Better Game Data Processing Performance

I spent some time completely rewriting the backend service that fetches the stats data for everyone's Apex Legends account because I knew that the version that has been running for years was severely unoptimized. If you aren't interested in deeper details, just know that your stats are now retrieved faster with less chance of failure.

The service during peak times is requesting stats data for up to 15 accounts every second. These requests would take on average about 1.35 seconds (1350ms) to return with the data. 5% of the time the request would fail and need to be retried. (A retry meant it ended up being almost 3 seconds before the data would be available. This could be a bigger problem for some chat bots because they will error out if a command takes longer than 2-3 seconds.) Another down side is the service was VERY CPU intensive. The Kubernetes container the service runs in would regularly be completely consuming 8 CPU cores worth of processing. Thankfully I have more than enough computing power available but it was still not ideal.

After the rewrite of the service every metric is looking better. The average response time for a game data request has dropped down to 1.05 seconds (1050ms, 300ms improvement). The request retry is effectively at 0% now. CPU usage also dropped to ~0.12 of a single CPU core (essentially 1.5% of what it used to be).

Here's a look at the CPU usage of one of my Kubernetes nodes that the service ran on, you can definitely tell the difference when I switched to the new rewrite of the service. CPU usage dropped significantly.