TLDR A simple metrics-based ranking system is good enough to decide who gets how many resources.

Computational resources – CPU time, memory usage, network traffic etc – are limited. This may be more or less of a problem depending on project/company size and so on; if you’re working on a smaller product with limited traffic, it might not be meaningful at all.

Once past a certain threshold though, expenses with such resources become non-trivial and it begins to make sense to spend some time thinking about how to distribute them as efficiently as possible.

Here’s the problem that got me thinking about this: at work, we had a computational resource that needed to be consumed by a large fleet of workers (think several thousand concurrent), but each type of worker had different productivity, and that productivity changed over time. How can we decide who gets what?

Here at Sensor Tower we handle large volumes of data, so to keep things snappy for our customers we need to think carefully about how we process and serve that data.

Understanding the data we’re handling is a fundamental part of improving the way we serve it, and by analyzing how an important backend service worked, we were able to speed it up by a factor of four.