Package structure in SSIS

Hi

I have been working with DTC packages together with analysis service, and are

now begining on SSIS.

But I have some demands that's been giving me headache before, so I hope

there's a solution in the SSIS for the following.

My DTC packages was build to be executed i 5 steps/Packages

1. Transfeer data from my Oracle 9.2.0.4 to SQL 2000

2. Transform data

3. Quality check

4. Process dimension

5. Process cubes

BUT in 2005, I want a little more flexibilty. I want to execute smaller

packages, let's say, only my "general ledger".

This way I will be able to update some of my important cubes quicker. And

not wait until all data

is transferred.

So what I have done until now is that I have created SSIS packages per module. I

have a General_ledger.dtsx and a Sales.dtsx. Which I can execute separately.

I also have a MasterFlow.dtsx which contain an "Execute package Task"

for each package.

I can run the MasterFlow and it will run through all packages once every night.

Each package contains

1. Truncate package tabels

2. Transfer package

tabels

3. Execute dimensions

data

4. Execute cubes

5. Process

Dimensions

6. Process Cubes

Is this the best way?

I'm not satisfied

myself because,

when I run MasterFlow. I would like to execute all

"step 1" in all packages, before "step 2" and so forth.

Then I would be sure that all data is OK when all Cubes are processed, if I

have a virtual cube that contains data from 2 packages.

How would you design this kind of setup?

Thanx in adv.

Best Regards

Bjarne R Mikkelsen

A-TEX A/S

Why not have a third child package that processes the cubes? then you only have to execute the cube process upon successful completion of everything else.

Or, in each child package you could process the measure group (I'm assuming you are using AS2005) relevant to teh data that you have just transformed.

Your approach up to now seems pretty sound to me.

-Jamie

|||Thank you.

The processing of dimension and cubes are the least time consuming, so it would be a possibility, simply to arrange these in a 3. child package, which will be executed everythime a single package is run.

/Bjarne

OWC in Windows 98 connect to Analysis Services 2005

Is there a posibility to make OWC installed in Windows 98 connect to Analysis Services 2005?

The thing is we cannot install the OLE DB Provider 9.0 for Analysis Services 2005, because it needs Windows Installer 3.1, and we really need to connect OWC to 2005, because we could do it with Analysis Services 2000.

How can we do?

Thanks in advance.

If you want to use OWC, you must have the OLEDB provider 9.0 for Analysis Services. OWC executes on the client and uses OLEDB to talk to the SSAS server.

Unfortunately the only options I can think of if you cannot get the OLEDB provider installed is to either upgrade the client or to look at using something other than OWC. Some of the commercial "thin client" solutions would probably work. Mosha has a list on his site at www.mosha.com/msolap .

Overriding repeating totals from non-referenced slicer dimension

I've run across what appears to be a change in the default behavior of Analysis Services from 2000 to 2005. My question is how to get the default behavior to revert to the AS 2000 method.

If you have a cube with multiple measure groups or multiple linked cubes, queries that try to slice a measure using a dimension that is not related to the measure result in the total amount for any other slicers being repeated across the "invalid" slicer.

For example, the following query slices the reseller amount by calendar year and employee department. The results are as expected. Only the Sales department has any totals for the measure and all other departments return null.

SELECT { [Employee].[Employee Department].[All Employees].CHILDREN } ON COLUMNS ,{ DESCENDANTS( [Date].[Calendar].[All Periods], [Date].[Calendar].[Calendar Year] ) } ON ROWS FROM [Adventure Works] WHERE ( [Measures].[Reseller Sales Amount] );

The next query is identical, except it using the Internet Sales Amount measure, which comes from a measure group that does not have a link to the Employee dimension. The result is the total amount for internet sales is displayed across all employee departments. In AS 2000, the results would have been null or empty.

SELECT { [Employee].[Employee Department].[All Employees].CHILDREN } ON COLUMNS ,{ DESCENDANTS( [Date].[Calendar].[All Periods], [Date].[Calendar].[Calendar Year] ) } ON ROWS FROM [Adventure Works] WHERE ( [Measures].[Internet Sales Amount] );

At one level I can agree with the change, but non-the-less, it is a change in the default behavior, and users hate it. Is ther any way to revert this behavior back to the AS 2000 approach?

Thanks!

What you are running into is the fact that AS2K5 now applies ValidMeasure() automatically. You can revert back to AS2K behavior by setting the "IgnoreUnrelatedDimensions" property associated with your measure group to "False".

HTH,

- Steve

|||

Steve,

Many thanks! I knew it would be something simple, but was getting tired of looking...