How Stackery Cut Their AWS Costs 39% by Switching to Aurora Serverless
Visualizing and building scalable serverless architectures is now easier than ever to get started on the cloud.
People have been at the heart of Stackery’s ethos since day one. When the company was founded five years ago, co-founder Chase Douglas simply wanted to help people make better applications. He built a low-code visual editor for Infrastructure-as-Code templates, enabling developers to build scalable serverless architectures and making it easier than ever to get started on the cloud.
Over the past five years, the platform has expanded to include environments, automated security, and CI/CD pipelines. Through it all, everything Stackery does—from who they partner with to what products they build—puts the quality of developer and end-user experience first.
Chase Douglas and Farrah Campbell, Stackery’s Co-founder/CTO and Ecosystem Director, respectively, are both AWS Serverless Heroes with strong opinions about how things should be done in the cloud. “And Chase has been here since the very beginning,” says Campbell. “Stackery is his baby. It’s very rare anybody outside our organization gets to touch this code.”
But in late 2020, an opportunity slid across Douglas’ desk that was too good to pass up: AWS Jumpstart would help fund a tricky database migration, which he and the rest of the team had been putting off for months due to lack of time.
Together with Trek10, Stackery decided to migrate their database from RDS instances to RDS Serverless Aurora—and to carve out the time to do it, they would squeeze the migration into the handful of weeks over the end of year holidays, before the rest team of the team got back to work.
Simpler Architecture, Lower Cost
Trek10 and Stackery have a partnership that stretches back to early 2020. “We have a stellar team of engineers,” Douglas says, “and pride in what we’ve built. It takes a lot for us to bring someone else into work lockstep with us.”
If there was anyone outside the organization Stackery could rely on to execute this project, Douglas thought, it was Trek10.
For months, Douglas had this migration on his mind, but database migrations, he said, were always hairy. “Nothing about them is trivial.” And their development calendar was perpetually full; the migration got placed on the back-burner until the team had more time to focus on it.
The upsides, however, would be huge. Aurora Serverless’ data API feature would let Stackery connect to their database over stateless, encrypted connections while removing the need for their VPCs. They could then eliminate NAT Gateways and bastion hosts, slashing their total AWS costs by as much as 30-40% across all environments—a simpler architecture with much slimmer billing.
But Douglas was worried about two roadblocks he could see coming before the project even began. First, Aurora Serverless’ data API was a powerful construct, but a brand new one. There wasn’t much existing software that could use it, and the drivers Douglas did find had no test coverage; Stackery would have to write their own. Second, “We couldn’t have a second of downtime during the migration itself,” he said, “not one.” Douglas laughed, adding, “I brought that up so many times I bet Trek10 got tired of hearing me say it.”
Three-Week Sprint to Launch
“I have something to protect here,” said Douglas, “and I never felt like Trek10 was taking that for granted.”
While Douglas got to work building a data API driver, Trek10 got ready to migrate Stackery’s data from RDS instances to RDS Aurora Serverless. That was where the two teams hit their first blocker—the migration tool they had decided to use was occasionally failing to sync record deletions, creating inconsistencies.
“I won’t lie,” said Adam Baker, Cloud Engineer at Trek10, “I got nervous for a second. It was a solvable problem, but this could easily balloon our timeline and scope.” The two teams went back to square one, creating a database backup (including snapshots for each region) and manually addressing any data that was out of sync, and then working to validate the data before cutting over to the new database. While Trek10 worked to get everything running, Stackery worked in tandem on DMS debugging.
“Working with Chase and Farrah at Stackery was such a cool experience,” said Baker. “There were these moments where it really seemed like we weren’t going to push this through in time, but we got it done, and fast, thanks to the way we could problem-solve with them on the spot and make quick decisions.”
With the pieces falling into place, Stackery began testing and error fixes on their custom-built API client and driver. On January 5, Douglas sent Trek10 a Slack saying they were about to go live—the entire database migration took a total of four weeks from start to launch.
Open-source & the Advantages of Aurora Serverless
While the lack of data API tooling was one of their biggest concerns before tackling the project, Stackery now has a highly-tested and proven AWS Aurora data API driver and SQL client that gives them more stability than ever, Douglas says, with their database.
Stackery’s overall cost savings after migrating to Aurora Serverless are a sizable 39%, and there have been clear efficiency gains in their development processes as well.
On RDS Instances, Stackery had to micro-manage connections to their database to prevent exceeding limits on the number of concurrent connections; with RDS Aurora Serverless, they can now leave connection management to the Aurora Data API, which simplifies their code. They can also connect to their own database using the database’s IAM permissions, meaning from a development and support standpoint they don’t have to deal with tunneling into a VPC to access instances as they work.
And as an added bonus, they never have to worry about scaling.
“Working with Trek10, we had this strong sense of trust,” says Campbell. “I could tell Trek10’s values aligned with ours, that, like us, they truly want to help others learn.”
“Aside from one early-days contractor who was really more like an employee, Adam at Trek10 is the only person outside of our company who’s touched our codebase,” says Douglas. “In my opinion, that says it all.”
Stackery has open-sourced their data API tools, which you can find, and use, below: