Amazon Data Firehose
Dependable real-time stream loading into analytics services, data lakes, and warehouses.
- Capturing, transforming, and loading streaming data is simple. With a few clicks, you can create a delivery stream, choose your destination, and begin streaming data in real time.
- Provide and scale network, memory, and processing resources automatically without constant management.
- Without creating your own processing pipelines, you may dynamically segment streaming data and convert raw streaming data into formats like Apache Parquet.
How it operates
The simplest method for obtaining, transforming, and sending data streams to analytics services, data lakes, and data warehouses in a matter of seconds is offered by Amazon Data Firehose. Setting up a stream with a source, destination, and necessary modifications is necessary in order to use Amazon Data Firehose. The stream is continuously processed by Amazon Data Firehose, which scales automatically according to the volume of data available and provides it in a matter of seconds.
Source
Choose your data stream’s source, such as a stream in Kinesis Data Streams, a topic in Amazon Managed Streaming for Kafka (MSK), or write data using the Firehose Direct PUT API. You can build up a stream from sources like Amazon CloudWatch Logs, AWS WAF web ACL logs, AWS Network Firewall Logs, Amazon SNS, or AWS IoT because Amazon Data Firehose is connected into more than 20 AWS services.
Data Transformation (optional)
Choose whether you wish to decompress the data, execute custom data transformations using your own AWS Lambda function, convert your data stream into formats like Parquet or ORC, or dynamically partition input records based on attributes to send into separate places.
The destination
Choose a destination for your stream, such as Splunk, Snowflake, Amazon Redshift, Amazon OpenSearch Service, Amazon S3, or a custom HTTP endpoint.
Use cases
Flow into warehouses and data lakes
Without creating processing pipelines, stream data into Amazon S3 and transform it into the formats needed for analysis.
Increase security
Use supported Security Information and Event Management (SIEM) solutions to keep an eye on network security in real time and generate warnings when possible threats materialize.
Create applications for ML streaming
To evaluate data and forecast inference endpoints as streams go to their destination, enhance your data streams with machine learning (ML) models.
Use Amazon Data Firehose to replicate database updates to Apache Iceberg tables (in preview)
A new feature in Amazon Data Firehose that records modifications made to databases like PostgreSQL and MySQL and replicates the changes to Apache Iceberg tables on Amazon Simple Storage Service (Amazon S3) is being made available in preview today.
An excellent open-source table format for large data analytics is Apache Iceberg. Open-source analytics engines like Apache Spark, Apache Flink, Trino, Apache Hive, and Apache Impala can operate with the same data simultaneously with Apache Iceberg, which also adds the simplicity and dependability of SQL tables to S3 data lakes.
This new feature offers a straightforward, end-to-end way to stream database updates without affecting database applications’ transaction performance. To transmit change data capture (CDC) updates from your database, you can quickly set up a Data Firehose stream. Data from various databases can now be readily replicated into Iceberg tables on Amazon S3, allowing you to access current data for machine learning (ML) and large-scale analytics applications.
Typical Enterprise clients of Amazon Web Services (AWS) utilize hundreds of databases for transactional applications. They wish to record database changes, such as the addition, modification, or deletion of records in a table, and send the updates to their data warehouse or Amazon S3 data lake in open source table formats like Apache Iceberg so that they can do large-scale analytics and machine learning on the most recent data.
Many clients create extract, transform, and load (ETL) processes to read data from databases on a regular basis in order to accomplish this. However, batch tasks can cause many hours of delay before data is ready for analytics, and ETL readers affect database transaction speed. Customers seek the option to stream database changes in order to lessen the impact on database transaction performance. A change data capture (CDC) stream is the name given to this stream.
Installing and configuring several open-source components is necessary for the initial setup and testing of such systems. Days or weeks may pass. The operational overhead is increased by the engineers’ need to validate and implement open source updates, monitor and manage clusters after setup.
CDC streams from databases can now be continuously replicated to Apache Iceberg tables on Amazon S3 using Amazon Data Firehose’s new data streaming feature. A Data Firehose stream is created by defining its source and destination. An initial data snapshot and all ensuing modifications made to the chosen database tables are captured and continuously replicated by Data Firehose as a data stream. Data Firehose minimizes the impact on database transaction performance by using the database replication log to obtain CDC streams.
AWS Data Firehose automatically splits the data and keeps records until they are sent to their destination, regardless of how frequently the number of database updates changes. Cluster management and fine-tuning, as well as capacity provisioning, are optional. As part of the initial Data Firehose stream creation, Data Firehose can automatically generate Apache Iceberg tables with the same schema as the database tables in addition to the data itself. It can also dynamically develop the target schema, adding additional columns, for example, in response to changes in the source schema.
You don’t need to use open source components, install software upgrades, or pay for overhead because Data Firehose is a fully managed service.
Amazon Data Firehose offers a straightforward, scalable, end-to-end managed solution for delivering CDC streams into your data lake or data warehouse, where you can execute extensive analysis and machine learning applications. It does this by continuously replicating database updates to Apache Iceberg tables in Amazon S3.
Things to be aware of
Here are some other things to be aware of.
The following databases on Amazon RDS and self-managed PostgreSQL and MySQL databases on Amazon EC2 are supported by this new feature:
- Amazon Aurora PostgreSQL-Compatible Edition with Amazon RDS for PostgreSQL
- Amazon Aurora MySQL-Compatible Edition with Amazon RDS for MySQL
Throughout the trial period and beyond general availability, the team will keep adding support for other databases. They informed me that support for MongoDB, Oracle, and SQL Server databases is already in the works.
Data Firehose connects to databases in your Amazon Virtual Private Cloud (Amazon VPC) via AWS PrivateLink.
You have two options when configuring an Amazon Data Firehose delivery stream: you may define a class of tables and columns using wildcards, or you can specify particular tables and columns. When using wildcards, Data Firehose will automatically construct new tables and columns in the destination if they match the wildcard and are added to the database after the Data Firehose stream is created.
Accessibility
With the exception of the Asia Pacific (Malaysia), AWS GovCloud (US), and China regions, all AWS regions now offer the new data streaming feature.
Amazon Data Firehose pricing
At the start of the preview, there are no fees for your use. In the future, the price will be determined by your actual usage, such as the number of bytes read and supplied. There are no upfront costs or obligations. To learn more, be sure to read the pricing page.
