Archives of the TeradataForum

Message Posted: Wed, 21 Mar 2001 @ 17:20:08 GMT

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

Greetings to All:

Although all ideas have merit, I offer this consideration:

Delimiters are fine for small amounts of character data. If you are exporting a small amount of data, you should be using BTEQ because it is faster than FastExport. FastExport was written for very large amounts of data. FastExports speed comes during the sending of the blocks, not the selection of the rows. A SELECT is a SELECT whether submitted by BTEQ, FastExport, Queryman, etc.

FastExport must arrange the data rows into sequencial blocks for transmission whenusing ORDER BY. This means that rows must be redistributed across all AMPs at least 2 times to build the sequencial blocks. BTEQ only needs select the rows and return them 1 row at a time.

If you are exporting millions of rows, the delimiter processing will:

A. turn FastExport into Half-FastExport

B. it will take more disk space than having the numeric data in native format

C. increase the chance of impacting what FastExport does well

Now that Teradata is communicating more with Baby databases, these things seem to be a consideration. But, we should think twice before requesting that NCR turn everything into a hammer. There are jobs for screwdrivers, wrenches and an assortment of tools in our respective tool belts.

This is my opinion and you know what "they" say about opinions. I do not expect everyone to agree with me. As stated earlier, it is for consideration only.

Best regards,

Mike

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