1 Comment

Summary:

Cloud-based application performance and security specialist CloudFlare finally announced on Tuesday that it closed a $50 million funding round in December 2012, but the real story is about the company’s rapid growth and how it’s rethinking its infrastructure to support all that traffic.

128-gb-of-ram
photo: CloudFlare

What do you do when you’re a web company with 1.5 million users around the globe that pass about 350 billion page views across your platform each year? If you’re website security and performance startup CloudFlare, you quietly raise $50 million in venture capital, sock it away somewhere and tell the world a year later. You also start growing your platform like crazy and tinkering with the design of your servers and switches in order to handle all that traffic.

The official news from CloudFlare on Tuesday is that it revealed the $50 million it raised from Union Square Ventures a year ago and is growing like mad. Those 1.5 million customers are twice as many as had it last year — a 100 percent increase that led to a 450 percent increase in annual revenue. But while the numbers tell part of the CloudFlare’s story, an equally important part is how the company is managing all this growth.

“We kind of operate at that internet-giant-like scale, but we’re still a relatively small company,” CloudFlare Co-founder and CEO Matthew Prince said. (For more information on the company and Prince, check out this profile Gigaom’s David Meyer wrote as part of our Cloud Trailblazers package in June, or watch his brief Structure presentation below.)

Global expansion and custom gear

However, it turns out that 70-person CloudFlare is managing growth in very much the same way that umpteen-thousand-person companies like Facebook, Google and Amazon do: It’s expanding its data center footprint and designing its own gear. Already in 23 data centers globally (of course, it rents cages in those facilities instead of building them), Prince told me the company is budgeting to add more than 100 new locations in 2014.

“There are not a lot of companies that this is a rational decision for,” he acknowledged, but CloudFlare has serious performance demands already — handling about 5 percent of all web requests — and has “a legitimate shot to be one of the largest” edge-service providers around.

So early last year, CloudFlare began working with server manufacturer Quanta to design boxes that can handle its traffic and lower costs as the company scales. That meant switching from 1 Gbps network ports to 10 Gbps ports, tripling RAM capacity and generally doing everything it can (including using platinum power supplies) to ensure performance and reliability. CloudFlare detailed its latest-generation servers in a July blog post (although it has since replaced some Intel SSDs with Fusion-io cards for better performance, Prince said).

A couple things that didn’t make it into that July post: CloudFlare is testing out ARM-based server processors as a way to save energy, and because the company’s old Juniper network switches began becoming a bottleneck during the summer, Prince said CloudFlare has also been working with a company called Pluribus Systems to design its own switches.

“This is a business that has just incredible scale economies,” he added, referencing the ability to turn all computing his company is putting at the network edge into products it can deploy as a service.

complete-g4-hardware

Source: CloudFlare

Getting the most out of gear with AWS breathing down your neck

Locations that used to be filled with specialty appliances from vendors like Riverbed, Imperva and F5 are now filled with general-purpose (albeit custom-built) gear that can perform a variety of tasks. For example, there are times of day when CloudFlare knows certain locations will be quiet traffic-wise, and Prince suggested it could use the excess capacity to power services like malware or vulnerability scanning — really anything that’s CPU-intensive and can run remotely — that would complement its current firewall and network-management services.

Growing its suite of services might be necessary as CloudFlare increasingly finds itself squaring off against web and cloud computing giants like Amazon Web Services, Akamai and Google. Right now, Prince claims CloudFlare already provides DNS management for more companies than does AWS, and that CloudFlare’s CDN service is growing at a faster pace than is Amazon CloudFront. Large hosting providers such as Rackspace and Media Temple already use CloudFlare to expand their edge capabilities so they can better compete with Amazon Web Services, he added.

However, Prince sees AWS becoming a serious competitor in the long run as it tries to own even more of the stack from basic compute and storage all the way out to the network edge. CDN kingpin Akamai already has a large edge presence and is getting serious about security, he acknowledged, and Google has the means to match CloudFlare in terms of scale and products should it choose to do so.

So maybe that $50 million (most of which is still in the bank, Prince said) is a big deal after all. It’s a drop in the bucket to CloudFlare’s potential competitors, but it could be enough to help CloudFlare beef up its capabilities and reach critical mass before they decide to get into its game more deeply. Even if CloudFlare can’t beat them in the long run as an independent company, it can still command a high price tag when one of those giants decides it wants to buy its way into CloudFlare’s space.

You’re subscribed! If you like, you can update your settings

  1. It’s excellent to see CloudFlare doing so well, and at Pluribus Networks we’re thrilled to be part of their next generation infrastructure that drives their business model and customer offerings.

Comments have been disabled for this post