Loading Petabytes of Data to Azure
Published Sep 10 2020 08:54 PM 5,256 Views

My customer is having On-premises Hadoop environment with a data volume of approximately 80 Petabytes. As they started evaluating Azure to store their cognizant data which will be extracted from the On-Premises Hadoop environment and will be loaded to Azure Storage.

This cognizant data will be then further processed and analyze using Azure Synapse Analytics  

We started evaluating this cognizant data which is coming close to 5 Petabytes. And this volume needs to be loaded to the Azure Data Lake Storage Gen 2 for further processing.


  1. As Azure Data Factory had already proved to be one of the best tools to do data migration between data lake by its performance, robustness, and cost-effective mechanism to migrate data at scale. Some of the customers have successfully migrated petabytes of data consisting of hundreds of millions of files from On-premises to Azure via ADF with a sustained throughput of 2 GBps and higher.  

With the latest resume copy feature, customers can get benefits when they are loading very large data with control the process on data migration, and saving effort and time on data partitions or any solutions explicitly created for high resilience.

Below is the way how to resume the copy activity from the last failure point at file level:


More guidance on best practices of using ADF for data lake migration:

Use Azure Data Factory to migrate data from your data lake or data warehouse to Azure

Use Azure Data Factory to migrate data from an on-premises Hadoop cluster to Azure Storage

Azure Data Factory copy activity documentation

Data Factory offers two basic approaches for migrating data from on-premises HDFS to Azure. You can select the approach based on your scenario.

  • Data Factory DistCp mode (recommended): In Data Factory, you can use DistCp (distributed copy) to copy files as-is to Azure Blob storage (including staged copy) or Azure Data Lake Store Gen2. Use Data Factory integrated with DistCp to take advantage of an existing powerful cluster to achieve the best copy throughput. You also get the benefit of flexible scheduling and a unified monitoring experience from Data Factory. Depending on your Data Factory configuration, copy activity automatically constructs a DistCp command, submits the data to your Hadoop cluster, and then monitors the copy status. We recommend Data Factory DistCp mode for migrating data from an on-premises Hadoop cluster to Azure.
  • Data Factory native integration runtime mode: DistCp isn't an option in all scenarios. For example, in an Azure Virtual Networks environment, the DistCp tool doesn't support Azure ExpressRoute private peering with an Azure Storage virtual network endpoint. In addition, in some cases, you don't want to use your existing Hadoop cluster as an engine for migrating data so you don't put heavy loads on your cluster, which might affect the performance of existing ETL jobs. Instead, you can use the native capability of the Data Factory integration runtime as the engine that copies data from on-premises HDFS to Azure.

Network security- By default, Data Factory transfers data from on-premises HDFS to Blob storage or Azure Data Lake Storage Gen2 by using an encrypted connection over HTTPS protocol. HTTPS provides data encryption in transit and prevents eavesdropping and man-in-the-middle attacks.

Security recommendations for Blob storage

Azure Storage encryption for data at rest


  1. Azure Databox: Use Data Box family of products such as Data Box, Data Box Disk, and Data Box Heavy to move large amounts of data to Azure when you’re limited by time, network availability, or costs. All data is AES-encrypted, and the devices are wiped clean after upload in accordance with NIST Special Publication 800-88 revision 1 standards

Migrate from on-prem HDFS store to Azure Storage with Azure Data Box






  1. Azure Import/Export service- Azure Import/Export service is used to securely import large amounts of data to Azure Blob storage and Azure Files by shipping disk drives to an Azure datacenter.




Version history
Last update:
‎Sep 11 2020 06:56 AM
Updated by: