Effectively deriving insights from big data can be a huge challenge for today's organizations. Few companies have the resources to collect and analyze large volumes of data in real-time. Sovrn set out to change that dynamic for publishers and content creators.
As one of the largest publisher networks in the world, Sovrn collects, combines, and analyzes an immense amount of data from across its network. By aggregating this data and sharing it through its meridian platform, Sovrn provides smaller publishers with advanced insights, regardless of the amount of site traffic.
Challenge: Modernizing Sovrn's Infrastructure Capabilities
On any given day, Sovrn scales up and down an order of magnitude anywhere from four to 10X, depending upon the amount of traffic on the ad exchange.
According to Sovrn's VP of Platform Engineering, Kyle Gilliland, it takes a lot of horsepower and compute power to process big data effectively and efficiently. "We build machine learning -- or artificial intelligence -- algorithms to manipulate all the data we collect. Scalable technologies that allow us to do that fast and efficiently are crucial to our business."
However, to scale up its infrastructure to meet ever-changing levels of demand, Sovrn was manually acquiring and provisioning systems – a process that took weeks. As a result, the business was not able to move as quickly as it wanted, nor ensure necessary infrastructure resiliency.
Solution: Mesosphere Eases the Transition to Containers
Gilliland knew that driving an as-a-service-based software architecture would allow Sovrn to isolate services in containers and easily scale those services. However, though he wanted to get into containers, Gilliland knew Sovrn couldn't afford to jump into the complex world of Kubernetes. Moreover, he didn't want to Sovrn to develop its own open source capabilities using other open source tools.
As far as Gilliland was concerned, Mesosphere represented the perfect marriage of the two, removing barriers to entry. "We saw that Mesosphere could help orchestrate our workloads, enabling us to handle this efficiently in any location," says Gilliland.
While Sovrn had successfully been able to deploy tools like Kafka and Cassandra, Mesosphere enabled it to do so with a push-button capability. "Mesosphere enabled our developers to do this easily and to automate that whole structure end-to-end, without any interruption of configuration or management in between."
Currently Sovrn is running 18 applications in a production state across its Mesosphere cluster, including Java applications, Spark, and Apache, and is looking at moving ephemeral workloads to Mesosphere.
Impact: Running Dynamic Workloads in Production
Using customized open source technologies that specialize in delivering real-time data processing, Sovrn can enable analysis of nearly any event within 20 seconds of its occurrence. Plus, by scaling up new infrastructure within days or hours, it can more quickly and reliably deliver this capability.
Mesosphere has given Sovrn the opportunity to put in place repeatable automated processes. As a result, Sovrn has matured through its container journey. Now it builds real-time data processing pipelines with continuous integration in mind from the start. And it deploys applications seamlessly in the same fashion from Development to QA to production.
"It's a build-once/deploy-anywhere construct that has helped evolve our engineering organization. The push-button capabilities of Mesosphere enable my engineers to develop and create more value," says Gilliland.
Read the full Sovrn customer story by downloading the case study.