Frequency v.s Heigh Balanced Histogram

Frequency

·         Each bucket has its own size (counts of rows with that value)

·         NDV <= Buckets request <=254

·         One distinct value per bucket

·         Bucket_size = rows with “that” particular value

  

Height Balanced

·         Bucket size = (num_rows – num_nulls) / num_buckets

·         Values are sorted before splitting them into buckets

·         One or more distinct values per bucket

·         Only values of end-points are stored in histogram

·         Bucket_size = (Orig_CarD – Nnulls) / Buckets_total