Hi guys!

First of all: I´m from Austria so excuse my bad english!
I´m still workin on an OLAP project since 3 month! During this time I found
some important questions but no answers. These are as follows:

1. In the Dimension-Editor the property "member count" is only set during
the creation of the dimension and not during the preparing-process (is this
the english expression?) What for do I need this property? Only for
erstimation-values during the preparing-process?

2. Is there a way to pass over "ring-connections" in the cube-editor?

3. How can I include one dimension two times in a cube connected to
different forein keys in the fact-table?

4. I have only one Cube with a lot of dimensions. Is it better to make this
dimenions private or public?

5. When du the creation of the aggregations occur? During the
memory-administration (in german "Speicheraufbereitung") or during the
preparation of the cube? The second one I guess!?

6. I have to do the memory-administration only one time! Is this correct?

7. The database for the dimension is growing every time and the preparation
works automatically. What happens when member-count exceed the limit of
64000 members?

thx in advance!!
D. Österreicher

Hi Dietmar,
first of all, don't worry about your english, there are
austrians in California who speak terrible english and
get elected as governor ;o).

Anyway, let me take a shot at your questions:
It is a good indicator on dimension size at creation time.

I do not quite understand the word "ring-connection". In
the cube editor one (and only one) dimension table out of
every dimension is linked 1:many to the fact table. There
should not be any circular structure or AS will complain.

do this you need to give it an alias.

case you might need them for a second, future cube. If
you are sure, that you will never need the dimension
again, you should make it private. Shared dimensions will
be loaded into memory just once (for every user role, I
think)

they can be redesigned later.

Yes and no. You have to design aggregations just once,
but as I said before, you can redesign them later, if
they don't fit your needs.

cube. (This has not happened to me yet)

Hope that helps (and that I do not get corrected too much)

Lutz

Hi Lutz!

Arnie for president

thanks a lot for your answers! But I have to ask a second time:

@1: thanks, I´m quite happy!

@2:
+---------------+
+---------+ +------+-----+---------------+ +------+-----+

I have this constellation with this two dimensions:
- coast center > article group 1 > articel
- coast center > article group 2 > articel

It´s not possible to built this constellation in the cube editor, but of
course you can do it in the database! I tried to insert an alias of the
table "coast center" but when I a delete the relation from the "ring" AS
removes the whole dimension from the cube! Andy ideas?

@3: I´ve to test it but is there a second dimension in the
client-application when you do that?

@4: Ok, so I make private dimensions. Is there a way to convert my existing
public dimensions?

@5-6: thx

@7: thats correct! So, if I expect more than 64000 elements I have to make
an additional grouping level.

@8: new one

I have some virtual dimensions, wich are parts of other regular dimension.
Are they slower in the client application?

@8: an other new one

Is there a way to use "actions" in MS Excel. Maybe Excel 2003 supports
this??

Thanks for this BIG HELP! Thank you very much!!!!
Dietmar

group can contain several articles (m:n). There are
several ways to solve this I guess.

a) choose a primary article group for each article thus
avoiding the m:n relationship
b) create a cube for every constellation
c) use article and article group as separate dimensions
and reflect the m:n via the fact table

I am sure there are more solutions, including a weighting
factor or something like that, but these are all I can
come up with now.


convert unless you have to, though. Shared dimensions are
useful.


dimensions. You will not notice unless you have very
large and complex dimensions. Usually virtual dimensions
are used, if you want to preserve storage space. i.e.
base a v-dim on a large customer dimension.

tell.

HTH Lutz

In regards to the dimension level counts, since these values are not updated
during dimesion processing, I would recommed that you set this value to the
value that you expect the dimension size to grow to. In many cases, people
begin to design their data warehouse with just a fraction of their
dimensional data. If the dimension level count is set at that point, it
will not be reflective of how many members will be in the levels during
production. These values are important for designing aggregations. Please
review the following whitepaper for more information.

http://www.microsoft.com/technet/tre...e/ANSvcsPG.asp

And if you are using partitioning, it is a good idea to set the dimension
level counts to values that will be represented in the partitions. For
example, if you partition by month and you have 5 years of data, set the
dimension level count to 1 for the month level and 31 for the day
level.....not 60 and 1800...since there will only be one month in each
partition and there will only be 31 days at most in any partition.

Sean


--
Sean

--
Sean Boon
SQL Server BI Product Unit

--
This posting is provided "AS IS" with no warranties, and confers no rights.
Use of included script samples are subject to the terms specified at
http://www.microsoft.com/info/cpyright.htm.

"Dietmar Österreicher" <dietmar.oesterreicher (AT) at (DOT) bosch.com> wrote

These counts are part of the algorithm behind the aggragation (storage)
wizard. It's a good thing if they are semi-accurate. You can hand-enter
estimates, or there's a command in the dimension editor Tools menu to
re-count.

tom @ the domain below
www.tomchester.net


"Dietmar Österreicher" <dietmar.oesterreicher (AT) at (DOT) bosch.com> wrote

Hi! Thank you for your explanations!

One more question: is there a way to set the member-count property to
"select count(*) from ..." in a DTS-workflow? I have a preperation-workflow
in which it would be best to be placed!
Any alternatives to set this property?

Dietmar