This message was deleted.

that will heavy depend on your use case, maybe if the number of segments increase you could have problems just to keep the segments meta info in the heap I can't find any material over this exact subject, but I use 200GB for each 4vCPU/8GB RAM, but I want querys with < 100ms latency

We've not hit any limitations on disk size for the historicals (we have considerably more disk than you're suggesting moving to). You will have to scale your memory and CPU up though but that will depend on the number of segments your data is held in. I've not found any useful ratio for this, it's very much trial and error utilising metrics... We've had to scale well above the rule-of-thumb 24GB of memory as per https://druid.apache.org/docs/latest/operations/basic-cluster-tuning#historical