ETL and Data Pipelines with Shell, Airflow and Kafka

ETL or Extract, Transform, and Load processes are used for cases where flexibility, speed, and scalability of data are important. You will explore some key differences been similar processes, ETL and ELT, which include the place of transformation, flexibility, Big Data support, and time-to-insight.

You will learn that there is an increasing demand for access to raw data that drives the evolution from ETL to ELT. Data extraction involves advanced technologies including database querying, web scraping, and APIs. You will also learn that data transformation is about formatting data to suit the application and that data is loaded in batches or streamed continuously.

Extract, transform and load (ETL) pipelines are created with Bash scripts that can be run on a schedule using cron. Data pipelines move data from one place, or form, to another. Data pipeline processes include scheduling or triggering, monitoring, maintenance, and optimization. Furthermore, Batch pipelines extract and operate on batches of data. Whereas streaming data pipelines ingest data packets one-by-one in rapid succession. In this module, you will learn that streaming pipelines apply when the most current data is needed. You will explore that parallelization and I/O buffers help mitigate bottlenecks. You will also learn how to describe data pipeline performance in terms of latency and throughput.

The key advantage of Apache Airflow's approach to representing data pipelines as DAGs is that they are expressed as code, which makes your data pipelines more maintainable, testable, and collaborative. Tasks, the nodes in a DAG, are created by implementing Airflow's built-in operators.​

In this module, you will learn about Apache Airflow having a rich UI that simplifies working with data pipelines. You will explore how to visualize your DAG in graph or tree mode. You will also learn about the key components of a DAG definition file, and you will learn that Airflow logs are saved into local file systems and then sent to cloud storage, search engines, and log analyzers.

Apache Kafka is a very popular open source event streaming pipeline. An event is a type of data that describes the entity’s observable state updates over time. Popular Kafka service providers include Confluent Cloud, IBM Event Stream, and Amazon MSK. Additionally, Kafka Streams API is a client library supporting you with data processing in event streaming pipelines.

In this module, you will learn that the core components of Kafka are brokers, topics, partitions, replications, producers, and consumers. You will explore two special types of processors in the Kafka Stream API stream-processing topology: The source processor and the sink processor. You will also learn about building event streaming pipelines using Kafka.