How to Decide Where a Rapidly Expanding Global CDN Should Go Next
By Kyle Okamoto, Chief Network Officer
As internet access expands across the world at a breathtaking rate, it’s a challenge for any content delivery network (CDN) to know where to build out next. In order to deliver high-quality media experiences without delays or other setbacks, it’s imperative to have infrastructure in place before a newly wired region starts seeing spikes in traffic.
But how can a CDN anticipate where those regions will be?
How can it foresee the demand far enough in the future to meet it, while staying flexible enough in the short-term? And once the CDN is being used in a particular region, how can it grow alongside that ever-changing market? Here’s how I approach these important questions of CDN strategy:
Build a crystal ball to see the future
Ultimately, the most important factor in deciding where to build out comes from existing customers. At Verizon Digital Media Services, we’ve built a crystal ball – a tool that effectively allows us to see the future, at least as it pertains to our customers’ needs. This tool (or system of tools) ingests a large amount of data, both structured and unstructured, and allows us to trend and forecast impacts to the network with various sensitivities to accommodate the ever-changing landscape.
We use this tool with our Edgecast Content Delivery Network to predict organic and inorganic traffic growth based off factors, including: broadband and wireless penetration driving consumer demand, reachability from existing infrastructure, current quality and performance levels against targets, the composition of our current customers, expected new customers, how our customers are migrating to higher bitrates or new geographies and the costs to extend reach across different networks and different geographies at various capacity levels.
It isn’t always perfect science, and a lot of human intuition helps to make the crystal ball clearer.
When applying insights from this forecasting system, it’s important to also be sensitive to each individual customer’s needs. We care most about the users; they drive everything we do at our core. We love to build platforms that our customers love, and that they can use to make their users happy. Say we have a live event provider that delivers popular fútbol matches in Latin America – they need a platform that can handle a large temporary surge in traffic, which is very different than 24 x 7 linear channels traveling across our network at all times. If our crystal ball tells us that Latin America will see a spike in internet usage that may exceed an acceptable capacity threshold, we’ll plan for that and overbuild to ensure we have more than enough capacity for our customers to be successful. There are also different platforms: a storage platform versus an eCommerce platform or an application acceleration platform. That’s why these predictions are based on a collection of tools rather than just one, giving us a multi-dimensional view into our network and capacity planning across unique platforms.
Even within a particular country, we need to know where specifically to build out our CDN infrastructure. For that, we rely on our relationships with networks and data center providers in those countries. If we were looking to expand in Chile, for example, we’d reach out to networks like Telefónica or VTR or Entel as well as smaller internet service providers, and they can tell us what the topology of their respective networks look like and how that will affect our site selection. For example, it is cheaper and easier to lay submarine cable off the coast of Chile instead of cable through the mountains. That’s the kind of information local internet service providers can share to help us decide where to go, as well as how big to go once we get there.
Plan ahead, but not too far ahead
In any business so closely tied to technology, the challenge is trying to predict the future while staying nimble enough to respond to the needs of the present. Verizon Digital Media Services works with a short-term roadmap extending to four months, a medium-term roadmap extending to 12 months and a long-term roadmap that can extend out to 18 months, sometimes more. All decisions are made based on data coming in from our many analytical tools, and depend on cooperation from multiple teams including Enterprise Resource Planning, Traffic Management, Project Management, Operations, Engineering and more. We aim to maximize performance for our customers above all else, so we want to take advantage of the best processors (which evolve quickly), the best storage, the best of everything. We have to balance this with operational churn and return on investment, but performance always wins and fuels our business.
This month, for example, we launched the newest addition to our network in Querétaro, Mexico. We had it on our medium-term roadmap as we worked through site selection and infrastructure sizing, and then moved it to a short-term one in January 2017, with a commitment to complete by May. We had to anticipate issues related to imports and exports; everything takes longer, as shipments need to be carefully inventoried in the U.S. and then double-checked when they arrive in Mexico.
But we did it. We built the PoP on schedule as promised to our customers and our partners. And that one build will last us for years without infrastructure augmentations as we continually update the home-grown software that supports our CDN to maximize performance.
That’s the difference roadmap thinking makes. Instead of focusing solely on the quick return on investment this quarter as we used to have to do as a startup, we can make bigger plans to increase our presence in an area for a longer period of time. The significant investments made by Verizon in this space also allow us to go big and plan for the future.
Expect the unexpected
Besides driving decisions about where to go, our customers also help us make decisions about where and when to expand once we’re there.
In India, for example, we initially put in four different PoPs at one time. As with the PoP we just installed in Mexico, we anticipated that initial infrastructure would last for a couple years. But, demand has far exceeded expectations. We also have a strategic relationship with Bharti Airtel, one of the major telecommunications providers in the country, who is aggressively building out its own 3G and 4G mobile infrastructure. So 14 months after we initially expanded our network in India, we’re already augmenting our infrastructure to make sure we have enough storage and network capacity to deal with this accelerated growth – a good problem to have. These consumers may be experiencing an amazing wireless connection for the first time in their lives, and it’s important to us that we’re delivering the highest quality content from our customers to fulfill our end of the bargain.
Deciding where our CDN should expand next is both art and science. It takes a carefully constructed forecasting system, deep consideration of feedback from customers and a healthy dose of human intuition with an eye on the consumer’s needs. And of course, it requires an immense amount of investment – money, resources, human capital and time. In the past few years, we’ve used this balanced decision-making process to drive the broadest and fastest expansion of our Edgecast CDN infrastructure in its history – and we look forward to continuing to do so for many more years.