Defining innovation capacity, part 2: Flexibility capacity

This is the second post in a series dedicated to discussing how companies driven by innovation evaluate technology stack expenditures and the cloud-based services used to deliver their products and services. Read part 1: “The innovation value of your technology stack.”

No one sets out to design for inflexibility. However, many organizations have discovered delay is built into their IT systems just as they were about to deploy new capabilities. Spinning up a virtual machine (VM) could take minutes or hours. Tying it into your development environment takes even more time. Integrating it with the rest of your infrastructure is complex and daunting.

I believe that inflexible technology can promote an aversion to innovation. When building a stack, cost and performance are traditionally prioritized. That makes sense for immediate needs, but companies are finding that the flexibility of that stack matters in the long run.

Reduce deployment steps

The number of steps to deployment will determine flexibility. The fewer steps to a test or deployment, the quicker your team will be. We think about that when we design Fastly services. While customer support and consultation is always available, providing complete self-service is a huge point of emphasis for Fastly. We created the ability for our customers to implement Custom Varnish Configuration Language (VCL), enabling them to code their own specific caching or other behavior needed to tune their service.

country edition

VCL code that selects an edition by country of requesting IP.

Shopify is a Fastly customer that constantly drives for more and better. Their technical staff is working to provide their 160,000 online store owners with a fast and reliable ecommerce solution. To do that, they need to test and improve their architecture. With their previous CDN, it would take up to two days to implement a change.

Now, with access to the Fastly Dashboard and updates that propagate in seconds, Shopify can deploy and test configurations faster, in real time. This means they can deploy a new config to global clients in two seconds instead of two days — making the phrase “faster and more cost effective” a major understatement. Fastly lets them know when configurations and content have propagated as well. They don’t need to guess if the new config is in place or that the content has been updated. Check, know, and move on.

"We really like the fact that we can write our own VCL and fully control the experience on Fastly. If we need to deploy a hot-fix or revert a change, we have the power and control to fix it ourselves. That's huge for us. With our former provider, we always had to engage our assigned customer support engineer to have changes made."
— Dale Neufeld, Director of Technical Operations, Shopify

Scaling without resource worries

The possibility of collateral damage can crimp an innovator’s style. When you scale a new project in a data center or with a traditional colocation, you must first confirm if the computer, network, and storage resources are in place. Does the new service threaten the resources of existing applications? If so, in the short term, the result could be service degradation. Long term, you have to fill out purchase orders to acquire new resources. Another option is a CDN that means new services and content can be added without significantly increasing compute.

By employing Fastly and Google Cloud Platform, you can also reduce the cost of running your application.

“Once we get one of our endpoints on Fastly, we see a dramatic reduction in the traffic hitting our servers — sometimes over 80% — which has allowed us to cut out additional servers.”

— Harlow Ward, former Rails Lead, Hotel Tonight

By using the additional capabilities of Fastly, our customer Hotel Tonight improved their cache hit rate from 75% to 90%. That is what you see on the Fastly Dashboard, but the flexibility they realized was at the origin: a five-fold efficiency increase of origin resources. This means that if they had been running on 16 instances, they could have cut it down to four.

The line-item savings is interesting, but the reduction in complexity is really the compelling story. An innovation-ready technology stack can use tools like a smart CDN, containerized ops, and quick deployment computing to be ready without having to implement complex scaling plans.

Essentially, existing services no longer hold innovation hostage if the stack is able to flexibly expand capacity. Of course, the pooling and pricing structures of cloud service technology stacks have reduced this problem across the board. We can get even more flexible, though.

Fastly can take the scaling question off the table. Shopify saw the benefit when a Super Bowl ad created an 18,000% spike in traffic to their client GoldieBlox. Fastly handled the load and protected their origin from getting swamped. Because that is what we do: we provide the flexibility for when opportunity presents itself.

Innovation is not just the domain of new code. Content changes need to be responsive as well. When live blogging for an Apple product release, Gizmodo peaked at 55,841 RPS. As a Fastly customer using Origin Shield and caching event-driven content, only 10% of those requests went to the origin. It is a very different server architecture to handle 55,000 RPS than 5,500.

A lot of our customers are able to serve huge amounts of end users in demanding environments with surprisingly small server profiles. For some applications, a single Google Compute Engine region can serve a global reach.

For example, Fastly customer Upworthy saw that they needed to provide better service for mobile users. By implementing Fastly, they were able to do so without amping up the compute at their origin.

"Upworthy had a very large increase in mobile traffic, and Fastly helped us reduce the load on our origin. We use Heroku, and once we moved to Fastly, we cut our dyno usage in half, reducing infrastructure costs."

— Ryan Resella, Senior Engineer, Upworthy

"Fastly has allowed us to improve performance overall, and we’re actually running a much leaner setup. Now, our latency is really good. Before Fastly, our miss latency was terrible; it could be anywhere from 5-10 seconds. Because Fastly gives us real-time insight into our site’s performance, we were able to make the necessary changes to our infrastructure. Now our miss latency is well below 1 second."

— Pavel Repin, Senior Engineer, Upworthy

Another kind of inflexibility can come about with cloud platforms — the customization to code and infrastructure required for the platform can bind you to a single vendor. Sometimes these customizations have great value, but that does not mean they aren’t a potential business problem. Google Cloud Container Engine continues to develop and employ Kubernetes for Docker containers. Although still in Alpha, the vision is about long-term flexibility. You can start your project on a cloud service but if you become wildly successful and need to adopt a hybrid cloud model, your Docker containers can function in private cloud and Google Cloud.

Don’t work without a net

You require a certain amount of bravery to innovate, but it doesn’t need to be risky. You don’t want fear to prevent experimentation.

When a change doesn’t work like you’ve planned, Fastly and Google Cloud Platform can both push changes very quickly. If you are not happy with the results of a new VCL script or policy, you can roll back to the previous configuration. Just click a button, and in less than five seconds the settings will propagate to the client experience.

When it comes to content, Fastly’s Instant Purge gives you the ability to change in seconds at any time, on any scale. Our customers prioritize innovative content, but use surrogate keys to most efficiently control the content their users view.

Along with price and performance, flexibility

When you adopt Fastly, the immediate benefit is reduced cost and increased performance. There is a long-term value, though, one that we focus on maximizing. Fastly will enable the innovative next service. By reducing cache misses with Fastly’s Origin Shield and real-time caching of all HTTP types, you can scale a service or add a new one without having a conversation about infrastructure.

The innovations and evolutions that I get to talk about come from companies that were able to prioritize creative thinking and experimentation. A CDN is usually thought of as a cost savings or performance-improving tool. It is, and Fastly does all that extremely well. But our customers also realize a greater flexibility with their deployment, infrastructure, and content. That flexibility means that when the new idea presents itself, they have the willingness to get started.

Fastly is joining Google Cloud Platform on their Next tour to share more about how we help companies innovate. Keep an eye out for us in your city.

This post is the second in our series. In the next post, we will talk about how the design of a technology stack can make innovation more informed.

Lee Chen
VP, Corporate Development and Strategic Partnerships

6 min read

Want to continue the conversation?
Schedule time with an expert
Share this post
Lee Chen
VP, Corporate Development and Strategic Partnerships

Lee Chen is the Vice President of Corporate Development and Strategic Partnerships at Fastly and sits as the executive sponsor for our media products. At Fastly, he has led a wide range of functions across product, marketing, and partnerships. He has spent the past 20 years in the media and entertainment space, pioneering live broadcasts over the internet. Prior to joining Fastly, he founded and led several startups in technology and eSports.

Ready to get started?

Get in touch or create an account.