Hi,

Config:
Win 2003 SRV SP1 x64
SQL2005 + AS 2005 EE x64

I have an multiuser application wich should allow users "write" data in
the cube. I want to use the push mode processing for that. So my
application is generating an XMLA push processing command and it send it
to the AS server.

So my questions are:

1. It is possible that many users in the same time to push data in the
cube? (up to 30 users can push data concurentlly)

2. I'm not using SSIS for pushing the data and I want to build the XMLA
in my application. I've build the XMLA but I don't know/understand where
and how should the recordset look like. I presume the recorset should
be beteen:

....
<DataSource xsi:type="PushedDataSource">
<root Parameter="InputRowset"/>

<--**** here should be the recordset=? which format? ***

<EndOfData Parameter="EndOfInputRowset"/>
</DataSource>
<Bindings>
....

3. How the aggregations are affected by the push-mode processing? There
are aggregations build or not? Should I process the cube on a time basis
that the aggregations are build or should I activate lazy aggregations?


Thanks for your help,
Gerald

Hi,

you can write into a cube, I'm not sure about the XML/A command to do this,
but you can and this works in a transaction, so multiple users can write at
the same time.
but AS is NOT a database, if you plan to have intensive writes, maybe its
not an option to write into the cubes.

when you write in a cube, the aggregations are reevaluated, so the "sums"
will display the new value in real time (like total sales will be the new
sales value after the new user input)

"Gerald Aichholzer" <gerald.aichholzer (AT) gmx (DOT) net> wrote

Hi Jeje,

The original post was made by me but under the account of a friend because I
was having at the time no internet access :P

SO, actually I'm not trying to write in the cube in the AS2000 way. The cube
is MOLAP and is not write enabled. I'm trying to "inject" data in the cube
using push-mode processing with the ProcessAdd option.

This means, I need to make a planning application and I need to be able to
splash the data across dimensions. Using write enabled cubes I've got very
bad query performance. So now I write an AS stored procedure who is
computing the recordset generated by splashing and I want, in parallel to
write the recordset in the facts table and to "inject" it in the cube.
Theoretiocally this is possible using the push-mode processing wich is
generating in the background a temporary partition and then merge it to the
destination partition.
I'm wondering now how this thing will work in an multiuser enviroment.

Regards,
Radu


"Jéjé" <willgart_A_ (AT) hotmail_A_ (DOT) com> wrote