TLDR: A fast, free, and open-source Modern Data Stack (MDS) can now be fully deployed on your laptop or to a single machine using the combination of DuckDB, Meltano, dbt, and Apache Superset.
There is a large volume of literature (1, 2, 3) about scaling data pipelines. “Use Kafka! Build a lake house! Don’t build a lake house, use Snowflake! Don’t use Snowflake, use XYZ!” However, with advances in hardware and the rapid maturation of data software, there is a simpler approach. This article will light up the path to highly performant single node analytics with an MDS-in-a-box open source stack: Meltano, DuckDB, dbt, & Apache Superset on Windows using Windows Subsystem for Linux (WSL). There are many options within the MDS, so if you are using another stack to build an MDS-in-a-box, please share it with the community on the DuckDB Twitter, GitHub, or Discord, or the dbt slack! Or just stop by for a friendly debate about our choice of tools!
What is the Modern Data Stack, and why use it? The MDS can mean many things (see examples here and a historical perspective here), but fundamentally it is a return to using SQL for data transformations by combining multiple best-in-class software tools to form a stack. A typical stack would include (at least!) a tool to extract data from sources and load it into a data warehouse, dbt to transform and analyze that data in the warehouse, and a business intelligence tool. The MDS leverages the accessibility of SQL in combination with software development best practices like Git to enable analysts to scale their impact across their companies.
Why build a bundled Modern Data Stack on a single machine, rather than on multiple machines and on a data warehouse? There are many advantages!
If you contribute to an open-source community or provide a product within the Modern Data Stack, there is an additional benefit!
One key component of the MDS is the unlimited scalability of compute. How does that align with the MDS-in-a-box approach? Today, cloud computing instances can vertically scale significantly more than in the past (for example, 224 cores and 24 TB of RAM on AWS!). Laptops are more powerful than ever. Now that new OLAP tools like DuckDB can take better advantage of that compute, horizontal scaling is no longer necessary for many analyses! Also, this MDS-in-a-box can be duplicated with ease to as many boxes as needed if partitioned by data subject area. So, while infinite compute is sacrificed, significant scale is still easily achievable.
Due to this tradeoff, this approach is more of an “Open Source Analytics Stack in a box” than a traditional MDS. It sacrifices infinite scale for significant simplification and the other benefits above.
Given that the NBA season is starting soon, a monte carlo type simulation of the season is both topical and well-suited for analytical SQL. This is a particularly great scenario to test the limits of DuckDB because it only requires simple inputs and easily scales out to massive numbers of records. This entire project is held in a GitHub repo, which you can find here: https://www.github.com/matsonj/nba-monte-carlo.
The detailed steps to build the project can be found in the repo, but the high-level steps will be repeated here. As a note, Windows Subsystem for Linux (WSL) was chosen to support Apache Superset, but the other components of this stack can run directly on any operating system. Thankfully, using Linux on Windows has become very straightforward.
In this example, Meltano pulls together multiple bits and pieces to allow the pipeline to be run with a single statement. The first part is the tap (extractor) which is ‘tap-spreadsheets-anywhere’. This tap allows us to get flat data files from various sources. It should be noted that DuckDB can consume directly from flat files (locally and over the network), or SQLite and PostgreSQL databases.
