フリー問題

問題文

Cleaned output will be queried almost always by a date range and occasionally by customer identifier. Which partition key is right?

選択肢

1. The date, because the common access pattern then prunes whole directories, while the occasional identifier lookup can be served by reading the range and filtering.
2. Both, with the identifier nested inside the date.
3. The customer identifier, because it has the highest cardinality and a high-cardinality key produces the largest number of partitions, which gives the reader the finest possible granularity for pruning whichever of the two access patterns is used at query time.
4. Neither; store one flat directory and rely on column statistics.