|
|
Archives of the TeradataForum
Message Posted: Fri, 10 Nov 2006 @ 16:07:28 GMT
Subj: | | Re: UTF-8, Fastload and linux |
|
From: | | Michael Larkins |
The fixed location data can normally be a lot smaller than the delimited file too. If you think about it, reading a numeric value like 2800
requires 4 bytes character. If this were a smallint which can go up to 32767, it is saving 2 bytes of space on disk in the record, not wasting
it. If you are talking about numeric values bigger than 32000 than you are talking about integer and that is only 4 bytes vs potentially 11 in
character format. Plus, you do not need space for each and every delimiter. In other words, 8 fields means 7 or 8 delimiters = 7 or 8 bytes of
exact junk in the record that you do not want in the database.
So contrary to wasting space, the fixed location - "native" data format saves space. My posting had two parts, one was finding and
eliminating the delimiter to get to the actual good stuff. The other was avoiding the cost of converting data from varchar to int, dec,
smallint, et al. To get optimum speed and efficiency, you need to eliminate both of these extraneous activities. Remember, you are not
converting data nor finding delimiters in parallel. These are both serial operations on the host computer, not in Teradata.
Regards,
Michael Larkins
Certified Teradata Master
Certified Teradata SQL Instructor
| |