Skip to content
Close
How Big Data Challenges Shaped Data Storage and Computing
12 min read

How Big Data Challenges Shaped Data Storage and Computing

The term “big data” has become a buzzword in the tech industry over the last few years. As data generation continues to accelerate, traditional storage and computing systems are compelled to evolve, driving the development of innovative technologies that have fundamentally transformed how we process and manage information. In this blog, we will explore the impact of big data challenges on the evolution of data storage and computing.

The Early Days: COBOL and Relational Databases

The beginning of business data processing was marked by the invention of COBOL—Common Business-Oriented Language. COBOL was designed to handle large volumes of business data and served its purpose well during the early stages of digital evolution. However, as data volumes grew, the limitations of COBOL became apparent. It required extensive coding to perform complex queries, and COBOL's file-based processing could be inefficient, especially when working with large datasets or complex data relationships.

Subsequently, relational databases, introduced in the 1970s, marked a significant advancement in data management. Relational Database Management Systems (RDBMS) provide three main features:

  1. SQL – Structured Query Language
  2. Scripting languages (PL/SQL and Transact-SQL)
  3. Interfaces for other programming languages such as JDBC and ODBC

SQL became the standard for managing structured data, offering a powerful way to query and manipulate information stored in tables. Using these technologies, data processing applications were developed. However, as new data formats like JSON, XML, PDFs, and JPEGs emerged, data began to be categorized into three main types:

  • Structured data: standardized form (Rows and Columns)
  • Semi-structured data: formats like JSON, XML
  • Unstructured data: formats like PDFs, JPEGs


RDBMS could not store and process these unstructured data types.

The Rise of Big Data Challenges

By the early 2000s, the term "big data" began to gain traction as organizations faced challenges posed by the massive amounts of generated data. The three Vs of big data—Volume, Velocity, and Variety—highlighted the limitations of existing data storage and computing solutions.

Google's Solution: GFS and MapReduce

Google was the first company to develop a viable solution to big data problems and establish a commercially successful business—the Google Search Engine. However, the company faced several key challenges:

  • Data Collection and Ingestion: Discovering and crawling web pages to capture content and metadata
  • Data Storage and Management: Storing vast amounts of data efficiently
  • Data Processing and Transformation: Applying algorithms, such as PageRank, to index and organize data
  • Data Access and Retrieval: Providing quick and accurate search results

To address these challenges, Google created GFS (Google File System) and MapReduce, two technologies designed to process data on large clusters. Google detailed this solution in a series of influential white papers in 2003 and 2004, which captured the attention of the developer community and led to the creation of Hadoop.

Hadoop: Reliable, Scalable, Distributed Computing

Hadoop introduced a distributed data processing platform with several key components:

  • YARN (Yet Another Resource Negotiator): A cluster management system that efficiently allocates resources across nodes, enabling better utilization of computational power and scaling large workloads across a distributed environment.
  • HDFS (Hadoop Distributed File System): A distributed storage system that divides large datasets into smaller blocks, stored across multiple nodes. This allows for horizontal scalability, fault tolerance, and quick access to massive volumes of data.
  • MapReduce: A programming model that simplifies large-scale data processing by breaking tasks into smaller, parallelizable jobs. It enables distributed computation across clusters, making it possible to handle vast datasets that would be infeasible with traditional systems.

These components allowed Hadoop to solve big data problems by providing a scalable, fault-tolerant framework for both storage and computation. However, limitations eventually arose, particularly in terms of speed due to reliance on disk I/O and the complexity of the MapReduce programming model.

The Rise of Lakehouse Architecture

To overcome the limitations of Hadoop and address the Velocity challenge, a new data architecture emerged, combining the strengths of data warehouses and data lakes. This architecture is known as the Lakehouse. Technology advancements that enabled the data lakehouse architecture include:
  • Metadata layers for data lakes
  • New query engine designs providing high-performance SQL execution on data lakes
  • Optimized access for data science and machine learning tools

Apache Spark: The Heart of the Lakehouse

Apache Spark has become a central component of the lakehouse architecture by overcoming several challenges faced by Hadoop:

  • Speed: Spark offers in-memory data processing, which drastically reduces the time required for data operations compared to Hadoop's disk-based system.
  • Ease of Use: Spark provides a more user-friendly API compared to MapReduce, simplifying the development and management of complex data processing tasks.
  • Flexibility: Spark supports various data formats and sources, making it a versatile tool for handling diverse data types.
  • Metadata and Governance: Spark integrates more seamlessly with the metadata and governance layers of modern data architectures, enhancing data management, tracking, and compliance—areas where Hadoop traditionally struggled.

These improvements have made Spark the preferred choice for many organizations building a lakehouse architecture.

The challenges posed by big data have catalyzed innovation in data storage and computing. From the early days of COBOL and relational databases to the rise of Hadoop and cloud computing, the evolution of these technologies has been driven by the need to manage ever-growing volumes of data. As we move into the future, the intersection of big data, AI, and real-time processing will continue to shape the landscape of data storage and computing, pushing the boundaries of what is possible.