Currently the managed database service from DigitalOcean does not provide support for automatically sharding your database.

Typically sharding is great when you are dealing with a massive data set that is no longer feasible to run on a single instance - think something along the lines of Twitter.

It’s benefits are obvious in that it was designed to basically serve datasets that can’t fit on a single node, however you can run into hotspots, where specific shards are being accessed more than others depending on the data that resides there.

This is why in 9 out of 10 cases you are better off just scaling with master slave replicas to horizontally and equally distribute your load. This will ensure consistent performance across each of your replicas and you won’t have to rebalance the shards or deal with hot spot issues.

Only in the case that you can’t fit all of your data on a single node would it be advisable to look into sharding, but at that point given the size of your dataset, you have no alternative but to shard.

So the basic answer is use sharding when you have no other choice, otherwise just scale horizontally with slave replicas.

I got your answer completely, thank you :)