How AppsFlyer cut compute costs by 30% with no code changes

Download Case Study

25%

Spark cost reduction

35%

Kafka cost reduction

30%

Annual cost reduction

AppsFlyer turned to Granulate's real-time continuous optimization solution to support the company's fast growth and increased demand for their AWS infrastructure, accompanied by rising costs. Following Granulate's activation, AppsFlyer immediately achieved a 30% CPU utilization reduction in Spark-Hadoop cluster, 25% CPU utilization reduction in Kafka consumers cluster, leading to an overall 30% cost reduction.

About

AppsFlyer, a global attribution leader empowers marketers with a suite of measurement and analytics solutions. AppsFlyer's platform helps 12,000+ brands and 6000+ technology partners make better decisions daily.

With Granulate, we were able to save hundreds of thousands of dollars on two of our clusters without any code changes

Eli Zilbershtein, Platform Group Team Leader

The Challenge

As the market leader of the growing mobile attribution market, AppsFlyer heavily relies on AWS usage, and with the rapid growth also comes an increased demand for their AWS infrastructure and rising compute costs. AppsFlyer’s highly professional team was looking for an innovative complementary solution to the already implemented best-practices and cost-management solutions

The Results

Following Granulate’s activation, AppsFlyer saw immediate performance improvements.

Monitoring performance in their Grafana showed a 30% CPU utilization reduction in the Spark-Hadoop cluster along with reduced job completion time. In addition, they witnessed similar results in their Kafka workers nodes, with 25% CPU utilization reduction leading to an overall 30% cost reduction. These results enable AppsFlyer to reduce the number of machines required for these services, amounting to an annual cost reduction of $320,000.

The following graph shows the 30% Spark CPU utilization reduction following Granulate and the CPU utilization following the cluster size reduction.