Our general requirement is 5k IOPS with < 1ms latency for the database container(s). As for a single collocated instance, we can make this work but right sizing the instance is important (you can “overbuild” in the cloud for no real benefit, and it can even work against you).
As Aram indicates, the capacity planning guide talks about all of this, and is based on vCPU cloud units. I also provide a capacity planning tool at the same link as the capacity planning guide, that can simplify sizing.
If you want to discuss in more detail based on your specific installation, feel free to contact me directly and we can summarize here. Thx.