SQL Server – Merging Partitions – Does it physically move data from a data file to another?

Partitioning is a common practice when maintaining large number of records in tables regardless of the database type: whether the database is an OLTP database or a data warehouse based on dimensional modeling. Partitioning requires a Partition Function that describes boundary values and number of partitions required, and Partition Scheme that assigns File Groups to partitions.
Here are few posts I have made on partitioning;
Once a table is partitioned, based on the workloads, we add new boundary values for introducing new partitions to the table and we remove boundary values for combining partitions (or removing partitions). These two operations are done with two partition related functions: SPLIT and MERGE. Now the question is, when we remove a partition using the MERGE function, if partitions are distributed with different file groups (means different data files), does SQL Server moves data from one file to another? Do we really have to consider it?
Let’s make a database and do a simple test. Let’s create a database called AdventureWorksTest and add multiple file groups and data files.
As you see, we have five file groups now.
Let’s create a Partitioned Table and load it using AdventureWorks2014 Sales.SalesOrderHeader table. Note that, below code creates a Partition Function with 2014, 2015, 2016 and 2017 as boundary values. And it creates the Partition Scheme setting with multiple file groups.
Here is the result of the query.
Here is another query to see how data is distributed with each file group.
If I illustrate the same in different manner;
This is how the data is distributed with my system now. If I try to delete a data file, example, AdventureWorksTest_Data_03 that is belong to FileGroup3, SQL Server will not allow me to do as it holds data.
If I decide to combine 2015 and 2016, in other words, if I want to remove the 4th partition that holds 2016 data, what will happen. Let’s do and see. The below codes merge the partition;
Once delete, let me run the same query to see how data is partitioned now.
As you see, FileGroup3 is no longer used by the table and number of records in the FileGroup2 has been increased. The partition which was 5 is now 4. This means, data that was maintained with FileGroup3 (or AdventureWorksTest_Data_03) has been moved to FileGroup2 (or AdventureWorksTest_Data_02). This is what has happened;
This clearly shows that data is moved when partitions are merged. Okay, do we have to really consider about this? Yes, we have to, when this is a small table, we do not see any performance issue but if the table is large, moving data from one file to another will surely take time, hence, think twice before merging, specifically when they are distributed in multiple files.
** Now, if you need, you can delete the AdventureWorksTest_Data_03 file and SQL Server will allow you to delete as it is empty.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

w

Connecting to %s