Beat Saves on Costs and Improves Database Performance Using AWS Graviton2 Processors in Amazon EC2

Beat aims to make every city a better place to live offering more efficient and easy to use mobility solutions, with safety, reliability and sustainability serving as our fundamental pillars. Beat's work is focused towards growing and developing the cities where the company operates in, always relying on innovation." Founded in 2011, the ride-hailing company’s mission is to develop seamless mobility for a safe and sustainable urban life. Beat focuses on growing and developing the cities where it operates, continually searching for innovative ways to improve its application. While searching for optimization opportunities, the company realized it could use the latest Amazon Web Services (AWS) innovations to further improve its core database built on Amazon Aurora, a MySQL- and PostgreSQL-compatible relational database built for the cloud.

Beat chose to migrate its Aurora clusters to AWS Graviton processors, which are designed by AWS to deliver optimal price performance for cloud workloads. Using AWS Graviton processors alongside Aurora, Beat further optimized its costs and at the same time, improved its database’s performance. Beat was able to deliver speed and efficiency benefits to customers and employees alike. 

Beat is one of the fastest-growing ride-hailing applications in Latin America, with more than 26 million registered users and over 800,000 registered drivers worldwide. Headquartered in Greece, Beat serves users in several countries, including Peru, Argentina, Chile, Colombia, and Mexico. To support its expansion, Beat often looks for solutions to improve its database performance. “We were growing really quickly, and we didn’t want to spend a lot of time managing our database,” says Alexios Polyzos, site reliability engineer at Beat. “We chose to use Aurora as our core database in early 2017, and we never looked back. We’ve been very happy with the performance and reliability of Aurora for our internal teams and our customers.”

Beat hosts more than 65 production clusters and more than 210 databases on Aurora. To support databases for specific use cases, the company relies on additional AWS services. For example, Beat uses Amazon ElastiCache, a fully managed in-memory caching service, to support a sudden influx in users after its main competitor was banned by the Colombian government in February 2020. By focusing on finding the right tool for the right job, Beat has continually improved its database performance while optimizing its costs.

When Beat learned about the newer generation AWS Graviton2 processors, which deliver a major leap in performance and capabilities over first-generation AWS Graviton processors, it recognized another opportunity for optimization. The company could use AWS Graviton2–based instances to support a wide range of complex workloads on Aurora, delivering a faster and higher-performing ride-hailing service to its customers. “We realized that we could improve database performance and reduce costs by migrating to AWS Graviton2–based instances,” says Polyzos. “It was a logical choice for us to try the solution and see the results.”

Beat initiated the migration by upgrading to the MySQL 5.7–compatible edition of Aurora, which is required in order to use AWS Graviton2–based instances. Then, it conducted tests to verify that the migration would not interrupt service to its customers or employees. In early 2021, Beat began migrating its production clusters on Aurora to AWS Graviton2, which was a simple process for the company. “We migrated to AWS Graviton2–based instances without any major operations on our side,” says Polyzos. “We also gained significant performance improvements and cost savings immediately after the migration.”

On the average production cluster, Beat sees approximately 2,000 queries per second on the writer endpoint and more than 18,000 queries per second on the reader endpoint. After completing the migration, Beat saw significant improvements in database performance. “There were some cases where performance increased by 5–10 percent,” says Polyzos. “We had one specific case where we saw 40 percent performance gains.” Using higher-performing clusters, Beat can continue to provide fast, reliable, and high-quality service to its customers.

Beat has a custom automatic scaling solution configured for its clusters on Aurora. When a cluster reaches a certain traffic threshold, the mechanism adds a new instance. However, it would take 6–7 minutes for the instance to begin accommodating traffic. Using AWS Graviton2-based instances, Beat can better accommodate spikes and support more traffic without needing to create a new instance. “After migrating to AWS Graviton2–based instances, the spikes that used to initiate the automatic scaling mechanism became less frequent,” says Polyzos. “There’s no need for the mechanism to add new instances.” Using this solution, Beat reduced the automatic scaling mechanism’s sensitivity by 50 percent and decreased its monthly operational costs on Aurora by 10 percent.

By using AWS Graviton2 processors, Beat has delivered benefits to its employees as well as its customers. Internal teams sometimes perform jobs using bots, which would use nearly 100 percent of CPUs on certain instances. “Under our previous instances, our CPU usage would peak around 95–100 percent for about 15–20 minutes,” says Polyzos. “Using AWS Graviton2–based instances, it often peaks at around 85 percent, and for less time—around 10–12 minutes.” By reducing CPU usage, Beat can better manage power generation and further optimize time spent and costs on Aurora.

By migrating to AWS Graviton2–based instances, Beat has optimized its database performance, reduced costs, and continued its pattern of constant innovation. In the future, Beat will continue to use AWS services and improve its application to deliver high-quality ride-hailing services to millions of customers around the globe. For example, it plans to test AWS Graviton3–based instances, the latest generation in the AWS Graviton processor family, and will upgrade to them if the results show performance improvement.

With AWS, Beat can select the database that delivers the right level of performance and tools for each workload at an optimal cost, to help accelerate efficiency and innovation across its organization. “By migrating to AWS Graviton2, we can run the same number of instances on Aurora, see better performance, and save money without a lot of effort,” says Polyzos. “We’re really happy that we can keep innovating on AWS.”