Hi,

we have cubes partitioned on the time dimension. The majority of data that
comes in if for the last 4 days or so. The new data is loaded into a seperate
fact table, and loaded into a partition. This partition is then merged into
the current month partition.

However, we can have data that comes in which is a month or longer old. So
would be merged into the incorrect partition. And would remain there until a
full reprocess of the Fact tables are cube partitiions.

I understand that when data slices are defined for a partition, that AS2000
will only visit the partitions that contain the relevent data. What happens
if we use data slices in our situation?
If the data is in the incorrect partition, as a result of merging, is it
possible that the data is missed out, depending on the query? Or would the
merging not be allowed to take place?

Would we want to define data slices for all partitions, except for the
current month ?

Or would we want to define a filter, and rather than a build and merge, do
an incremental build against the cube ?

We are really looking for the quickest load time possible, while having
correct data of course.

Thanks in advance for any advice and help.

Is this for AS 2000 or 2005? In 2000, slices that you defined are used for
querying. In 2005, slices for MOLAP partitions are detected automatically.
For ROLAP partitions the slice you specify will be used to eliminate
partitions from queries.

But besides that, only your latest partition would have "incorrect" data,
right? So just don't set a slice for that last partition -- it would mean
that this partition would be unnecessarily scanned for some queries, but it
should be a fairly small partition anyway...

Thanks,
Akshai

--
--
This posting is provided "AS IS" with no warranties, and confers no rights
Please do not send email directly to this alias. This alias is for newsgroup
purposes only.

"Al" <Al (AT) discussions (DOT) microsoft.com> wrote