Apache Pinot

And I don’t see anything about when to use onHeapDictionaryColumns, does that matter at all?

We found it useful in case of (conjunction of all below):
```1. Very high read qps (&gt; 10k)
2. Very low read latency SLA (p99 &lt; 100ms)
3. Long queries with too many values in IN clause for a string column.```
In such cases, we found that a lot of time was spent in off-heap dict lookup to find dict id's for those strings, and that resulted in too many String deserializations (from byte-buffer) that was slow as well as generated garbage.

My recommendation is to not use it, unless there are no other ways to improve performance, or one is a super user of Pinot and knows what they are doing.