Amazon

With a stated mission of being “the most customer-centric company on earth”, Amazon.com has a big job of ensuring an exceptional experience for their tens of millions of customers worldwide. To do this, Amazon.com’s Client Experience Analytics (CXA) Team runs customer simulations against Amazon’s global web properties on an ongoing basis. These customer simulations help the team measure website latency across the globe, identify trends or issues, simulate website activity, and more. These simulations are done on a massive scale to mimic the 98 million active customer accounts across more than 10 web properties. As a result, these simulations produce a lot of data.

Amazon.com

Until recently, the CXA team was using a traditional MySQL Relational Database Management System (RDBMS) to store and query the data. But the size of their data set was increasing by 12-15 GB each day, which required the team to regularly upgrade to beefier, more expensive hardware. Even among the team’s MySQL-savvy staff, this required a significant time investment to provision the hardware and set up and configure the software. Not to mention the ongoing admin hours required to run regular backups and keep the software up-to-date with the latest patches.

The CXA team needed a more scalable solution and turned to a new service from Amazon Web Services: Amazon Relational Database Service (Amazon RDS). The CXA team was able to quickly migrate to the new database service because Amazon RDS offered the same capabilities of a MySQL database. “It was easy to get started. We were able to use the same code, applications, and tools we were using on our physical servers, without modification,” recalls Tomi Yiu, Software Engineer, Amazon.com. “With RDS, it’s so easy to add new instances, so we changed our system to do application-level database sharding. Our system is now using the RDS API to create new instances on-demand.”

Yiu continues, “We immediately noticed the DB admin time...