Why I like ODA

Not too shabby:

SQL> col rows format 999,999,999,999
SQL> set timing on 
SQL> select /*+parallel(24)*/ count(*) as "ROWS" from daily_xxx_summary; 

            ROWS 
---------------- 
   7,099,066,073 

Elapsed: 00:00:04.51 
SQL> 

Processing a lot of fairly sizeable datasets (incoming data stream is currently at 70,410 row inserts/sec), and parallel processing scales extremely well (and almost linear in some cases, which is unexpected).

Question: any specific statistics to look at for performance and scaling of PQ on ODA? Interconnnect plays a minor role with only 2 nodes (as compared to my 12 node RAC) I assume - or should I spend time on looking at gc events too?