Archives of the TeradataForum
Message Posted: Wed, 08 Mar 2006 @ 12:47:06 GMT
| Subj: | | Re: Performance of Fastload vs Multiload |
| |
| From: | | Dieter Noeth |
Ole Dunweber wrote:
| | According to my understanding both utilities uses the concepts of "internal checkpoints" during their respective application
phases. | |
| | This is not to be confused with checkpoints taken in their respective acquisition phases. | |
| | In a 3rd party vendor book it's described that MLoad logs a checkpoint each time a data block is successfully written to its target
table. | |
| | I can't find this in the manuals - but I reckon this "data block checkpoint" is the same as an "internal checkpoint"? | |
There's hardly anything missing from the manuals, but if, then you'll find it in the 1st party training materials :-)
From the Application Utilities 3.9 training:
"The following principles guide the MultiLoad checkpoint strategy:
Acquisition phase
Checkpoint is performed according to user specification as defined in the .BEGIN MLOAD statement. This checkpoint can be based on time or on
number of records processed. The default checkpoint interval is fifteen minutes.
Application phase
Checkpoint is performed each time a data block is written to the target table. Each block is written one time.
Sort phase(s)
Sort operations do their own internal checkpoint that overrides the higher level checkpoints."
There're no details about FastLoad checkpoints within the training, but i understand it this way:
- Fastload's phase 1 checkpoint is the same as MLoad's Acquisition phase checkpoint
- FastLoad's phase 2 checkpoint is a "Sort phase" checkpoint
| | If so, and if a Mload acquisition phase equals a Fastload Acquisition phase + a Fastload Application phase, then Mload would issue
all these internal checkpoints against the work table in the Mload acquisition phase. And this doesn't seem correct with regards to how restarting
from an "acquisition-checkpoint" is described? | |
According to the training the Acquisition phase consists of several steps:
First:
"Get host data to the appropriate AMP worktables" -> Acquisition phase checkpoint
When all data is sent:
"Sort the reblocked records in the work tables" -> Sort phase checkpoint
Dieter
| 
