Fast is slow, slow is fast – rethinking Our Data Engineering Process | Blog | bol.com

Rethinking Our Information Engineering Course of

If you’re beginning a brand new staff, you are usually confronted with an important dilemma: Do you stick together with your present method of working to stand up and working shortly, promising your self to do the refactoring later? Or do you’re taking the time to rethink your method from the bottom up?

We encountered this dilemma in April 2023 after we launched a brand new knowledge science staff targeted on forecasting inside bol’s capability steering product staff. Inside the staff, we frequently joked that “there’s nothing as everlasting as a short lived resolution,” as a result of rushed implementations usually result in long-term complications.These fast fixes are inclined to grow to be everlasting as fixing them later requires vital effort, and there are all the time extra fast points demanding consideration. This time, we have been decided to do issues correctly from the beginning.

Recognising the potential pitfalls of sticking to our established method of working, we determined to rethink our method. Initially we noticed a chance to leverage our present know-how stack. Nonetheless, it shortly grew to become clear that our processes, structure, and total method wanted an overhaul.

To navigate this transition successfully, we recognised the significance of laying a robust groundwork earlier than diving into fast options. Our focus was not simply on fast wins however on guaranteeing that our knowledge engineering practices might sustainably assist our knowledge science staff’s long-term objectives and that we might ramp up successfully. This strategic method allowed us to deal with underlying points and create a extra resilient and scalable infrastructure. As we shifted our consideration from speedy implementation to constructing a stable basis, we might higher leverage our know-how stack and optimize our processes for future success.

We adopted the mantra of “Quick is sluggish, sluggish is quick.”: dashing into options with out addressing underlying points can hinder long-term progress. So, we prioritised constructing a stable basis for our knowledge engineering practices, benefiting our knowledge science workflows.

Our Journey: Rethinking and Restructuring

Within the following sections, I’m going to take you alongside our journey of rethinking and restructuring our knowledge engineering processes. We’ll discover how we:

• Leveraged Apache Airflow to orchestrate and handle our knowledge workflows, simplifying complicated processes and guaranteeing clean operations.
• Realized from previous experiences to determine and get rid of inefficiencies and redundancies that have been holding us again.
• Adopted a layered method to knowledge engineering, which streamlined our operations and considerably enhanced our skill to iterate shortly.
• Embraced monotasking in our workflows, bettering readability, maintainability, and reusability of our processes.
• Aligned our code construction with our knowledge construction, making a extra cohesive and environment friendly system that mirrored the way in which our knowledge flows.

By the top of this journey, you’ll see how our dedication to doing issues the correct method from the beginning has set us up for long-term success. Whether or not you’re going through comparable challenges or seeking to refine your individual knowledge engineering practices, I hope our experiences and insights will present invaluable classes and inspiration.

Float

We rely closely on Apache Airflow for job orchestration. In Airflow, workflows are represented as Directed Acyclic Graphs (DAGs), with steps progressing in a single path. When explaining Airflow to non-technical stakeholders, we frequently use the analogy of cooking recipes.