How to Break into Data Engineering : 2024

The One & Only Way: Mastering the Fundamentals

Richard Warepam
11 min readDec 11, 2023


Photo by Christina @ on Unsplash

Hey there, future data engineering stars! Let’s unpack a common puzzle you might be facing: navigating the maze of data engineering tools and technologies.

Are you wondering whether to dive into Spark or Snowflake?

Pondering over Python or Scala?

Trying to choose between Airflow or Argo, Flink or Spark Streaming, AWS or GCP, Superset or Tableau?

I get it, it’s confusing!

Here’s the key takeaway:

Focus on the fundamentals, not just the flashy tools.

Understanding the Core, Not Just the Tools


Fundamental #1: Distributed Computing vs. Spark/Snowflake:

Grasping the basics of distributed computing is far more crucial than picking between Spark or Snowflake.

It’s about how to manage and process massive datasets, not just which tool to use.

What is Distributed Computing?

Imagine you have a big puzzle, and instead of trying to solve it all by yourself, you get help from your friends. Each friend takes a piece of the puzzle and works on it. Once everyone is done, you put all the pieces back together.

That’s the essence of distributed computing — breaking down big tasks into smaller ones, distributing them across multiple computers (or nodes), and then combining the results.

Key Concepts in Distributed Computing

1. Nodes and Clusters:

  • Nodes: These are individual computers in a distributed system.
  • Clusters: A group of nodes working together is called a cluster.

Think of nodes as workers, and a cluster as a team.

2. Parallel Processing:

  • This is like multitasking. Instead of one computer doing all the work (sequentially), multiple tasks are done at the same time by different computers.

3. Fault Tolerance and Redundancy:

  • In a distributed system, if one node fails, others can take over. This is fault tolerance.
  • Redundancy means having extra nodes or copies of data as a backup.

4. Scalability:

  • This is about handling more work by adding more nodes. If your data grows, your system can grow too.

5. Load Balancing:

  • This is about evenly distributing work among nodes so that no single node is overwhelmed.

Why is Distributed Computing Essential in Data Engineering?

1. Handling Big Data:

  • Distributed computing makes it possible to process and analyze vast amounts of data (Big Data) efficiently.

2. Speed and Efficiency:

  • By dividing tasks, you can process data faster than a single machine could.

3. Reliability:

  • If one node fails, others can pick up the slack, making the system more reliable.

4. Cost-Effective:

  • It’s often more economical to use a cluster of lower-cost computers than a single high-end server.

Real-World Applications in Data Engineering

1. Hadoop and Spark:

  • These are popular frameworks that use distributed computing for data processing.

2. Data Storage and Retrieval:

  • Systems like Google’s Bigtable and Amazon’s DynamoDB use distributed computing for efficient data storage and retrieval.

Fundamental #2: Batch or Streaming Over Flink/Spark Streaming:

Knowing when to use batch processing and when to opt for streaming is more significant than the choice between Flink and Spark Streaming.

It’s the strategy of data handling that counts.

Batch Processing: The Methodical Chef

Imagine you’re a chef preparing a big feast. You gather all your ingredients (data) over time and then start cooking (processing).

That’s batch processing. You process data in large, collected chunks at a scheduled time. It’s not about speed; it’s about thoroughness.

Use Batch Processing When:

  1. You’re Not in a Hurry: If real-time analysis isn’t crucial, batch processing is your go-to. It’s perfect for end-of-day reports or monthly summaries.
  2. Accuracy Over Speed: Batch processing can handle complex calculations that require data from multiple sources. It’s like making sure you have every ingredient before you start cooking.
  3. Cost-Effectiveness: Often, batch processing can be more cost-effective, especially when dealing with large volumes of data that don’t need immediate action.

Streaming: The Quick-Response Cook

Now, picture a short-order cook, whipping up dishes on the fly as orders come in.

That’s streaming. It’s about processing data in real-time as it arrives. Speed and responsiveness are the essences here.

Opt for Streaming When:

  1. Immediate Action Required: If your data needs to be acted upon instantly (like fraud detection in financial transactions), streaming is the way to go.
  2. Continuous Input: When data is continuously generated (think social media feeds or sensor data), streaming ensures you’re always up-to-date.
  3. Handling Large, Unpredictable Workloads: Streaming can better handle unpredictable influxes of data, adapting to the flow as it happens.

Choosing the Right Approach

Think of your data objectives.

If you need to analyze vast amounts of accumulated data for comprehensive insights, batch processing is your ally. On the other hand, if you’re dealing with time-sensitive data that requires immediate attention, streaming steps into the spotlight.

Fundamental #3: Programming Principles vs. Scala/Python:

The debate between Scala and Python takes a back seat to understanding functional and object-oriented programming.

These programming foundations are what make you versatile in the data engineering landscape.

Functional Programming: The Precision Tool

Functional programming (FP) is like a precision tool in your kit.

It emphasizes the use of pure functions — think of them as predictable machines where the same inputs always give the same outputs, without altering any state outside the function. This makes your code more predictable and easier to debug.

Key Aspects of Functional Programming:

  1. Immutability: Once created, data structures aren’t changed. This is like using ingredients in a recipe without altering their original form.
  2. Statelessness: FP avoids shared states and relies on immutable data, leading to fewer bugs, especially in systems where data flows across different parts of the application.
  3. Higher-Order Functions: These are functions that take other functions as parameters or return them as results. It’s like having a master recipe that can modify and use other recipes in various ways.

Why Use FP in Data Engineering?

  1. Predictability and Easy Debugging: With pure functions, you can easily predict outcomes and debug issues.
  2. Concurrency: FP’s immutable data and statelessness make it a great fit for concurrent operations, crucial in processing large data sets.

Object-Oriented Programming: The Structured Organizer

Object-Oriented Programming (OOP) is like a structured organizer for your code.

It’s all about creating objects — collections of data and methods that represent something in your application. Imagine each object as a mini-robot programmed to perform specific tasks and store certain information.

Core Concepts of Object-Oriented Programming:

  1. Encapsulation: This involves bundling the data and methods that work on the data within one unit, or object. It’s like a capsule containing everything you need for a specific task.
  2. Inheritance: Objects can inherit properties and behaviors from other objects, creating a hierarchy. This is akin to a family tree, where traits are passed down.
  3. Polymorphism: This allows objects to be treated as instances of their parent class rather than their actual class. Think of it as different tools sharing the same interface but performing their unique actions.

Benefits of OOP in Data Engineering:

  1. Modularity: Breaking down complex systems into manageable objects makes it easier to understand, develop, and maintain.
  2. Reusability: Objects and classes can be reused across different parts of the application, saving time and effort.
  3. Clear Structure: OOP provides a clear structure for your code, making it easier to navigate and scale.

Bringing It Together in Data Engineering

In data engineering, both functional and object-oriented programming have their place.

Use FP for tasks that require high levels of predictability and concurrency.
OOP, with its emphasis on modularity and reusability, is excellent for structuring large systems and applications.

Fundamental #4: Cloud Deployment Basics vs. AWS/GCP:

Grasping how pipelines are deployed in the cloud is more vital than choosing between AWS or GCP.

It’s about understanding cloud infrastructure and how to best use it.

Understanding Cloud Infrastructure

Cloud infrastructure refers to the virtual technology and resources that make up cloud computing.

Think of it as a big pool of resources (like servers, storage, databases, networking, software) that you can access over the internet.

Key Components of Cloud Infrastructure:

  1. Compute Power: This includes servers and virtual machines. Just like your laptop or desktop, but these are located in remote data centers.
  2. Storage: Cloud storage allows you to store and access data over the internet, offering scalability and security.
  3. Networking: This involves the virtual network resources that let you connect your cloud services and users securely.
  4. Databases: Cloud databases provide flexibility and easy scalability, essential for managing large sets of data efficiently.

How to Best Use Cloud Infrastructure

  1. Scalability: The cloud can easily scale up or down based on your needs. Use it for projects that have varying workloads or need to grow over time.
  2. Cost-Efficiency: Pay for what you use. The cloud is great for projects where you want to control costs — no need to invest in physical hardware.
  3. Flexibility and Accessibility: Access your data and applications from anywhere. This is perfect for remote teams or if you need to provide services across different geographies.
  4. Disaster Recovery and Backup: Use the cloud for backing up data and disaster recovery. It’s much easier and more reliable than traditional methods.
  5. Experimentation and Innovation: The cloud provides a perfect environment to experiment with new ideas, as you can set up and dismantle systems quickly without heavy investment.

Tips for Navigating the Cloud

  1. Understand Different Service Models: Know the difference between IaaS (Infrastructure as a Service), PaaS (Platform as a Service), and SaaS (Software as a Service). Each has its unique use case.
  2. Security and Compliance: Always prioritize security. Understand the shared responsibility model — while cloud providers secure the infrastructure, you’re responsible for securing your data.
  3. Monitor and Optimize Costs: Keep an eye on your usage and costs. Use tools provided by cloud services to optimize and reduce expenses.
  4. Stay Up-to-Date with Trends: Cloud technology evolves rapidly. Stay informed about new services and features.

Conclusion: The Sky’s the Limit!

Cloud infrastructure is a game-changer in data engineering. It offers flexibility, scalability, and a host of services that can be tailored to your specific needs.

Whether you’re a small startup or a large enterprise, understanding and utilizing cloud infrastructure effectively can propel your data projects to new heights. So, embrace the cloud, experiment, and watch your data capabilities soar!

Fundamental #5: Comprehensible Data Visualization Over Tools:

The ability to create clear and effective data visualizations is more important than the tool you use, be it Tableau or Superset.

It’s about communicating insights, not just about the software.

The Art of Communicating Data Insights

1. Understand Your Audience: Before you dive into the data, ask yourself who your audience is. Are they technical experts or laypeople? Executives or clients? Understanding their perspective and what they value most helps tailor your message for maximum impact.

2. Translate Data into Stories: Humans are wired for stories. Transform your data insights into narratives. Rather than just presenting numbers, explain what these numbers mean in the real world. How do they affect your audience? What story do the numbers tell about trends, challenges, or opportunities?

3. Simplify, but Don’t Oversimplify: Your goal is to make complex data understandable, but not at the cost of accuracy. Avoid jargon, but ensure that the core message of your data is not lost in an attempt to oversimplify.

4. Use Visuals Wisely: A well-crafted chart or graph can convey what paragraphs of text cannot. Use visuals to support your narrative, but be mindful of not overcrowding them. Each visual element should serve a clear purpose.

5. Focus on Key Takeaways: What do you want your audience to remember or act upon? Highlight these points. Whether it’s a significant trend, a surprising insight, or an urgent call to action, make sure these elements stand out.

6. Invite Engagement: Encourage questions and discussions. This not only ensures that your audience has understood the insights but also fosters a collaborative environment. Be open to feedback and ready to delve deeper into aspects that interest your audience.

7. Practice Empathy: Put yourself in your audience’s shoes. How would you react to the data if you were them? This empathetic approach can guide you in choosing the right tone, level of detail, and presentation style.

Tools Are Just the Beginning

Remember, tools and software are just the starting point. They’re like instruments in an orchestra — it’s how you use them to create a symphony that matters.

Your ability to communicate insights effectively can turn data into action and decisions, bridging the gap between numbers and real-world impact.

Fundamental #6: Pipeline Schedule and SLAs Over Airflow/Argo:

Understanding how to manage pipeline schedules and Service Level Agreements (SLAs) is more crucial than the specific tool, whether it’s Airflow or Argo.

It’s about reliability and efficiency in your data processes.

Understanding Pipeline Scheduling

1. What is Pipeline Scheduling?

  • Think of a pipeline schedule as a timetable for your data processes. It dictates when and how often your data tasks should run.

Just like a train schedule ensures trains run on time, pipeline scheduling ensures your data tasks are executed efficiently and reliably.

2. Why is it Important?

  • Proper scheduling prevents bottlenecks. It ensures that data flows smoothly through your systems, and resources are optimally used.

It’s like coordinating traffic in a busy city to avoid jams.

3. Key Considerations:

- Frequency: Decide how often your pipelines need to run. This could be hourly, daily, or weekly, depending on your data needs.

- Dependencies: Some tasks depend on the completion of others. Understanding these dependencies is crucial to avoid errors and delays.

- Resource Allocation: Ensure that your schedule aligns with the available computational resources to prevent overloading systems.

Service Level Agreements (SLAs)

1. What are SLAs?

- SLAs are like promises you make to your users about the performance and reliability of your data services. They set expectations for things like data processing times and system uptime.

2. Importance of SLAs in Data Engineering:

- SLAs build trust. By meeting your SLAs, you assure users that they can rely on your data systems. It’s about maintaining a consistent level of service.

3. Managing SLAs:

- Clear Metrics: Define clear, measurable targets (e.g., processing time should not exceed 2 hours).

- Monitoring: Regularly monitor your systems to ensure you’re meeting SLAs.

- Communication: Be transparent with your users, especially if you anticipate SLA breaches.

Beyond Tools: Focusing on Process and Efficiency

While tools like Airflow or Argo are important, they are just a means to an end.

The real skill lies in how you manage these tools to maintain efficient and reliable data pipelines.

- Automation: Use tools to automate scheduling and monitoring tasks, but don’t forget to periodically review and adjust these automations based on changing data needs.

- Best Practices: Stay informed about industry best practices in pipeline management and SLA fulfillment.

- Flexibility: Be prepared to adapt your strategies as your data environment and business needs evolve.

Conclusion: The Heart of Reliable Data Operations

In summary, understanding and effectively managing pipeline schedules and SLAs are foundational to the success of any data engineering operation.

It’s not just about which tool you use, but how you use it to ensure your data processes are reliable, efficient, and trustworthy. Embrace these aspects, and you’ll set yourself apart as a data engineer who not just handles tools, but truly powers data-driven success.

Adapting to Change:

The Power of Fundamentals Grasping these fundamental concepts arms you with adaptability.

Technologies in data engineering are always evolving.

If you have a solid understanding of the underlying principles, adapting to new tools and technologies becomes a breeze.

Remember, in the fast-paced world of data engineering, it’s not about the latest tools or software. It’s about the timeless principles that underpin them.

By focusing on these core concepts, you’re setting yourself up for long-term success, ready to navigate and adapt to the ever-changing technological landscape.

So, dive into the fundamentals. They are your true compass in the exciting journey of data engineering!

Currently, I am working on a book that teaches all these fundamentals in detail.

Stay Connected by subscribing on the newsletter, to get early offers when I publish the book.



Richard Warepam

Data Scientist & Writer | Google Certified Data Analyst | As a Mentor - Writes on Data Science and AI | My eBooks: