I need to associate customers' order with their "level" (Silver, Gold, etc.) when they placed the order:

CRM server::CRM db::CRM table
CustomerID   PreviousLevel    NewLevel    NewLevelGrantedOn

Order server::Order db::Order table
OrderID     CustomerID    OrderPlacedOn

In an SSIS package I did this:

  1. Execute SQL against Order db to extract the orders and put them in an object variable;
  2. Loop through each order using "foreach container", in which I put a data flow task select top 1 * where CustomerID = ? and LevelGrantedOn < ? order by LevelGrantedOn desc that extract data from CRM db (both parameters come from step one), derive some columns and write the output to another table.

There are more then twenty thousand records in the order db, which means the data flow task will be executed for more than twenty thousand times. The CRM db will be queried for more than twenty thousand times, too. It takes more than an hour to do these.

Can I utilize some built-in features to speed up these (or do it in a "smart" way)? And, is an hour a long time, in the context of ETL and / or SSIS?

  • If you have the option, consider actually copying the results of the OrderDB query to a table in the CRM db. You might then be able to restate your CRM query so it's set-based, so you only run the one query (or a few sequential queries), instead of running your queries once for each row from OrderDB. And, an hour's only a long time if it's interfering with other processes, or if future growth will make it take longer until it does. A one-off process only requires enough optimization to complete successfully - a recurring process is worth spending a bit more time on.
    – RDFozz
    Commented Jun 21, 2017 at 15:48
  • @RDFozz no I can't write to either order or crm db.
    – Ryan
    Commented Jun 22, 2017 at 2:17
  • Can you add a user-created database to the CRM server, and copy the order info into that? Or, for that matter, set up the server the Order DB is on as a linked server on the CRM server? As much as I like and have used SSIS packages, trying to use them to join data from two servers is simply not something it generally does quickly, in my experience.
    – RDFozz
    Commented Jun 22, 2017 at 4:00
  • No I have no write permission on either server. I can only "select". The only db / server I have full control is the local server where I run the package.
    – Ryan
    Commented Jun 22, 2017 at 7:27

2 Answers 2


You can achieve this without using Foreach enumerator, just use one data flow task to achieve this.

Building the Package

First of all, add a Data Flow Task to the control Flow

1. OLEDB Sources

In the DataFlow task add an OLEDB Source that Read from Orders Table (the same command used in the Execute SQL Task (first step in your question)

Select * FROM Order

Also add a second OLEDB Source that read from customer table:

select * FROM CRM

2. Sort

After each OLEDB Source add a source component:

  1. The First (order table) you have to select CustomerID and OrderID columns for sorting and select the sort type as ascending
  2. The Second (customer table) you have to select CustomerID (sort type = ascending) and LevelGrantedOn (sort type = descending) columns for sorting

3. Merge Join

Add a merge Join component to join both sorted outputs. and select CustomerID column from both outputs as Join Key and select the output columns you need from both tables

4. Conditional Split

After merge Join add a conditional split to filter rows only that match the following expression

[LevelGrantedOn] < [orderdate]

5. Script Component

The final component we have to use is a script component to get only the first row for each customerID (because both source are well sorted just grabbing the first row is similar for Select top 1 ... ORDER BY LevelGrantedOn desc)

In the script component add an Output Column OutFlag of type DT_BOOL and Use the following script:

enter image description here

This script will set OutFlag to True when the customerID occur for the first time (similar to TOP 1)

    Public Class ScriptMain
        Inherits UserComponent

        Dim lstCustomerID As New System.Collections.Generic.List(Of Integer)
        Public Overrides Sub Input0_ProcessInputRow(ByVal Row As Input0Buffer)

            If lstCustomerID.Contains(Row.ID) Then

                Row.OutFlag = False

                Row.OutFlag = True

            End If

        End Sub

    End Class

6. Conditional Split

Add a second Conditional Split to filter Top 1 rows:

[OutFlag] == True

7.OLEDB Destination

Connect the Conditional Split Output top the OLEDB Destination and map the columns.

The Dataflow task must looks like

enter image description here

  • The LevelGrantedOn variable is from order table. It is, actually, the order date. How can I still have that variable if I use oledb source?
    – Ryan
    Commented Jun 22, 2017 at 2:15
  • Just edited my question to make it more clear.
    – Ryan
    Commented Jun 22, 2017 at 2:22
  • Can you create a linked server. It is the best way to achieve your goal.
    – Hadi
    Commented Jun 22, 2017 at 7:11
  • No I can't use linked server.
    – Ryan
    Commented Jun 22, 2017 at 7:25
  • 1
    I think both of your two answers are very informative, although the first one may not fit my situation very well. I'll try out the second one after I go back to work.
    – Ryan
    Commented Jun 22, 2017 at 13:58

Consider using SQL Server "Index with included columns" feature which can speed up the query performance during the data collection process.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.