Building Verizon Digital Media Services’ Fourth-Generation Edgecast Content Delivery Network
Verizon Digital Media Services is constantly updating our platform’s system and infrastructure to provide our customers with the fastest, most secure, most reliable and best possible quality, delivery and performance.
The first generation of our Edgecast Content Delivery Network was built in 2007, but our rapid growth meant we had outgrown after only a year. At the time, we experienced a large influx of customers coming onboard who catered to many customers themselves, which meant more traffic. Eventually, all this became too taxing for our infrastructure. Fast forward to 2016, during the time of our third-generation network, and we were seeing the exact same problems, only at a larger scale.
We had reached the physical limits of the third-generation network, which was based on Juniper MX960 routers and QFX5100 switches. On our busiest PoPs (points of presence) we were unable to add capacity. We had used every slot in the router and every available port on the line cards. Additionally, there were bandwidth limitations on the uplinks between the Top of Rack (TOR) switches that prevented us from making use of the entire throughput.
The increase in demand showed no signs of slowing, so we set out to create a new network design that was prepared to handle all this traffic and more. We launched a program to figure out exactly what this next generation would be. We wanted a design that could scale horizontally, essentially creating a flatter architecture. What we needed most was to provide quality service, performance and delivery. That way, as more and more customers came onboard with greater bandwidth we would be able to handle it and always keep pace with the growth in demand.
With the plan set, it was now time to take action. We cut the router into border and core layers, giving us the ability to process routes and not have to worry about the ones that were not in use. The core PoPs in this layout would no longer have top-of-rack switches; every server would be directly connected to the core device. This would allow us to take full advantage of the bandwidth of each server. Every core would be connected to each of the borders so that if one border happened to go down, each PoP would continue transmitting and delivering traffic.
This new architecture solved many of our challenges and set us up to confidently move forward and take on new customers. It allowed us to expand our throughput by a factor of 10, plus it added redundancy and fault tolerance to the design, along with easier management and the ability to perform maintenance without impacting customer traffic in any way.
With our new fourth-generation network architecture, core PoPs do half the amount of work, but with even greater capacity. The new layout also has redundant borders, which means greater fault tolerance. Traffic can now be separated into different PoPs based on type.
Just how powerful is our next-generation network? Each node is equal in size to the entire capacity of some of the smaller content delivery networks (CDNs) on the market today. Overall, our new design is high performing, more streamlined and has greater throughput and quicker delivery.
Want to learn more about our next-generation architecture? Get in touch with us today.