Real-time data ingestion in Athena — Apache Kafka to Amazon Athena

This guide explores a common use case: optimizing and transforming event data from Apache Kafka to Amazon Athena.
Apache Kafka is an open-source distributed event streaming platform set apart by its ability to handle a high volume of data from various systems and applications. Popular data warehouses such Snowflake and Amazon Redshift take data from Kafka and enable you to run analytics and derive insight. But these warehouses are designed to support batch processing. They also struggle with data consistency, particularly when there’s a constant flow of evolving data. This can increase both the cost of preparing data and the time it takes to design, implement, and maintain your data architecture.
This guide uses example data that contains nested fields. It shows how you move data from the raw zone of your data lake to the staging zone, and then to the refined zone for further analysis by engines like Snowflake.
The raw zone is used to store all data. As the name implies, the data in question is uncatalogued, uncleaned, unverified, and unoptimized for analysis. At this point, the data is only being collected and stored.
The staging zone is where SWQLake automatically catalogs data for easy discovery and converts it to an optimized file format. Within this file format, compression is improved, in turn reducing storage costs. It also increases data access, which improves the read performance. The data has still not been transformed, but you can define data pipelines that read and transform the data from this zone for storage in the refined zone.
Finally, SQLake automatically optimizes data in the refined zone. First, it converts the files into Apache Parquet, then partitions the data by date. Following this, the schema evolves over time as SQLake continuously compacts small files. The end result is a fully-optimized data lake that is prepared to produce queries in a matter of minutes instead of days.
The process of transforming data in data lakes
The following steps explain how to create a data pipeline that reads from a Kafka raw zone, stages the data in the staging zone, and then performs basic transformations before writing the results to a refined zone such as Snowflake, from where you can query the results.

Before you begin

SQLake runs on AWS Virtual Private Cloud (VPC). An Upsolver Customer Service representative can help you set up your user account and deploy SQLake on AWS.
When you first deploy SQLake in the customer VPC, you create an Identity and Access Management (IAM) role that gives you access to any AWS resources you might require to build data pipelines. More about these permissions here. In this scenario, your cluster has the permissions to read and write to the AWS Glue Data Catalog and to S3.
However, if your Kafka cluster is in a VPC different from the one created for Upsolver and isn’t accessible via a public IP, you must create a VPC peering connection.
The procedure for optimizing and transforming event data from Apache Kafka to Amazon Athena consists of 5 steps:
  1. 1.
    Connecting to Apache Kafka in the raw zone
  2. 2.
    Staging the data in the staging zone
  3. 3.
    Copying from Kafka
  4. 4.
    Creating an output table in Amazon Athena
  5. 5.
    Prepareing your data for the refined zone and reading your data from Amazon Athena

Step 1

Connecting to Apache Kafka in the raw zone

Create a Kafka connection to consume data from your cluster. This connection gives you the ability to configure the AWS IAM credentials, bucket names, and prefixes that SQLake needs to access the data. To do this, all you need to know is familiar SQL syntax.
Create a Kafka connection:
CREATE KAFKA CONNECTION my_kafka_connection
HOSTS = ('')
security.protocol=SASL_SSL required username="XXXXXXXX" password="-----------";

Step 2

Staging the data in the staging zone

You can use SQLake to stage your data. To achieve this, create a job in SQLake that reads from Kafka and creates a table with the appropriate schema in the AWS Glue Data Catalog.
Here’s the code to create a staging table to store your raw data:
CREATE TABLE default_glue_catalog.default.kafka_staging_table;

Step 3

Copy from Kafka

Create a job to extract that data from Kafka and load it onto the staging table.
There are several types of jobs you can create, such as copying the data from the source to your destination. To copy the data to the staging zone, use the COPY FROM job type.
The following SQL statement creates a job that reads data from the Kafka topic to a staging location and catalogs the metadata in the AWS Glue Data Catalog. SQLake automatically creates a Glue Data Catalog table and uses the name you designate in the INTO statement for the database and table names. It also creates an S3 prefix in your source S3 bucket using the name given in the PREFIX statement.
Create a COPY FROM job:
INTO default_glue_catalog.<DB_NAME>.<STAGING_TABLE_NAME>;

Step 4

Create an output table in Amazon Athena

When you build your data lake you should maintain schemas, partitions, and other relevant metadata information as new data arrives and existing data changes. It’s ideal to maintain this metadata in a central catalog that data practitioners and query engines can access to find and query the data. The AWS Glue Data Catalog fills this purpose. The AWS Glue Data Catalog is a fully-managed metadata store integrated with a wide range of tools and services, including Upsolver SQLake.
With a typical installation, SQLake automatically creates a connection to the Glue Data Catalog, which displays in your navigation tree. However, you can manually create a new connection using the following SQL:
Create an output table in AWS Glue Data Catalog:
CREATE TABLE default_glue_catalog.upsolver_samples.orders_transformed_data(
partition_date date)
PARTITION BY partition_date;
You don’t need to define all of your output columns, as SQLake automatically adds any missing columns. If you do want to strictly control which columns to add, the when you CREATE JOB you can define them in the CREATE TABLE statement and set the ADD_MISSING_COLUMNS job property to FALSE.

Step 5

Prepare your data for the refined zone

You’re ready to start transforming your data. To do this, you can choose from three types of transformation jobs:
  1. 1.
  2. 2.
  3. 3.
To take full advantage of these jobs, we recommend that you explicitly create your output table, which enables you to define partitions, primary keys, and more.
After you define the output table, create an INSERT job to read data from the staging table and write the results into the target table inside your refined zone.
Create an INSERT job to read from staging and write to the refined zone:
CREATE JOB transform_orders_and_insert_into_athena_kinesis
AS INSERT INTO default_glue_catalog.upsolver_samples.orders_transformed_data MAP_COLUMNS_BY_NAME
orderid AS order_id, -- rename columns
MD5(buyeremail) AS customer_id, -- hash or mask columns using built-in functions
nettotal AS total,
$commit_time AS partition_date -- populate the partition column with the processing time of the event, automatically casted to DATE type
FROM default_glue_catalog.default.kinesis_staging_table
WHERE eventtype = 'ORDER'
AND $commit_time BETWEEN execution_start_time() AND execution_end_time();

Read your data from Amazon Athena

To view your pipeline results, query your table using SELECT.
SELECT * FROM default_glue_catalog.<DB_NAME>.<TRANSFORMED_TABLE_NAME> LIMIT 10;Conclusion
This how-to guide explained how to take raw data from Kafka, stage the data, and then transform it to meet your business needs. The approach was modeled after a common data lake design pattern, in which data transitions through raw, staging, and refined zones. Using only familiar SQL syntax, you can use SQLake to create data pipelines and organize your data to more easily perform analytics and ML.
And as your business needs evolve, so can your data. In the future, you can create additional jobs that use the staging table as the source of creativity and innovation, while your pipelines indefinitely keep your data fresh.

Try it yourself

To launch the template:
  1. 1.
    Launch SQLake by navigating to
  2. 2.
    Sign up by entering your email address and setting a password.
  3. 3.
    After you click through the onboarding panel, a panel with a list of templates displays.
  4. 4.
    Select the Kafka to Athena template.