Theoretically maximizing garbage collection and cache performance on low-end devices

[larssn]

I need a better understanding on how the Firestore SDK cleans up its cache, how it utilizes it, and how that affects performance.

We have an app with the default cache size of 40MB written in Flutter. It's an app that generates orders, and we've recently acquired a customer that generate a large amount of them (on Android devices).

They've seemingly run into a bottleneck with the caching layer: Performance slows to a crawl on the payment devices (which are low-end to say the least), until you clear the data storage for the app, after which it performs fine again.

I wish to improve the situation but I'm wondering what is better: Increasing the cache or decreasing it.

Increasing it:

• Maybe will cause the SDK to be able to garbage collection less often, but do more of it, in one go, improving general performance? Maybe.
• Might help active snapshot listeners by not interfering with their "active" data, that might otherwise get cleared out? Not sure if that's even a thing.
• Maybe the increased amount of data will cause all queries to take longer, due to having to emit from the cache first? Maybe.

Decreasing it:

• Less to clean up, meaning improved performance on an CPU bottlenecked device. Maybe.
• Maybe the devices will perform worse, having to garbage collect more often? Maybe.

I want to make an informed action on this, as having the customer running around and clearing 20+ devices each day is not sustainable. But it's hard to find any good information about this.

Any suggestions would be appreciated. 😊