Create a database with as few tables as possible, insert as few rows as possible and a single query Q$.$

Record the query plan generated by the optimizer.

Insert as few $$additional$$ rows as possible such that the newly inserted rows result in a change in the

query optimizer$$s generated plan. The best case is to insert one single row that causes the query to $$fall

over the cliff$$ and to get changed to a different query plan.

Repeat the previous step $($insert$/$delete rows from the table$)$ till you get five different query plans.

Once you get the various query plans, force every query plan over every data set you used and record

the execution time of every single query over every data set. Please figure out if the query optimizer

made the right decision or not for every data set$/$query plan pairs.