"But I heard that the Data Warehouse no need to create any indexes"...
- that is false

"and statistics is not gathered"
- that is also false

This is one of those things that is common in the IT industry. Someone says

"With an Exadata, there is a good chance that many of your indexes are redundant for DSS style queries"

and that is true. The problem is, someone takes that sentence and turns it into:

"Drop ALL of your indexes on Exadata"

which is definitely NOT true.

But onto the issue at hand - how to tune SQL on Exadata.

Just like *any* platform, your aim is to have your SQL take advantage of what works best on that platform. So for exadata, that means taking advantage of compression, predicate offloading, storage indexes, or probably the more common way of looking at it, is making sure your SQL's are not doing things that stop these optimizations from taking place, eg, single row lookup, expressions around columns, case-insensitive queries, buffered reads etc.

Similarly, you also want to tackle the *creation* of data, namely, how its loaded, because that too can have an impact on how efficiently it can be accessed by queries later.