Fastload Phases

[BillDHS]

During Phase 1, Fastload loads the data in blocks on the amps associated with the sessions. During Phase 2, Fastload sorts the data by hash on the "final" amp.

Question: does fastload distribute the data from the "load" amp to the "final" amp during Phase 1 or 2, or start distribution in 1 and complete during 2?

 

[dnoeth]

Phase 1:
raw data is sent in blocks to the next available session/AMP.
A receiver task on that AMP unpacks the blocks, does CHECKs/CASTs/FORMATs, hashes the row and sends it to the final AMP, which writes unsorted blocks to disks.

Phase 2:
Each AMP sorts it's blocks and writes it to disk.
After that FALLBACK rows are generated.

Dieter