Adding a method to get accurately the bandwidth consumption (download/upload) per CID

[sameh-farouk]

Describe the feature you are proposing

Adding a method to get the bandwidth consumption per CID

Additional context

I'm not aware of any (straightforward) way/method to get this info from the cluster client or from the IPFS client.

I'm implementing a pinning service that uses the cluster client to pin data to a cluster of IPFS peers, I need to be able to query the cluster for information about the bandwidth consumption per CID, and the storage. for storage, I can get an IPFS shell and query the node about the recursive data size of a root CID. but I have a hard time thinking about how I can get the bandwidth-related info in an accurate way that can be used for a billing service.
I read somewhere about the possibility of getting such info by monitoring some logs, or by querying the bit swap ledger for how many times a block was requested. but none is reliable or straightforward to production use.

This was the most helpful info i was able to find:
https://gitcoin.co/issue/23631
https://github.com/tzdybal/cidtrack

Given that it was 2 years ago, probably a lot of things was changed/improved.

[hsanjuan]

This is a problem beyond what cluster can solve right now (it is Kubo that downloads the things after all).

The problem about how much storage a Pin takes is also difficult (given overlaps). Kubo/ipfs needs to be instrumented to do that sort of accounting.