Guide to Migrating from Databricks Delta Lake to Apache Iceberg

Kartik Sharma 28 Mar, 2024 • 7 min read


In the fast changing world of big data processing and analytics, the potential management of extensive datasets serves as a foundational pillar for companies for making informed decisions. It helps them to extract useful insights from their data. A variety of solutions has been emerged in past few years , such as Databricks Delta Lake and Apache Iceberg. These platforms were developed for data lake management and both offer robust features and functionalities. But for the organizations it is necessary to grasp the nuances in terms of architecture, technical and functional aspects for migrating the existing platform. This article will explore the complex process of transitioning from Databricks Delta Lake to Apache Iceberg.

Guide to Migrating from Databricks Delta Lake to Apache Iceberg

Learning Objectives

  • Understanding the features of Databricks and Apache Iceberg.
  • Learn to compare the architectural components between Databricks and the Apache Iceberg.
  • Understand the best practices for migrating the delta lake architecture to open source platform like Iceberg.
  • To utilize other third party tools as an alternative to the delta lake platform.

This article was published as a part of the Data Science Blogathon.

Understanding Databricks Delta Lake

Databricks Delta Lake is basically a sophisticated layer of storage built on the top of Apache Spark framework. It offers some modern data functionalities developed for seamless data management. Delta Lake have various features at it’s core :

  • ACID Transactions: Delta Lake guarantees the foundational principles of Atomicity, Consistency, Isolation, and Durability for all the modifications in user data, thus ensuring robust and valid data operations.
  • Schema Evolution: Flexibility comes predominantly with Delta Lake, because it seamlessly supports schema evolution thus enabling industries to carry out schema changes without disturbing existing data pipelines in production.
  • Time Travel: Just like the time travel in sci-fi movies, the delta lake provides the ability to query data snapshots at particular points in time. Thus it provide users to deep dive into comprehensive historical analysis of data and versioning capabilities.
  • Optimised File Management: Delta Lake supports robust techniques for organising and managing data files and metadata. It results in optimised query performance and alleviating storage costs.

Features of Apache Iceberg

Apache Iceberg provides a competitive alternative for companies looking for enhanced data lake management solution. Icebergs beats some of the traditional formats such as Parquet or ORC. There are lots of distinctive advantages:

  • Schema Evolution: The user can leverage the schema evolution feature while performing the schema changes without expensive table rewrites.
  • Snapshot Isolation: Iceberg provides support for snapshot isolation, thus guarantees consistent reads and writes. It facilitate concurrent modifications in the tables without compromising data integrity.
  • Metadata Management: This feature basically separates metadata from the data files. And store it in a dedicated repo which are different from the data files themselves. It does it so to boost the performance and empower efficient metadata operations.
  • Partition Pruning: Leveraging advanced pruning techniques, it optimises query performance by reducing the data scanned during query execution.

Comparative Analysis of Architectures

Let us get deeper into comparative analysis of architectures:

Databricks Delta Lake Architecture

  • Storage Layer: Delta Lake take advantage of cloud storage for example Amazon S3, Azure Blob as its underlying layer of storage , which consists of both data files and transaction logs.
  • Metadata Management: Metadata stays within a transaction log. Thus it leads to efficient metadata operations and guarantee data consistency.
  • Optimization Techniques: Delta Lake utilizes tons of optimization techniques. It includes data skipping and Z-ordering to radically improve query performance and reducing the overhead while scanning the data.
Databricks Delta Lake Architecture

Apache Iceberg Architecture

  • Separation of Metadata: There is a difference with comparison with Databricks in terms of separating metadata from data files. The iceberg stores metadata in a separate repository from the data files.
  • Transactional Support: For ensuring the data integrity and reliability, Iceberg boasts a robust transaction protocol. This protocol guarantees the atomic and consistent table operations.
  • Compatibility: The engines such as Apache Spark, Flink and Presto are readily compatible with the Iceberg. The developers have the flexibility to use Iceberg with these real-time and batch processing frameworks.
Apache Iceberg Architecture

Navigating Migration Landscape: Considerations and Best Practices

It needs immense amount of planning and execution to implement the migration from Databricks Delta Lake to Apache Iceberg. Some considerations should be made which are:

  • Schema Evolution: Guaranteeing the flawless compatibility between the schema evolution feature of Delta Lake and Iceberg to preserve consistency during schema changes.
  • Data Migration: The strategies should be developed and in place with the factors such as volume of the data, downtime requirements, and data consistency.
  • Query Compatibility: One should check about the query compatibility between Delta Lake and Iceberg. It will lead to the smooth transition and the existing query functionality will also be intact post-migration.
  • Performance Testing: Initiate extensive performance and regression tests to check the query performance. The utilization of resources should also be checked between Iceberg and Delta Lake. In that way, the potential areas can be recognized for optimization.

For migration developers can use some predefined code skeletons from Iceberg and databricks documentation and implement the same. The steps are mentioned below and the language used here is Scala:

Step1: Create Delta Lake Table

In the initial step, ensure that the S3 bucket is empty and verified before proceeding to create data within it. Once the data creation process is complete, perform the following check:

Step1: Create Delta Lake Table
val data=spark.range(0,5)
Create Delta Lake Table
Create Delta Lake Table

Adding optional vaccum code

#adding optional code for vaccum later
val data=spark.range(5,10)

Step2 : CTAS and Reading Delta Lake Table

#reading delta lake table"delta").load("s3://testing_bucket/delta-table")

Step3: Reading Delta Lake and Write to Iceberg Table


Verify the data dumped to the iceberg tables under S3

Reading Delta Lake and Write to Iceberg Table
Reading Delta Lake and Write to Iceberg Table

Comparing the third party tools in terms of simplicity, performance, compatibility and support. The two tools ie. AWS Glue DataBrew and Snowflake comes with their own set of functionalities.

AWS Glue DataBrew

Migration Process:

  • Ease of Use: AWS Glue DataBrew is a product under AWS cloud and provides a user-friendly experience for data cleaning and transformation tasks.
  • Integration: Glue DataBrew can be seamlessly integrated with other Amazon cloud services . For the organizations working with AWS can utilize this service.

Feature Set:

  • Data Transformation: It comes with large set of features for data transformation (EDA). It can come handy during the data migration.
  • Automatic Profiling: Like the other open source tools , DataBrew automatically profile data. to detect any inconsistency and also recommend transformations tasks.

Performance and Compatibility:

  • Scalability: For processing the larger datasets which can be encountered during migration process, Glue DataBrew provides scalability to handle that as well.
  • Compatibility: It provides compatibility with broader set of formats and data sources , thus facilitate integration with various storage solutions.


Migration Process:

  • Ease of Migration: For the simplicity , Snowflake does have migration services which helps end users to move from existing data warehouses to the Snowflake platform.
  • Comprehensive Documentation: Snowflake provides offers vast documentation and ample amount of resources to start with the migration process.

Feature Set:

  • Data Warehousing Capabilities: It provides broader set of warehousing features, and has support for semi-structured data, data sharing, and data governance.
  • Concurrency: The architecture permits high concurrency which is suitable for organizations with demanding data processing requirements.

Performance and Compatibility:

  • Performance: Snowflake is also performance efficient in terms of scalability which enables end-users to process huge data volumes with ease.
  • Compatibility: Snowflake also provides various connectors for different data sources, thus guarantees cross compatibility with varied data ecosystems.
Performance and Compatibility of Data Lake with UniForm


To optimize the data lake and warehouse management workflows and to extract business outcomes, the transition is vital for the organizations. The industries can leverage both the platforms in terms of capabilities and architectural and technical disparities and decide which to choose to utilize the maximum potential of their data sets. It helps organizations in the long run as well. With the dynamically and fast changing data landscape, innovative solutions can keep organizations on edge.

Key Takeaways

  • Apache Iceberg provides fantastic features like snapshot isolation, efficient metadata management, partition pruning thus it leads to improving data lake management capabilities.
  • Migrating to Apache Iceberg deals with cautious planning and execution. Organizations should consider the factors such as schema evolution, data migration strategies, and query compatibility.
  • Databricks Delta Lake leverages cloud storage as its underlying storage layer, storing data files and transaction logs, while Iceberg separates metadata from data files, enhancing performance and scalability.
  • Organizations should also consider the financial implications such as storage costs, compute charges, licensing fees, and any ad-hoc resources needed for the migration.

Frequently Asked Questions

Q1. How the migration process from Databricks Delta Lake to Apache Iceberg is performed?

A. It involves exporting the data from Databricks Delta Lake, clean it if necessary, and then import it into Apache Iceberg tables.

Q2. Are there any automated tools available to assist with the migration without manual intervention?

A. Organizations generally leverages custom python/Scala scripts and ETL tools to build this workflow.

Q3. What are the common challenges organizations encounter during the process of migration?

A. Some challenges which are very likely to happen are – data consistency, handling schema evolution differences, and optimizing performance post-migration.

Q4. What is the difference between Apache Iceberg and other table formats like Parquet or ORC?

A. Apache Iceberg provides features like schema evolution, snapshot isolation, and efficient metadata management which differs it from Parquet and ORC.

Q5. Can we use Apache Iceberg with cloud-based storage solutions?

A. Definitely , Apache Iceberg is compatible with commonly used cloud-based storage solutions such as AWS S3, Azure Blob Storage, and Google Cloud Storage.

The media shown in this article is not owned by Analytics Vidhya and is used at the Author’s discretion.

Kartik Sharma 28 Mar 2024

Frequently Asked Questions

Lorem ipsum dolor sit amet, consectetur adipiscing elit,

Responses From Readers