Accelerating Metadata Discovery with Web-Based Applications

The industry is seeing a growing interest in enhancing the productivity of data specialists through the use of metadata. For instance, tools developed in this area include Airbnb's Dataportal, Uber's Databook, Netflix's Metacat, Lyft's Amundsen, and most recently, Google's Data Catalog.

Many companies are continuously working on scaling their infrastructure to keep pace with their expanding big data ecosystems. As the volume and variety of data grow, it becomes even more time-consuming for data specialists and engineers to discover available data assets, understand their lineage, and take appropriate actions based on the data insights. In this article, we’ll discuss accelerating the metadata discovery process using web-based applications.

Why is Collecting Metadata Important

Data becomes valuable only when accompanied by metadata that provides context. However, metadata needs efficient discovery capabilities to be useful; without it, metadata remains unexploited. Below I’ve listed the benefits metadata collection provides.

Regulatory Compliance

• Metadata discovery streamlines compliance processes like GDPR, allowing for quick identification and management of sensitive data.
  Identifying Changes

• It simplifies the locating and updating of data fields across systems processes, reducing the workload and potential errors during updates.

Data Migrations

• During BI system migrations, metadata discovery aids in deciding which data to transfer, ensuring an efficient and cost-effective migration.

Automated Data Lineage

• Metadata discovery is crucial for automated data lineage, enabling efficient root cause and impact analysis by mapping the data journey.

Improved Data Governance

• By capturing detailed information about data sources, usage, and ownership, we can facilitate robust data governance. This means that data policies will be effectively enforced and the company can gradually come to more standardised data handling practices.

Enhanced Data Quality

• Apart from the above, metadata helps in identifying data anomalies, redundancies, and inconsistencies, this allows us to introduce timely corrections and results in higher data quality.

Support for Data Analysis and Business Intelligence

• If analysts have detailed metadata they can understand the data transformations and usage. This is essential for a more accurate and meaningful analysis.

Operational Agility

• Metadata collection contributes to operational agility. With its help, we can clearly see the data flows and dependencies, meaning we can quickly respond to changing business needs.

Data Catalog Creation

• It powers the creation and updates of data catalogues, which organise and define data assets, fostering consistency and trust across the company’s data landscape.

In other words, we need efficient metadata discovery to transform data collection into actionable insights, which will let us enhance productivity and compliance while reducing errors and inefficiencies.

In the context of data catalogues and metadata management, metadata is typically classified into three main categories: Technical Metadata, Business Metadata, and Operational Metadata.

Technical Metadata includes details about the structure of data assets such as schema definitions, table structures, column types, and data relationships. It also encompasses information about ETL processes, data lineage (tracing data origins and transformations), and storage details like database locations and file paths.

Business Metadata focuses on providing clear definitions of data elements, specifying business rules and constraints applied to data, and identifying data owners, stewards, and subject matter experts responsible for data assets. It also includes contextual information that explains the purpose, usage, and business relevance of data.

Operational Metadata involves capturing usage statistics to understand how data assets are accessed and used, performance metrics related to data systems, audit logs tracking changes and user activities for compliance, and data quality metrics assessing data accuracy, completeness, consistency, and timeliness.

What is a Data Catalogue and How to Build It

A data catalogue is a comprehensive tool that organises all the data assets within a company's information landscape. Each entry in the data catalogue includes definitions, descriptions, ratings, and details about data owners and stewards. This organisation makes it simple to search for and identify the data needed for various purposes.

Self-service BI Users

Self-service BI users, who bridge the gap between IT and business, benefit significantly from data catalogues. They can discover and evaluate data easily, finding relevant data assets to build effective reports. Additionally, they can engage with subject matter experts within the catalogue framework, ensuring questions are answered and documented for future reference.

Business Users

Business users gain independence in creating value from data by:

• Locating all relevant reports across multiple tools.
• Checking details about datasets used in reports, understanding their scope, and collaborating directly within the catalogue to preserve and share knowledge.
• Identifying existing reports to avoid redundant work. General Data Users Any data user can use the catalogue to answer questions such as:
• Where to find specific data.
• The relevance and importance of data.
• Data origin and responsible parties.
• Proper data usage and current users.

How to Build a Data Catalogue

Below I have listed steps to create a data catalogue that can be further used for above discussed purposes:

1. Determine which metadata to record, including technical and business descriptions, asset types, responsible parties, and tags.

2. Choose between programming your framework or using existing tools, focusing on essential functionalities like searchability, filtration, data asset evaluation, and automation.

3. Use an automated catalogue solution to survey your BI landscape and pull out existing metadata from various tools, populating the data catalogue efficiently.

4. A human review is necessary to enrich definitions and ensure that data consumers can independently use the data.

5. Integrate a data lineage tool to gain visibility into the data’s journey through the BI landscape. This integration provides quick access to end-to-end lineage, helping data stewards trace data flows, answer accuracy questions, and build trust in the data.

6. Regularly update and maintain the data catalogue to ensure it stays accurate and relevant.

Top 5 Tools to Accelerate Data Catalogue Creation

Apart from the manual creation of the data catalogue, there is a range of tools that can facilitate the process, making it dramatically faster and more efficient.

• Apache Atlas It is an open-source framework for comprehensive metadata management and governance within various data ecosystems.

Apache Atlas provides an extensive REST API for programmatic interactions to help with metadata management, governance, and integration with other systems. With this tool, companies can conduct studies much quicker and with solid, reliable data. Moreover, the platform efficiently handles any IT overflow and ensures seamless distribution and management. As massive data searches and report generation are always challenging to manage, Apache Atlas is a great tool for data-driven enterprises.

• Apache Airflow Apache Airflow is a tool for orchestrating complex computational workflows and data processing pipelines. It offers extensive cross-platform integration, seamlessly connecting with almost all necessary technologies. One of its standout features is the ability to create Directed Acyclic Graphs (DAGs) using Python, making workflow creation intuitive and flexible. This capability is central to many processes, particularly for integrating workflows with reports, enhancing efficiency and automation across various tasks.

Airflow automates the data workflows that populate and update the data catalogue managed by Apache Atlas, including tasks like data ingestion, transformation, and periodic metadata refreshes.

• Docker/Kubernetes with Helm
  Docker containers standardise the packaging of an application and its dependencies. Kubernetes manages the deployment, scaling, and operations of containerised applications. Helm charts can help to simplify Kubernetes application configurations and management.
  These tools ensure that all components of the data catalogue (including Apache Atlas and Apache Airflow) are deployed consistently and can be dynamically scaled across different environments.

• MongoDB
  This is a flexible NoSQL database. It is known for its high performance and scalability.
  In a data catalogue context, MongoDB is used for storing secondary, non-metadata information such as user profiles, access logs, and interaction data. This helps in managing user access and enhancing personalization features within the catalogue.

• Grafana
  Grafana is a visualisation tool used to create dynamic dashboards. It is always useful to have visual insights into the data catalogue’s operations and user interactions. It helps to monitor the performance of data workflows and maintain the health of the entire system.

Wrapping Up

Undoubtedly, the mastery comes with practice. Yet I would also recommend going through some theory as you can find really valuable insights there. Here are the books I can advise:

Data Management for Researchers: Organize, Maintain and Share Your Data for Research Success by Kristin Briney. Although aimed at researchers, this book covers the principles of data management, including metadata, which apply to any field. Another one is a great start for beginners, Metadata by Jeffrey Pomerantz. The book provides a thorough introduction to metadata, its purposes, and its applications. It covers the basics of metadata and its importance in data discovery and management.