Flyclops case study

Flyclops switched to PlanetScale to handle 100% growth without a DBA

Flyclops is a game studio in Philadelphia, PA specializing in casual multiplayer games, both asynchronous turn-based and real-time. They developed “Domino!”, the leading multiplayer dominoes game on iOS and Android, and are the team behind “Why Are We Running?” with its ever-growing user base. Flyclops’ games have been played by millions across the globe. Notably, they have accomplished all of this without having a single Database Administrator (DBA) on staff.

Challenges and goals

Dave Martorana is the co-owner and partner at Flyclops, in addition to being a software engineer. When high-level company-wide technical decisions need to be made, they usually start with him. Before migrating to PlanetScale, Flyclops was using AWS Aurora to handle their 1.5TB of legacy data, with specifically selected instances for their memory-bound workload. When things went wrong, things went really wrong — and AWS couldn’t be there to help. With more traffic and new products coming, it was obviously time for a change.

In 2021, with the cyclical nature of their usage spiking higher, a database with more than 1.5TB of data that was rapidly growing, and 11 years of legacy data to contend with, it was time for Dave to make a decision about the future of Flyclops’ data.

When Flyclops took an honest look at their set of existing challenges, it added up to a ton of work. Their main challenges were:

• Support from AWS was expensive and severely lacking
• Lack of tools for gaining insight into how to guarantee consistent performance
• An aging data set deployed with increasingly outmoded technology
• Desire to keep staffing down and continue to grow the business without a DBA
• Zero confidence in their ability to make essential schema changes to data once it has been deployed

Technology requirements

Whatever tool they ended up switching to, it became clear that the following criteria were necessary conditions:

• Easy-to-use migration tools for their 1.5TB and 11-year old data set
• A serverless architecture that wouldn’t involve provisioning hardware themselves
• The ability to choose memory-intensive instances to run their infrastructure on, as they do not have the typical CPU-bound workload that characterizes many dynamic database installations
• Developer tools for schema changes, making forward progress easier
• Auto-failover and replication, both for migrating and for daily use
• The ability to scale up to 2× current workloads to meet user growth

When it comes down to it, PlanetScale is much cheaper than hiring DBAs, and that’s something we couldn’t ignore.

When the team looked around at potential solutions, PlanetScale quickly floated to the top:

• PlanetScale is MySQL-compatible and migrating didn’t require changing a single line of application code
• The level of support from the beginning was unmatched by anything they had experienced in over 10 years of working with AWS
• The migration story and tooling were also unique — the PlanetScale team clearly understood the inherent challenges in migrating business-critical legacy data

We were on AWS Aurora… but we were not getting anywhere near the level of support we needed, especially for the premium we were paying. We were looking for a solution that could actually support us. We saw that with PlanetScale, we could essentially leverage their team of experts as our own DBAs when we needed them most.

Solution

To recap, Dave was looking at 1.5TB of legacy data in custom AWS instances with a support burden that just wasn’t being met, on top of a new product and lots of new traffic about to come online.

We’re confident that our PlanetScale deployment can safely handle more than twice the workload we’re currently experiencing.

One challenge was that Flyclops doesn’t have the typical CPU-bound workload that many dynamic database installations have. Instead, they are more memory bound. Thankfully, PlanetScale’s technology and team are both flexible enough to handle these requirements, and after some experimentation, they were able to land on the right instance types and sizes to keep them running smoothly far into the future.

I don’t want to run my own test installations. I don’t want to run my own databases. The entire value of our company is a database and we can run front-end servers and microservices, but when it comes to managing our data, that’s not something that we want to be doing ourselves. We want somebody else to do that, with a lot more expertise and a lot more systems in place and a lot more safety and security than we would be able to provide.

Outcomes

It’s been a little over a year since Flyclops switched to PlanetScale. How has it been going? Here is a summary of what they’ve experienced so far:

• Superior tooling: PlanetScale’s migration and runtime developer tools make day-to-day development easier
• Increased confidence: Schema changes have gone from something they had “zero confidence in“ before, up to a “4 out of 5.”
• Increased velocity: Better tooling + more confidence = shipping more and better software, more often
• Cost savings: Using PlanetScale is much cheaper than staffing up a team of database experts. Instead, they can just rely on PlanetScale

With PlanetScale, we’ve definitely made mistakes that put an undue tax on our deployment. PlanetScale helped us figure out the issues, came to our aid, and recommended the perfect fix. That level of support can’t be beat.