SundaySky is transforming customer experience for the world’s most demanding brands, enabling them to deliver video-powered experiences that drive breakthrough outcomes at key moments across customer journeys.
The platform and domain expertise equip brands to achieve business results and quantifiable value from increased revenue, reduced costs, lower churn and higher customer satisfaction.
As a leading video advertising platform, SundaySky serves personalized video experiences to millions of customers daily, so it is built with an emphasis on performance. The platform is hosted on AWS EC2 Graviton-based instances.
SundaySky wanted to reduce their growing AWS bill while also preserving performance and adhering to their strict SLA requirements. The company has undergone many cost-saving initiatives in the past, including moving most of its compute workloads to spot instances, and internal initiatives to identify and eliminate the main cost drivers.
The Intel Granulate optimization solution was deployed on the bidder workload with a single command-line on two different regions. The service was bound by CPU with autoscaling metrics set on bid time, which is equivalent to latency. The solution was deployed using a ‘canary deployment’, meaning half of each cluster had the Intel Granulate installed while the other half didn’t, which allowed for easy comparison during the benchmarking phase.
Immediately after activation, SundaySky witnessed a reduction in the bid time, followed by a decrease in the number of machines in the clusters, triggered by hitting the low threshold of the autoscaling metric. This reduction in the size of both clusters was translated into a 15% reduction in AWS spend.
Figure 1: Intel Granulate reduction in the size of the East cluster activation
Green without Intel Granulate, orange with Intel Granulate
Figure 2: Intel Granulate reduction in the size of the West cluster activation
Blue without Intel Granulate, orange with Intel Granulate
SundaySky turned to Intel Granulate's real-time continuous optimization solution to slash costs on its most costly service, the bidder, which places bids on its videos in real-time. Reducing the AWS bill on this service meant significantly improving the company's margins. This service is very performance-sensitive and mission-critical, meaning degradation in performance was not an option.
