[SteveLauC]

Regarding write-heavy workloads, especially for Postgres, I think we really need to distinguish between INSERTs and UPDATEs, because every update to a tuple in Postgres duplicates the whole tuple due to its MVCC implementation (if you use the default heap storage engine)

[brightball]

One thing that catches people by surprise is that read heavy workloads can generate heavy writes.

Queries that need to operate on more data than will fit in the allocated working memory will write to a temporary table on disk, then in some cases perform an operation on that temporary table like sorting the whole thing and finally, after it's done delete it which is even more disk write stress.

It's not really about whether it's ready heavy or write heavy, it's about whether it's usage creates Disk I/O stress.

You can write millions of increment integers and while technically that's "write heavy", there's no stress involved because you're just changing the value in a defined space that's already been allocated. Update space that is more dynamic, like growing a TEXT or JSON field frequently...it's a different story.