NoPI hashing algorithm


NoPI hashing algorithm


It seems that for a NoPI table , a query id is submitted to a randomized hashing algorithm to produce a hash value comprising of hash bucket number and a remainder. This number is used to assign a row into an amp. Once it reaches an amp, the amp software in turn generates a different hash bucket number and treats it as a RowID.

If it is so, then how does it do a lookup to the hashmap and it is  more time-consuming than a normal PI.

Can anyone please throw some light upon this for me?

Thanks and regards,