kukulcan VS Apache Spark

Check it out! https://github.com/mmolimar/kukulcan

We all know how easy it is to overlook small parts of our code, especially when we have powerful tools like Apache Spark to handle the heavy lifting. Spark's core engine is great at optimizing our messy, complex code into a sleek, efficient physical plan. But here's the catch: Spark isn't flawless. It's on a journey to perfection, sure, but it still has its limits. And Spark is upfront about those limitations, listing them out in the documentation (sometimes as little notes).

One specific use case of the IaaS model is for deploying software that would have otherwise been bought as a SaaS. There are many such software from email servers to databases. You can choose to deploy MySQL in your infrastructure rather than buying from a MySQL SaaS provider. Other things you can deploy using the IaaS model include Mattermost for team collaboration, Apache Spark for data analytics, and SAP for Enterprise Resource Planning.

In this project, I'm exploring the Medallion Architecture which is a data design pattern that organizes data into different layers based on structure and/or quality. I'm creating a fictional scenario where a large enterprise that has several branches across the country. Each branch receives purchase orders from an app and deliver the goods to their customers. The enterprise wants to identify the branch that receives the most purchase requests and the branch that has the minimum average delivery time. To achieve that, I've used Apache Spark as a distributed compute engine and Apache Hadoop, in particular HDFS, as my data storage layer. Apache Spark ingest, processes, and stores the app's data on HDFS to be served to a custom dashboard app. You can find all about it, in this Github repo

In contrast, Databricks maintains internal forks of Spark, Delta Lake, and Unity Catalog, using the same names for both the open-source versions and the features specific to the Databricks platform. While they do provide separate documentation, online discussions often reflect confusion about how to use features in the open-source versions that only exist on the Databricks platform. This creates a "muddying of the waters" between what is open and what is proprietary. This isn't an issue if you are a Databricks user, but it can be quite confusing for those who want to use these tools outside of the Databricks ecosystem.

Recently I had to revisit the "JVM languages universe" again. Yes, language(s), plural! Java isn't the only language that uses the JVM. I previously used Scala, which is a JVM language, to use Apache Spark for Data Engineering workloads, but this is for another post 😉.

Consume data into third party software (then let Open Search or Apache Spark or Apache Pinot) for analysis/datascience, GIS systems (so you can put reports on a map) or any ticket management system

also, this knowledge applies to learning more about data engineering, as this field of software engineering relies heavily on the event-driven approach via tools like Spark, Flink, Kafka, etc.

Apache SeaTunnel is a data integration platform that offers the three pillars of data pipelines: sources, transforms, and sinks. It offers an abstract API over three possible engines: the Zeta engine from SeaTunnel or a wrapper around Apache Spark or Apache Flink. Be careful, as each engine comes with its own set of features.