NanoVMs Founder Ian Eyberg on Disrupting Linux

We run Linux software faster and safer than Linux itself. Now, that probably doesn’t mean almost anything to your average non-technical investor. However, it means quite a lot to any software developer. Let me explain. The most popular server-side operating system in the world is Linux. Google is entirely built on Linux. Facebook is entirely built on Linux. Uber is entirely built on Linux. If a company is using a public cloud like AWS, they are probably using Linux. All of your favorite startups websites are probably on a Linux server. Here’s the rub. Linux is roughly 30 years old now and is largely the same design as Unix which is 50 years old. Unix came out in 1969 and was built to run on machines like the PDP-7 and PDP-11. It was architected to run multiple applications by multiple users back then because the machines were too expensive and too big (literally taking up an entire wall) not to. Fast forward to today and many companies don’t just have one server anymore. Instead of one database they might have hundreds or thousands. Some banks use so much software that they own not just one datacenter, but multiple datacenters. The world is consuming vast amounts of software, and yet we are still using an operating system design from the late 1960s. Our goal is to provide a faster and safer system for these server-side applications to run on.

I was working on an application performance monitoring system at the time and had started reading a bunch of whitepapers on the subject of unikernels. To be clear we didn’t invent the idea — it was stuck in academia forever. At the time no one had built a Go unikernel, a programming language I was using heavily at the time. And so we had that done. The first time I booted it up on KVM I just immediately knew that this was the future and just dropped everything to focus on it.

I think it’s a confluence of various influences and circumstances. For instance I must’ve been 11 or 12 when I was handed a handful of slackware floppies to install Linux back in 1994 or so. Our scout troop met about a block away from the local university and all the people that helped out from there were electrical engineering students that were taking a handful of computer science courses. Remember the first dotcom boom was just getting started at this point too.

Second, my dad was a successful general contractor and so early on it was very clear that it was perfectly normal to build your own company as long as you don’t mind working. Finally, work ethic and perseverance. The news makes it look like some people get insane amounts of luck, but I think for the majority of us it’s just an awful lot of work. And as for perseverance there are so many times when the world just tries it’s hardest to make you give up. Especially here in the SF Bay area, a good engineer can make a *lot* more money, at least early on, just by working at an existing company. So money alone can’t really be your only motivator. If you’re going to invest this much time of your life, there’s a real opportunity cost.

We are firmly in the camp of “software infrastructure.” A lot of people compare/contrast us with the entire cloud native computing ecosystem, but there are many other companies with different takes on the space as well such as HashiCorp. We are basically creating a new category that combines traditional devops software with an opinionated take on security. Traditional security software is filled with very reactive types of solutions that scan for hacked systems or scan for systems that are going to be hacked.

Our approach is completely different. We realize that the whole point of an attacker trying to break into a system has nothing to do with the bug they found or the exploit they are using. Comparatively speaking if someone was breaking into your house, they come through a door or a window or something like that, but that’s not why they are breaking into your house. They have come for your guns, money, and flatscreen tvs. It is the exact same thing on a server. An attacker could care less about your software. They want to dump the database (eg: every single data breach news story ever) or install an illicit crypto miner or something else. The implication here is that it always involves running other software. Unikernels explicitly prohibit that behavior as they are only designed to run one program at a time, and they don’t even have the notion of an interactive session or the notion of users/passwords.

To us the ecosystem is going to be one giant land grab. Our company alone can’t even fulfill the size of this market. We already run things like redis and nginx and other popular Linux-based software faster and safer than those run on Linux. Redis Labs itself already has a valuation measured in multiples of billions. That’s just one company. How many different database companies are there? Not only can we run the existing software better, but we are creating an environment for other companies to come in and create better versions of the wealth of existing software out there. This is going to be such a massive, massive opportunity.

We have a core group of kernel engineers now finally — thankfully, because trying to do this yourself is completely insane. Some of our engineers have worked at networking giants such as Cisco and storage companies such as Coraid. Looking towards the future we are looking to help build our outbound team and expand with other roles such as devrel as well.

We have direct competitors and indirect competitors. Our direct competition working on unikernels and unikernel-like software I’m not really concerned about because we are so far down the road. And I know just how much work it takes to do what we’ve done so far. We position against a lot of indirect competitors though. So this includes companies in the cloud native space (eg: Docker/k8s) and other technologies used in the software infrastructure space. There are a handful of very large companies such as VMWare, NEC, IBM, ARM, etc. that have research teams working on unikernels. But none of them are working on products yet, and I’m not afraid of them either since their modus operandi tends to be buying their way into markets. What I find really interesting though is that technology-wise we are positioning against Linux itself. To be clear so I don’t rub anyone the wrong way — we’re not dishing on Linux at all. We’re all heavy Linux users ourselves and have been for quite some time. There are many things such as desktop Linux which we will never build support for. What’s interesting is that there is no PM for Linux. There is no one company that controls it. Think about that for a bit.

We have many different paths of engagement. We’ve done a handful of one-off project integrations and larger feature requests. We sell licenses for support and enhancement. We sell the Nanos C2 product which is a graphical administration tool for using unikernels on the cloud or in your own datacenter — helpful when you are running a fleet of them. We sell Nanos Radar which is an APM solution specifically tailored for unikernels so we can do really interesting things that some of the other solutions such as DataDog and NewRelic can’t — although to be clear a lot of that software works out of the box.

To be honest it happened on a whim. We did our first raise with Republic last year. We weren’t planning on doing another one, but the limits changed from $1 million to $5 million and the individual investor limits changed and so it made a lot more sense for companies like ours that are still in the seed stage. A lot of founders have to give up board seats even on their first institutional round, and that can be translated as control. It’s very important for a company that is this technical to maintain the vision. There have been other unikernel companies out there that were forced to pivot because they couldn’t get the R&D done fast enough to get enough market traction. A lot of non-technical people might not fully appreciate the amount of work that is necessary to work on a project like this. It’s not like the dog walking app where you can kick out the first rendition of it in a weekend and focus on go-to-market. Just to give you an idea — integrating with even one instance type on a new cloud could take a full-time engineer multiple months to add the corresponding drivers.

It’ll entirely depend on how much we raise, but market education is one area in particular we’d really like to shore up so that means hires such as devrel and pre-sales engineering.

We’re building a massive company. and it’s not just the company that is going to be big. The total addressable market is downright massive. If we are even remotely successful, there are going to be another 100 companies in our wake, and they won’t necessarily be our competitors because our software will be enabling them to do things we haven’t even thought about.

Some people might think that the ridesharing market was invented by Uber and Lyft, and by extension the gig economy. But the reality is that there was already an existing market before them to facilitate some of the same basic services — it’s called taxis, and that’s the nature with category creators like us.

As much as I’d like to ring the bell, an acquisition makes far better sense, and like any company at our stage we’ve had corp dev sniff around. I keep an Excel sheet dedicated just for them. Having said that, our focus right now is grow grow grow. We are not interested at all in trying to sell right now. Part of the plan is to affect real lasting change. We’re not trying to “raise the collective consciousness” to quote Adam Neumann, but I feel we can make a pretty large dent in the confluence of software infrastructure and cybersecurity.