By Kyle Okamoto, Chief Network Officer
Within the past few years, the number of people who regularly consume high-quality mobile content, such as high-definition video, has grown at an astronomical rate. According to Cisco, 82 percent of all internet traffic will be video by 2021. Most of that international growth has come from "mobile-first" developing nations as their internet infrastructure improves to support these applications. According to the UN International Telecommunications Union (ITU), mobile broadband subscriptions are growing by more than 30 percent year over year in developing nations, compared to around 10 percent in developed ones.
For content delivery providers, this international growth poses a difficult question: How can you prepare your network for this upswing in demand?
At Verizon Digital Media Services (VDMS), we've tackled this question directly through developing our next-generation content delivery network (CDN). Here are the biggest takeaways we have deployed against our CDN strategy:
The early days of our Edgecast CDN
When Edgecast was built in 2007, it was one of the first CDNs optimized for video. Instead of focusing on lightweight assets like text and images of the HTML web pages of the past, Edgecast's software was designed to cache large objects (like video chunks) for long periods of time—so, for instance, it could deliver a viral video over and over again from a given point of presence (PoP) without needing to ping the origin server over and over again. The basic network architecture of the original Edgecast CDN served us well for many years, but by 2016, it became clear we needed to retool to meet rapidly rising demand. VDMS could see that massive international growth was on the horizon, so we took advantage of the moment to upgrade our entire network from the ground up.
Upgrading our content delivery network to run faster and safer
When we began designing our next-generation CDN, we took a border-core (or leaf-spine) approach towards developing the architecture. The structure is similar to that of a tree whose trunk splits off into multiple branches, each of which holds leaves. If a single leaf falls off the tree, the series of leaves and branches around this spot ensure that the plant can survive and thrive.
This approach carried over to how we laid out our next-generation content delivery network to handle the new demands of video-heavy users. We had realized that due to massive international growth in traffic, the failure of a PoP in a large city like New York or London could no longer be solved by rerouting to another PoP without causing service degradation, which is not okay at VDMS. That meant we needed to make the PoPs themselves more resilient—and an easy way to do that was adding redundancy in the network layer.
Redesigning for redundancy
We already had redundancy in every other part of the CDN. Our CDN is inherently redundant in the software stack and the application layer; we also used multiple storage vendors, data centers and redundant power supplies. So, within individual servers, there would be multiple drives and other key components that would provide redundancy and prevent service degradation or outages.
In the old days, all of these servers were typically tied to a single router. If this router failed or needed maintenance performed, the servers would need to be taken down, causing its share of multiplying issues to troubleshoot. By introducing redundancy at the network layer, we were able to find a solution that improved our reliability and scalability. Now, our team can easily work on a set of servers or an individual router without having to worry about triggering any potential service issues.
During the redesign process, we also made performance increases a focus. On a technical level, we wanted to ensure that our CDN was never a performance bottleneck. By removing some hardware and replacing other pieces with higher-performing components, we were able to increase throughput while also simplifying the network structure, leading to greater reliability since there were fewer single points of failure in the system. Through this streamlining process, we can now guarantee that our next-generation content delivery network can handle the expected load from a video-heavy user base at scale.
Emerging markets point towards the future of streaming video, CDNs
The rise in streaming video demand worldwide means that CDNs must take a global perspective on international growth. Users now expect to watch high-quality online video regardless of where they are physically located. That's why it has been a priority for us in recent years to build out our capacity in emerging markets like Latin America, the Middle East, Africa and Southeast Asia.
Having the next-generation content delivery network in place has helped enormously with this expansion. In markets with developing, last-mile infrastructure, issues such as power outages and fiber cuts are common. But thanks to our next-generation content delivery network's redundancies, we're able to absorb the hits from these outages more smoothly than we would have been able to otherwise. Our upgrade enables us to meet high standards of performance and reliability around the globe, not just in established markets—and that alone has made the investment in international growth worth it.