Archives of the TeradataForum

Message Posted: Thu, 10 Apr 2003 @ 18:26:56 GMT

	<Prev	Next>		<<First	<Prev	Next>	Last>>

This is correct. The first step is to locate which part of Multiload is the bottleneck.

If it acquisition then it could be any number of things.

1. Perhaps you're channel is saturated.

2. If your input file has a large row length then you may not be getting many rows per data block. A solution to this could be to determine if all of the fields in the record are being loaded to the target table. If not then replace the .FIELD "FIELDNAME" with .FILLER "FIELDNAME".

3. Check the SESSIONS count that multiload is getting or asking for. Usually the rule is a SESSION per AMP.

4. If you are reading your input file from tape check to make sure you have adequate data buffers allocated on the file.

If it is the apply phase then:

1. Check your PI to make sure distribution is not a problem on the target table. I have to disagree with the statement below to make it a UPI but a badly skewed NUPI could have a definite affect on MLOAD performance. Besides if it is a NUPI then you obviously would have it there for end user query performance so I would take the MLOAD hit.

2. If you have SI's then drop them before MLOAD begins and recreate them after the MLOAD is finished. Much faster than letting MLOAD do the maintenance.

3. Check the system load and if you have Priority Scheduler what partition your load is running in.

Lastly 3 hours for loading 25 million rows may be the best you can hope for. It kind of depends on your configuration on both sides (hosts and Teradata).

Rob Bankston
Teradata Certified Professional
UnumProvident - Data Warehouse

	<Prev	Next>		<<First	<Prev	Next>	Last>>