Archives of the TeradataForum
Message Posted: Tue, 21 Mar 2006 @ 10:26:36 GMT
Subj: | | Re: Mload vs tpump ( based on no of records ) |
|
From: | | Bob Hahn |
For a batch load, MDL v TPump performance depends on the block hit rate--basically the ratio of the count of rows in the input to the
datablocks in the table.
If MDLs block hit rate is 1 or less, it is actually slower than TPump on Teradata. When the block hit rate>1, then the throughput (i.e.
rows/sec or mb/sec) starts increasing with the hit rate.
Anita Richards used to do a Partners Presentation on Load Strategies that described this very well. (Last time 2004 I think).
So it isn't just the no of records, it is the no of records compared to the no of datablocks.
Note also, MDL has restrictions (no USIs, JIs on the target table). While TPump doesn't have these restrictions, there can be cross-session
locking issues to think through in those cases.
|