 |
 |
Archives of the TeradataForum
Message Posted: Tue, 14 Jun 2005 @ 07:56:07 GMT
| Subj: | | Re: Compression and join indexes |
| |
| From: | | Victor Sokovin |
Hi Ole,
| | Yes, JI compression feature will help. | |
| | But field compression would help a lot more. | |
| | In our physical design we use a lot of "field splitting". | |
| | That is, any relevant field is split into say an 8 byte part and a 4-byte part. The value goes into only one of these fields, and the other
one is 0-compressed. The table is then accessed through a view summarizing the two parts of a given field. | |
I understand that saving of storage space is the main concern here? You might get less than optimal plans if you have to join on the combined
field. But it's OK if you just have to output the field in reports etc. A bit more work for the CPU but that's what the CPU are designed for.
| | Do you know, whether using JI compression have any cons? | |
I am not aware of any. I think JI compression is natural for the physical design of JI, it fits very well in it and it would just make the JI
more efficient. Of course, the question whether JI is the right instrument for the job must be answered before thinking about JI compression.
Compression is just a part of fine tuning.
It would be interesting to receive your feedback if you have a chance to compare typical query response time with different combinations of
repeating columns (columns in parentheses in the JI definition). That's the fun part of it.
Regards,
Victor
| |
