|
|
Archives of the TeradataForum
Message Posted: Thu, 08 Nov 2007 @ 11:24:06 GMT
Subj: | | Re: Calculating number of compressible values as a |
|
From: | | Dieter Noeth |
Etienne E Stieger wrote:
| The above information from the manual is a bit confusing because: | |
| a) Is 8192 characters the only absolute cut-off point I should worry about? In other words, is this the only real constraint I should use
to calculate how many values I could compress without running into systems limits? | |
I once tried to find out that limit, too. It's probably the 8k limit, e.g. i could compress up to 31 values for a CHAR(255).
| b) Under which conditions would you get a thin header vs a fat header? | |
If you're running V2R6 and the header size hits 64k.
| c) Does thin vs. fat header have any impact here? I.e., could this become 2x8192 for a fat header? | |
AFAIK no.
| d) Are MVC values stored as varying length values in the table header, or is the datatype&size of the compressed column used
here?(i.e. padded with spaces). | |
They are stored with trailing blanks, that's why it looks so ugly in SHOW TABLE
| e) Does UNICODE/2-byte char columns have a different impact on the amount of header space used than LATIN/1-byte character columns?
In other words, are these 8192 character positions, or 8192 Bytes? | |
The dbc.tvfields.compressvaluelist is a VARCHAR(8192), so these are characters.
Dieter
| |