Skip to main content

SSIS Performance Comparison

While working with a client this past week, I ran into a scenario where I needed to compare two dates in a SSIS package. One date would be the same for all rows, and one date would be different for each row. There are two possible implementations: include the "same date" in all rows and do a column comparison, or put the "same date" into a variable and do a comparison between the variable and column. My first instinct was that the column comparison would be faster because the extra trip to the server would be more expensive than extra column space. As I thought about it more, I realized that as the size of the dataset increases, the time to handle the extra column might overtake the hit from the extra trip.

So I ran a test to find out!

Scenario:

I created a temp table in the AdventureWorksDW database called FactLottaInternetSales that contains multiple copies of the FactInternetSales database, but with all orderdatekeys set to 1.

I created two packages to do the comparison.
Package 1: Column.dtsx

1 Data Flow Task containing:

  • 1 OLE DB Source:
    SELECT top x *
    FROM [AdventureWorksDW].[dbo].[FactLottaInternetSales]
    WHERE orderdatekey=1
  • 1 Conditional Split: OrderDateKey <>
  • 1 Row Count (as a terminator)

Package 2: Variable.dtsx

1 Execute SQL Task which assigns the result to the variable DateKey:
select max(orderdatekey)
from adventureworksdw.dbo.FactLottaInternetSales
where orderdatekey=1

1 Data Flow Task containing:

  • 1 OLE DB Source:
    SELECT top x [ProductKey] ,[OrderDateKey] ,[DueDateKey] ,[ShipDateKey] ,[CustomerKey] ,[PromotionKey] ,[CurrencyKey] ,[SalesTerritoryKey] ,[SalesOrderNumber] ,[SalesOrderLineNumber] ,[RevisionNumber] ,[OrderQuantity] ,[UnitPrice] ,[ExtendedAmount] ,[UnitPriceDiscountPct] ,[DiscountAmount] ,[ProductStandardCost] ,[TotalProductCost] ,[SalesAmount] ,[TaxAmt] ,[Freight] ,[CarrierTrackingNumber] ,[CustomerPONumber]
    FROM [AdventureWorksDW].[dbo].[FactLottaInternetSales]
    where orderdatekey=1
  • 1 Conditional Split: @[User::DateKey] <>
  • 1 Row Count (as a terminator)

Results:




The variable package takes longer up to about 2 million rows. After that, the column package takes longer! Happy performance tuning :)


Version: SQL Server 2005 SP2

Comments

Popular posts from this blog

SSIS Configuration to Configuration to Configuration Schema

I've gotten several requests to put down in writing the configuration schema that I use as the base of my SQL Server Integration Services framework. It contains a set of configurations: an indirect environment variable configuration, which points to an XML configuration file configuration, which points to a SQL Server configuration. I learned this configuration from the Project REAL Reference Implementation . If you're getting started with a BI implementation, I highly recommend that you download it for some great ideas and best practices. Steps to implement: 1) Create an environment variable on your machine with the name of SSIS_CONFIG_FILE and the value of: C:\SSIS\Config\MasterConfigFile.dtsConfig. 2) Create an SSIS configuration file at C:\ SSIS\Config\MasterConfigFile.dtsConfig with the line: <configuration valuetype="String" path="\Package.Connections[CONFIG_SERVER].Properties[ConnectionString]" configuredtype="Property"><configuredv

Execute SQL Task Designer Limit

After migrating a package from DTS to SSIS, I had a problem with an Execute SQL Task. I couldn't change any characters in the SQLStatement property; I couldn't add any new characters; I could delete characters, but not retype them! After googling several variations of "integration services" "read only" and "Execute SQL Task", I deleted about half of the entry in a fit of frustration. Lo and behold, I could type again. Apparently, there is limit on the size or number of characters that can be entered in the SQLStatement property. From my experimentation, I came up with a limit of 32767 characters. The interesting thing is that the restriction only seems to be on the designer. If you set the SourceType to "Variable" and use a variable that contains more than 32767 characters, the task will execute. Also, if you use the "Direct Input" SourceType and modify the package XML to set the SQLStatement longer than 32767 characters,

Reporting Services 2008 Configuration Mistake

To start working with the management side of SQL Server Reporting Services 2008, I decided to set up a report server and report manager. Unfortunately, I made a mistake while setting up my configuration that left me a little perplexed. Here are the steps I took to cause, track down, and solve the issue. Problem: I began by opening the Reporting Services Configuration Manager from the Start Menu. I clicked through each of the menu options and accepted the defaults for any question with a warning symbol, since warning symbol typically designate an action item. After two minutes, all of the warning symbols had disappeared, and I was ready to begin managing my report server. Unfortunately, opening up a browser and trying to open up the report manager resulted in the dreaded " The report server has encountered a configuration error. (rsServerConfigurationError) " message. Sherlock-ing it: I put on my sleuthing hat and went to the log file directory: C:\Program Files\Microsoft