Problem in getting data from Oracle to DS : Too Slow

[ray.wurlod]

Probably worth starting a new thread on this one.
Quick answer: hashed files work best when the average record size is small (less than 10% of group size) with the smallest possible variance, and when there is maximum variability (randomness?) in key values.
To determine the counts of distinct values in each column (field) assumes that there is a constant structure in all records, which is not a requirement in hashed files; indeed, the repository structure depends on this. It is sufficient to gather statistics on physical record sizes.