Pinterest has also simplified its build processes using Amazon EC2 Mac Instances. Because it no longer needs to procure machines and only pays for the macOS compute capacity it uses, Pinterest has significantly reduced its development costs.
Compared to running builds with on-premises machines, Pinterest can build and release iOS applications 18.4 percent faster using Amazon EC2 Mac Instances. It only takes a few minutes for Pinterest to spin up a new Amazon EC2 Mac Instance, compared to the hours or days it could take to procure a new machine. Developers no longer need to wait for capacity to become available during peak demand, improving development speeds. “During periods of low demand, we can easily scale them down,” says Koo. Pinterest uses Buildkite Agent Scaler, which runs as a function on AWS Lambda-a serverless, event-driven compute service-to efficiently scale the company’s CI/CD infrastructure by tracking the number of jobs that need to run and adjusting Pinterest’s Amazon EC2 Mac Instance fleet size accordingly. “On AWS, we can now increase the number of build machines by 400 percent at peak demand instead of managing a static group of machines,” says Koo. With this solution, the company has streamlined its CI/CD pipeline and can now scale its iOS build environment on demand. “Amazon EC2 Mac Instances is the first service that lets users run on-demand macOS environments in the cloud, which is a game changer.” In March 2021, Pinterest began migrating its CI/CD pipeline to Amazon EC2 Mac Instances.īy August 2021, Pinterest had completed its migration to Amazon EC2 Mac Instances. “We found that migrating to AWS would streamline the development experience,” says Koo. Pinterest realized that it could use this service to provision and access macOS environments within minutes, scale capacity as needed, and optimize its computing costs. Having used AWS services since it first launched in 2010, the company turned to AWS and learned about Amazon EC2 Mac Instances. Pinterest needed a scalable, efficient way to build, test, and deploy iOS features and upgrades. If one machine had an issue, we would experience service disruptions and higher latency.” This on-premises environment lengthened development, affecting Pinterest’s time to market for iOS features. “We needed staff to configure all the laptops and monitor and upgrade them continually. “Having to manage physical machines was very difficult,” says Oliver Koo, senior software engineer at Pinterest. To build, test, and deploy applications for its iOS users, Pinterest developers used on-premises computers, which the company needed to procure, configure, and upgrade manually. As a visual discovery engine that people use to find inspiration in categories like home, food, style, and beauty, Pinterest has grown to become a global platform.
0 kommentar(er)
