Fly.io is the First CDN for Docker Containers

Fly.io is a CDN startup that has been flying under the radar for some time. Founded in 2016, they’ve raised $37M in Series A and B. The team consists of engineers, and one product manager, with no marketing or sales, as of today.

The startup has a product stack that differs from anything we’ve seen in the market. They describe themselves differently, but one description that stands out is “CDN for Docker Containers.”

They are not a traditional CDN, but neither are Akamai, Cloudflare, Edgio, and Fastly. Today, there are no traditional CDNs, so comparisons fall flat when a competitor identifies them as such.

Apart from Fly’s software stack, they have a dedicated global infrastructure, which includes hardware and a network. That means they are not piggybacking off of AWS or other cloud providers.

Andreessen Horowitz (a16z) describes Fly as “one of the fastest-growing core infrastructure companies we have ever seen.” In the CDN industry, Fastly and Cloudflare reached the magic ARR of $100M in ~10 years since being founded. EdgeCast generated $100M ARR in a shorter period. So if a16z is talking about the fastest revenue growth they’ve ever seen, hats off to the Fly team.

Since Fly competes with the CDNs mentioned above, let’s examine their business model, specifically the product stack, growth model, messaging, people department, and gaps. Also, we’ll compare them against others at a high level.

Background

  • Company: Fly.io 
  • Founded: 2016
  • HQ: Chicago
  • # of Employees: ~32
  • Raised: $37M
  • Founders: Kurt Mackey, Jerome Gravel-Niquest, and Michael Dwan
  • Investors: Intel, Andressen Horowitz, Dell Technologies, and others
  • Product: CDN for Docker Containers
  • Customers: SMB market

Company Messaging

From the website, it’s challenging to get a clear picture of what they do. It requires much digging into blog posts, job postings, and documentation. Fly.io has described itself as the following:

There’s no consistency in the messaging. But that’s okay since Fly is a startup; it will improve over time. In the meantime, we’ll describe them in our own words at the end.

Product

Regarding their product stack, it’s dripping in awesomeness; job well done by the Fly team.
They use innovative technologies like Firecracker, a new virtualization technology developed and open-sourced by Amazon.

Firecracker is a revolutionary product likely to have the same impact as DynamoDB. Today, this product powers millions of container workloads and functions at AWS. Imagine before this; every function ran on an EC instance. Ouch.

Some of the other parts of the product stack, defined as the tools powering the Fly platform, are the following: 1) Container orchestration system developed in Go 2) Anycast network called fly-proxy developed in Rust 3) WireGuard for VPN tunnels 4) eBPF and XDP secret sauce 5) Support for Deno Javascript runtime, and a whole lot more. 

In addition, the Fly team is highly skilled in database technologies. For example, they managed to hack PostgresSQL into an edge database that launches in a single command. That means an ACID-compliant database will be spun up and deployed anywhere. Additionally, read replicas of the Postgres instance will spin up by the orchestration engine to support the demand if it is there. How cool is that? That experience comes from another startup called Compose.

The CEO of Fly started a company called Compose in 2011, which it sold to IBM. At its peak, Compose “spun up over 100k databases.” Innovation in the ACID-compliant edge databases has been slow in our industry. Cloudflare recently announced support and integration around SQLlite, and Akamai rolled out managed database services supporting MySQL, PostgresSQL, Redis, and MongoDB. Fly should be able to lead innovation in this area due to its heavy-duty experience.

 

Source: Internet Archive. Oct 2016

Growth Model

The Fly pricing model is the same as competitors. They charge for data transfer, compute, SSL certs, storage, etc. The sales growth engine is PLG. PLG is an online model that offers different pricing tiers, including a free tier. And for startups, there is no need to hire salespeople to sell the platform. Instead, the products must be in a position to sell themselves. Cloudflare, Fastly, and AWS started this way. Once Fly reaches a threshold, it makes sense for them to hire salespeople.

Today, the startup has ~33 employees even though it raised $37M. That’s good because, as we learned from Subspace, don’t go overboard in hiring many people and spending a ton on infrastructure. Financial discipline is necessary, especially with a recession banging at the door.

Product-Led Growth

Today’s big gap is the lack of a Growth Team responsible for PLG. With the right team, they can accelerate Fly’s growth.

In the PLG environment, multiple tools are a must. For example, tools such as Mixpanel or Amplitude, HubSpot, Google Analytics, Preset, etc. must be integrated and tailored to track the customer experience lifecycle.

The lifecycle starts from the first time a visitor lands on a page to when they sign up for a free trial, then upgrade to a paid subscription.

The startup has been around for six years; thus, even if they were to hire an entire PLG team today, it would take two years for the Fly team to see the results. By the way, PLG folks are in high demand. So it takes months to find them. If the team had been in place now, perhaps, ARR would be double or triple what it is today.

Security

The other gap is the lack of a security product. Hardening one’s infrastructure is one thing, but offering a cutting-edge security solution is critical. A WAF or variant thereof is needed at a bare minimum because security is top of mind for some use cases, and it’s more important than performance.

Competition

Fly has a head start of 1-2 years in the CDN industry. The competition will follow suit. At some point, deploying containers at the edge will become the standard. The CDN industry is a rat race, and unique products stay that way for only so long. Fly.io must have already developed and rolled out another killer product when the competition catches up.

Fly reminds us of Section. However, Section is a marketplace that runs on existing cloud provider infrastructure. StackPath has a container solution, but it’s nowhere near Fly’s product quality or depth. Cloudflare, Fastly, and Edgio don’t have a container solution today. I doubt Fastly ever will. Cloudflare and Edgio might have one someday. 

Akamai inherited a killer container solution, comprehensive in nature, from its acquisition of Linode and is in a position to catch up to Fly soon.

Summary

Fly.io is a Container Delivery Network where users can spin up containers globally in seconds with a few lines of code. When content becomes popular in different regions worldwide, the orchestration platform provisions VMs, applications, and edge databases to service the requests or demand.

In addition, dedicated infrastructure and a global network ensure latency is minimal. Hosting applications at the edge allows Fly to serve dynamic content to visitors without invoking centralized cloud services like AWS.

Fly.io is a fresh breath of air in the CDN industry. We hope to see more innovation in their edge database technology in the coming quarters.

We’ll check back with them next year.

Leave a Comment