Section: Giving Developers Control of Their Edge Computing
Application developers struggling with content delivery networks: Daniel Bartholomew feels your pain.
Six years ago, he and Stewart McGrath were working together at Australia’s largest e-commerce company, and they found themselves deeply frustrated with how focused traditional CDNs were on infrastructure problems and not on the needs of developers. Driven by a vision of empowering developers to get better control over their CDNs, the two men co-founded Section.
“As we’ve evolved our technology stack and our business, the industry has also been evolving to properly define the vision that Stewart and I had at the beginning, which in today’s language is called edge computing,” says Bartholomew, who serves as CTO to McGrath’s CEO. “Edge computing goes beyond what the traditional CDN does by enabling developers to push their code away from the mega-scale clouds like AWS and Google and move a lot of their programming logic out closer to where the consumers are.”
An Innovative Product
Section offers a multipurpose edge computing platform that allows web-based businesses to run their HTTP traffic through it. As with a traditional CDN, the platform sits between the end user and the business’s infrastructure (which typically runs in a large public cloud like AWS), and intercepts traffic.
However, unlike a CDN, which mandates what software a company can run in its data centers, Section allows developers to choose from a variety of modules that are tailored for specific use cases. “Then we open up a programming framework, which allows them to write their own code and deploy that out into this edge network,” says Bartholomew. “This means that software developers and architects can restructure their application so that part of it is running in Amazon, and part of it is running in the Section network. By getting that flexibility, they can build a much more efficient solution.”
(Section is based in Boulder, Colorado with operations in Sydney, Australia.)
A Packet Prototype
In 2017, Section began working on a prototype using Packet. “What attracted me to Packet initially was their high performance servers and the fact that we could get access to dedicated hardware,” says Bartholomew. “Packet allows us to treat dedicated hardware as though it was a virtualized cloud.”
Plus, “Packet’s support for Border Gateway Protocol (BGP) sessions is useful in an edge computing framework,” he adds. “Packet also allows us to use and control our own IP space, which has a lot of benefits.”
For 18 months this work remained a prototype without any workloads in production. However, by mid-2018, Section’s customer base had grown to a point at which the company required a higher degree of control and performance. “With our growth accelerating, we really needed to get into a more specialized relationship with our infrastructure partners,” says Bartholomew.
The Hotjar Bake-Off
The tipping point came when Section attracted the attention of Hotjar, a company that provides customer experience software to more than 300,000 websites. Hotjar was having issues with its CDN, and flirted with leaving until that CDN promised a network upgrade. That upgrade happened, and still “they saw no benefit to their system whatsoever,” says Bartholomew. “Within a week, they reached back out to us.”
Not only did they want support for their legacy CDN workload, but they were also interested in an overhaul of how their application is built, so that the system could scale more efficiently and they could reduce the costs of their hosting infrastructure.
Section’s proposed solution was to move some of Hotjar’s application logic out of Amazon and into the Section platform. “In order to win that deal, we had to match or beat the performance of their incumbent provider, but not only that, they actually put us into a bake-off,” says Bartholomew.
Using New Relic, Hotjar conducted benchmarking against four household-name CDNs and Section (running on Packet’s equipment). “Not only did we outperform all those big names, but the performance that we delivered was much more consistent,” he says. “Consistency in the performance is really important to a company like Hotjar because predictability means that operations and things like capacity planning are much easier to perform.”
Hotjar alone accounts for 2.56 billion requests and 43.5TB per 24 hours on Packet. And from that point on, says Bartholomew, “Section was really going all in on Packet.”
The Packet Difference
Everything Section runs in production today fits into default models that Packet provides, which is a big plus for the company. “As Packet expands, their capacity becomes available to us, and we don’t need to do any kind of custom orders or specialized offerings,” says Bartholomew.
Additionally, the Section team appreciates how Packet has made senior engineering resources available for brainstorming and prototyping. “I shared with them what our goals are, and they brought me working prototypes of how our Kubernetes-based edge delivery platform would work really well inside Packet, and still facilitate our model of allowing multiple vendors to participate in our solution,” says Bartholomew. “It’s been great getting access to the engineering team who really understand what we’re trying to achieve, but also are not steering us down a vendor lock-in path.”
The Present and Future
As Section’s customers are becoming more familiar with edge computing, expectations have started to shift. “One of the most delightful things that happens is when developers realize that Section is so much more than the traditional CDN that they’re used to, and they start building logic into our platform that we haven’t seen before,” says Bartholomew. “This is one of the beautiful things about edge computing.”
The gaming industry, for instance, has been gravitating to the Section platform for the promise it holds. “Computer game developers are launching large infrastructures inside high-scale clouds to collect data about how a gamer moves through a map or how much corn they’ve farmed,” he says. “Now these developers realize they can improve the efficiency by using an edge computing platform. Instead of scaling up Amazon so large to meet the surging demand for that game, they can collect this data at the edge and then just trickle back the small amount of data that they need to Amazon. What happens then is gamers get a better in-game experience and the company running the game can spend less money on the centralized infrastructure.”
As new use cases materialize, Section is determined to continue innovating. “Our developer capabilities are unmatched in the industry,” Bartholomew says proudly, “and we’re about to double down on that and introduce a lot of modern principles coming out of organizations like the Cloud Native Computing Foundation. We’re interested in moving our core diagnostics platform away from logging to tracing, and expanding our library of modules.”
Whatever path the Section platform ends up taking, Bartholomew vows, one thing will never change: “We think about the developer first in our engineering decisions.”