Backend

What are the best practices for databases in web3? Like do we want to host as much on-chain as possible and make all the calls to the chain or de want to put into DBs

Founder of Prognoze.xyz Forecasting and media.

Building paywithglide.xyz. ex Robinhood, Uber, Apple

Indeed. Would you also use a decentralised indexer i.e. the Graph to query the on-chain data or opt out for your own indexer + a fast DB infra?

See farcaster as a good example. Puts only the required data on-chain, everything else is implemented in either centralized dbs or through other off-chain mechanisms.

On-chain storage is slow and scarce. Use it, but only as needed

+1 to @tusharsoni.eth's opinion. Put the minimum required to provide users value on chain, e.g. proofs, and use more efficient solutions for everything else. If decentralization matters, there's IPFS, but most expensive storage and compute can be done offchain.

I could regurgitate just to hear myself talk but what the two above said really is the answer