NBCUniversal Uses Kubernetes and Diamanti to Reduce Auto-Scaling Time from 15 Minutes to Seconds
Jan 18, 2017 / in Main Blog / by Mark Balch
In a recent webinar, NBCUniversal’s Vice President of Engineering Operations, Derek Yovine, Google Kubernetes Engine product manager, Dan Paik, and Gopal Sharma, Diamanti’s Chief Technology Officer discussed how to deploy data-intensive, containerized applications with guaranteed performance.
With Diamanti, Kubernetes and containers, NBCUniversal was able to accelerate container time-to-market to seconds and provide guaranteed real-time container service levels
NBCUniversal partnered with Diamanti on an identity application that underlies many of its properties, including cnbc.com. NBCUniversal was able to reduce the cluster from 12 nodes down to 4 nodes and increased performance with the reduced infrastructure.
NBCUniversal deployed containers using Diamanti and Kubernetes to get faster application scaling, particularly with auto-scaling. With their previous architecture, it took 15 minutes to reach the auto-scaling threshold, code deployment, and for the application to spin up and accept connections. While the previous architecture was great initially, over time they wanted more.
The new container architecture with Kubernetes and Diamanti reduced auto-scaling time from minutes to seconds and included compute, network, and storage.
Identity application latencies were reduced by 20-40%, which means tremendous performance improvements with reduced infrastructure. Even after reducing hardware infrastructure with Diamanti, the infrastructure at NBCUniversal is still only 30% utilized, giving 3x the headroom for future scaling. Tests show that Diamanti can consolidate containers with 90% utilization
Moving to a containerized architecture is not easy. Diamanti was able to fast-track their deployment.