we switched to with 2.0 - we essentially moved to a cloud-first model. So we deliver that product continuously as like a SaaS service, and then over time, some of those features get pulled out into the open source InfluxDB. |
**Jerod Santo:** Is that because of a realization that the other way wasn't working well enough? Why did you switch to the cloud-first model? |
**Paul Dix:** It has nothing to do with open source versus closed; it has everything to do with software delivery cycles. Before, we looked very much like an enterprise software company. We'd have anywhere from two to four feature-bearing releases a year, which could then get shipped to our cloud customers and shipped ... |
The problem with that is you don't get that many cycles of iteration, and each release is super-painful to do, because there's so much code wrapped up in it... So I really wanted to move to a continuous delivery model, so we could get much faster feedback, features out to customers quicker, and the individual releases ... |
**Adam Stacoviak:** How does that trickle down to your open source then? How does InfluxDB, the open source benefit or not benefit from this switch? |
**Paul Dix:** I think the benefit is that by the time something lands in open source, we've already vetted the features and vetted its functionality and how it works inside our cloud product. The thing is with our cloud product we're able to iterate on it and release fixes very quickly. Once you ship something in an op... |
The drawback is I think it reduces the collaboration with the community in terms of what we're developing and how it's getting done, and all that kind of stuff. It basically makes the open source a downstream kind of product. |
**Adam Stacoviak:** Yeah. \[unintelligible 01:08:30.15\] tweet you've put out, which is "My own preference is to keep open code permissive, and open, and have a clear strategy (as you just depicted here) with how that code will be used in the commercial offering." So you're eating \[unintelligible 01:08:42.27\] which i... |
**Paul Dix:** Yeah, so what I've just described is basically our 2.0 model... But I'm actually trying to move even beyond that, over to what I call basically like a complementary software model. We have a new project that we announced a few months ago, called InfluxDB IOx, which is basically the new core of the databas... |
The way that we're building this out is essentially there's the open source thing, and then there's another piece of software that we have that is closed source. And as a whole, the system is designed to be two pieces of software, one of which is totally open and permissive, permissively licensed; you can do whatever y... |
And then the other piece, which is the software that we're writing to be able to run the open source software in our cloud offering. So the reason why I say it's complementary is because what I want is I want our cloud product to be running the open source bits exactly as is, exactly as the open source community experi... |
**Adam Stacoviak:** Yeah. Two masters. |
**Paul Dix:** Exactly, yeah... With IOx-- |
**Adam Stacoviak:** In terms of serving two masters, is what I mean. Like, literally, you're serving two masters. You have two different projects. It's very difficult to serve both easily. |
**Paul Dix:** Absolutely. Internally, we have a separate team that works on the open source bits, versus the people working on the closed source cloud product. |
**Adam Stacoviak:** Right. I asked you the benefits to the open source, and it seemed like the benefit to the open source was obvious, but they're different, because they're separate. That's what it seemed like. And I was gonna ask you about that, because it seems like, with your cloud too that you mentioned, you can o... |
**Paul Dix:** Yeah, it's a bit disconnected. Whereas with this new model, again, my goal is -- we offer it as a cloud product first. We're not doing that yet. But then later we'll offer it as an on-premise product. But the idea there is that people who adopt IOx and deploy a bunch of servers and stuff like that - if th... |
It's very, very different than our old InfluxDB 1.x enterprise model, where our enterprise product is a replacement for the open source InfluxDX. I think that's a heavier lift, and it's a bigger ask for users to replace their open source bits. I think it's better if they're able to run the open source bits and continue... |
**Jerod Santo:** There's some interesting ramifications that I think I would love to see play out as you go about deploying that new idea. If I'm understanding correctly, it's kind of like the open source bits is like the core software offering, and then the proprietary stuff is like a management controller, or like a ... |
**Paul Dix:** Yeah, that's a good way of thinking of it. It's basically all this code that we have to write to offer it as a service - operations, back-ups, deploying new versions of it, management, all this kind of stuff. |
**Jerod Santo:** Gotcha. |
**Paul Dix:** The other thing is we wanna be able to offer that as an on-premise piece of software. Another way to think about this is the open source thing represents the data plane, whereas our closed source product represents the control plane. But the way the two interact is the control plane interact with the data... |
**Adam Stacoviak:** Yeah |
**Paul Dix:** And the thing is, we don't have to worry about our open source bits competing with our commercial bits... |
**Jerod Santo:** Right. |
**Paul Dix:** Because the responsibilities of the two pieces of software are clearly delineated. So there's no reason for people to put control plane bits into the main open source project. They would have to create a second open source project to do it, which would make sense. |
**Jerod Santo:** Right. At that point you're kind of just deciding what is control plane and what is data plane. And that's kind of the same concept of like what goes in core and what goes in proprietary, isn't it? Like, what about back-ups? Well, it could go right into our core offering, but it's more of a control pla... |
It seems like you still make those decisions, you just make them in -- the two pieces of software are further apart, perhaps. |
**Paul Dix:** I view them as further apart. When I think of open core businesses, I think of businesses where the commercial product is a replacement for the open source product. This is not that, and it's designed specifically not to be that. Take DataStax, for instance. DataStax Enterprise is a replacement for Cassan... |
**Jerod Santo:** Gotcha. |
**Paul Dix:** I think a good example is Google and Kubernetes. Open source Kubernetes certainly doesn't represent the entirety of GCP and all the software that runs that... But obviously, Google has a vested interest in driving Kubernetes forward, and GCP happens to be one of the best places to buy Kubernetes, to opera... |
**Jerod Santo:** So what's your thoughts on the server-side pubic license and Elastic's move? You obviously prefer this other way of going about it, but... Do you think it was smart by them, short-sighted? What's your take on that? |
**Paul Dix:** It's not the move I would make... To be totally honest though, to me it's not really about a license choice, it's more about how they intend to drive innovation that drives commercial value. And the truth is, I own stock in MongoDB, which is obviously SSPL-licensed software, but I do not own stock in Elas... |
I think them changing the license is more a symptom of the fact that they're getting out-innovated on their cloud offering. If they had a cloud offering that was demonstrably better than Amazon's Elastic service, they would continue to be able to drive revenue and drive people to it. If they're so upset because they fe... |
**Adam Stacoviak:** A lot of sounds over there... \[laughter\] |
**Paul Dix:** Yeah... They are a disservice to the community, because then you can say "Oh, community members saw your code, or whatever. It just means that people can't start competing project with you without putting themselves at risk of being accused of taking your code, or whatever. I prefer "Open code is open, cl... |
The thing that kind of annoys me about the whole Elastic/AWS stand-off is both of them are trying to put forth this position that they have the moral high ground, they have moral superiority over the other one. Amazon's saying "Oh, we're protectors of open source, so we're gonna launch this fork", or whatever. And the ... |
It was always obvious when they launched that that fork is what it was ultimately going to become... Because Elastic was gonna take the stance that Amazon's stealing from us, so we're gonna change the license on more and more of our code... Which is then going to give Amazon no choice but to fork. |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.