Hi Tom,

On 4/12/07 1:40 PM, "Tom Lane" <[EMAIL PROTECTED]> wrote:

> I would suggest that *all* of those TODOs are premature in the absence
> of experimental evidence about the effect of varying the parameters.
> If we end up finding out that the existing settings are about right
> anyway across a range of test cases, who needs more knobs?  We've got
> too many mostly-useless knobs already.

This set of TODOs is really about a binary change in behavior that data
warehouse users will employ to shift given columns into a separate storage
mechanism while preserving their schema.  By contrast, the knob you describe
is about tuning the existing storage mechanism by offlining values that are
too large.

We can talk about these TODOs as a group using the name "vertical
partitioning" if that suits.

To demonstrate the effectiveness of vertical partitioning, we would write
queries that use the partitioned columns independently or as groups that
correlate with the storage mechanism.

The other demonstration applies to the use of compression techniques that
align with the column type(s) and operate across tuple boundaries within
pages.  Examples there include the segmentation of fixed width types and
variable width types into separate page storage classes, which allows for
the application of different compression and/or representations on pages.

- Luke 

---------------------------(end of broadcast)---------------------------
TIP 4: Have you searched our list archives?


Reply via email to