Enabling compaction optimizer

To enable compaction

1. Open the AWS Glue console at https://console.aws.amazon.com/glue/ and sign in as a data lake administrator, the table creator, or a user who has been granted the glue:UpdateTable and lakeformation:GetDataAccess permissions on the table.

2. In the navigation pane, under Data Catalog, choose Tables.

3. On the Tables page, choose a table in open table format that you want to enable compaction for, then under Actions menu, choose Optimization, and then choose Enable.

   You can also enable compaction by selecting the Table optimization tab on the Table details page. Choose the Table optimization tab on the lower section of the page, and choose Enable compaction.

   The Enable optimization option is also available when you create a new Iceberg table in the Data Catalog.

4. On the Enable optimization page, choose Compaction under Optimization options.

   

5. Next, select an IAM role from the drop down with the permissions shown in the Table optimization prerequisites section.

   You can also choose Create a new IAM role option to create a custom role with the required permissions to run compaction.

   Follow the steps below to update an existing IAM role:

   1. To update the permissions policy for the IAM role, in the IAM console, go to the IAM role that is being used for running compaction.

   2. In the Add permissions section, choose Create policy. In the newly opened browser window, create a new policy to use with your role.

   3. On the Create policy page, choose the JSON tab. Copy the JSON code shown in the Prerequisites into the policy editor field.

6. If you have security policy configurations where the Iceberg table optimizer needs to access Amazon S3 buckets from a specific Virtual Private Cloud (VPC), create an AWS Glue network connection or use an existing one.

   If you don't have an AWS Glue VPC connection set up already, create a new one by following the steps in the Creating connections for connectors section using the AWS Glue console or the AWS CLI/SDK.

7. Choose a compaction strategy. The available options are:

   • Binpack – Binpack is the default compaction strategy in Apache Iceberg. It combines smaller data files into larger ones for optimal performance.

   • Sort – Sorting in Apache Iceberg is a data organization technique that clusters information within files based on specified columns, significantly improving query performance by reducing the number of files that need to be processed. You define the sort order in Iceberg's metadata using the sort-order field, and when multiple columns are specified, data is sorted in the sequence the columns appear in the sort order, ensuring records with similar values are stored together within files. The sorting compaction strategy takes the optimization further by sorting data across all files within a partition.

   • Z-order – Z-ordering is a way to organize data when you need to sort by multiple columns with equal importance. Unlike traditional sorting that prioritizes one column over others, Z-ordering gives balanced weight to each column, helping your query engine read fewer files when searching for data.

     The technique works by weaving together the binary digits of values from different columns. For example, if you have the numbers 3 and 4 from two columns, Z-ordering first converts them to binary (3 becomes 011 and 4 becomes 100), then interleaves these digits to create a new value: 011010. This interleaving creates a pattern that keeps related data physically close together.

     Z-ordering is particularly effective for multi-dimensional queries. For example, a customer table Z-ordered by income, state, and zip code can deliver superior performance compared to hierarchical sorting when querying across multiple dimensions. This organization allows queries targeting specific combinations of income and geographic location to quickly locate relevant data while minimizing unnecessary file scans.

8. Minimum input files – The number of data files required in a partition before compaction is triggered.

9. Delete files threshold – Minimum delete operations required in a data file before it becomes eligible for compaction.

10. Choose Enable optimization.