How Granulate helps Singular reduce cost by 35% on mission-critical services on Amazon ECS

Download Case Study

43%

Average Job Completion Time Reduction

55%

Attribution Service Reduced Average Latency

30%

Data Pipeline Service Reduced Costs

Singular turned to Granulate's real-time continuous optimization solution to improve the overall latency of its application, but also to reduce its growing AWS bill. Singular deployed Granulate’s agents in passive mode on several of its platform’s core workloads to begin. After the learning phase, once Granulate’s agent was activated, Singular saw an immediate improvement of 55% in response time and a 34% reduction in CPU utilization on that service.

About

Singular is a marketing intelligence platform that unifies marketing analytics, giving marketers actionable insights from previously siloed data. By connecting upper funnel marketing data with lower-funnel attribution data, marketers can measure ROI from every touchpoint across multiple channels and optimize spend down to the most granular level. Singular customers include Lyft, Rovio, Warby Parker, Airbnb, Wish, Zynga, Netmarble, and other Fortune 500 brands.

We were able to achieve great cost-saving and improved responsiveness on some of our most mission-critical services all that within a matter of days

Ofir Nir, Head of Devops

The Challenge

In the past few years, Singular has experienced tremendous growth. As it scaled up, the company started looking for initiatives to optimize its cloud infrastructure in terms of cost and performance. Singular’s platform is hosted on a containerized environment on Amazon ECS, with spot-based instances across several regions to provide a great user experience to its global customer base. Singular was searching for a way to improve the response time of some mission-critical services on Amazon ECS, to improve the user experience of its services, and to allow for scaling infrastructure without jeopardizing its guaranteed SLA. Specifically, Singular was looking for a way to improve its application performance by focusing on the median latency to improve responsiveness, and the p95 latency to support traffic spikes.

The Results

Singular deployed Granulate’s agents in passive mode on several of its platform’s core workloads on Amazon ECS to begin. After the learning phase, once Granulate’s agent was activated, Singular saw an immediate improvement of 55% in response time and a 34% reduction in CPU utilization on that service. These quick results, which came only days after starting the POC, gave Singular the confidence to activate Granulate’s agents on the entire cluster. Additionally, the team activated agents on three more core services. The company witnessed a reduction in job completion time of 43% on Data Pipeline service- a big data cluster that is tasked with performing calculations throughout Singular’s platform. The performance improvements enabled them to significantly reduce the number of instances provisioned for all four services, resulting in over 30% reduction in cloud spend.