• Axiom Community Resources
  • Protolabs
  • Rydoo
  • Axesor
  • IBS
  • Docfinance
  • everis
  • enhesa
  • Enterprise Greece
  • Amina
  • Amina

Extract, Transform, and Load Big Data with Apache Hadoop

Convert Big Data into Magic Data - Big Data Analytics by Intel

Intel / 21 February 2020



Over the last few years, organizations across public and private sectors have made a strategic decision to turn big data into competitive advantage. The challenge of extracting value from big data is similar in many ways to the age-old problem of distilling business intelligence from transactional data. At the heart of this challenge is the process used to extract data from multiple sources, transform it to fit your analytical needs, and load it into a data warehouse for subsequent analysis, a process known as “Extract, Transform & Load” (ETL). The nature of big data requires that the infrastructure for this process can scale cost-effectively. Apache Hadoop* has emerged as the de facto standard for managing big data. This whitepaper examines some of the platform hardware and software considerations in using Hadoop for ETL.


– We plan to publish other white papers that show how a platform based on Apache Hadoop can be extended to support interactive queries and real-time predictive analytics. When complete, these white papers will be available at https://hadoop.intel.com.

The ETL Bottleneck in Big Data Analytics


Big Data refers to the large amounts, at least terabytes, of poly-structured data that flows continuously through and around organizations, including video, text, sensor logs, and trans-actional records. The business benefits of analyzing this data can be significant. According to a recent study by the MIT Sloan School of Management, organizations that use analytics are twice as likely to be top performers in their industry as those that don’t.1


Business analysts at a large company such as Intel, for example, with its global market and complex supply chain, have long sought insight into customer demand by analyzing far-flung data points culled from market information and business transactions. Increasingly, the data we need is embedded in economic reports, discussion forums, news sites, social networks, weather reports, wikis, tweets, and blogs, as well as transactions. By analyzing all the data available, decision-makers can better assess competitive threats, anticipate changes in customer behavior, strengthen supply chains, improve the effectiveness of marketing campaigns, and enhance business continuity.

Many of these benefits are not new to organizations that have mature processes for incorporating business intelligence (BI) and analytics into their decision-making. However, most organizations have yet to take full advantage of new technologies for handling big data. Put simply, the cost of the technologies needed to store and analyze large volumes of diverse data has dropped, thanks to open source software running on industry-standard hardware. The cost has dropped so much, in fact, that the key strategic question is no longer what data is relevant, but rather how to extract the most value from all the available data.


Rapidly ingesting, storing, and processing big data requires a cost-effective infrastructure that can scale with the amount of data and the scope of analysis. Most organizations with traditional data platforms—typically relational database management systems (RDBMS) coupled to enterprise data warehouses (EDW) using ETL tools—find that their legacy infrastructure is either technically incapable or financially impractical for storing and analyzing big data.


A traditional ETL process extracts data from multiple sources, then cleanses, formats, and loads it into a data warehouse for analysis. When the source data sets are large, fast, and unstructured, traditional ETL can become the bottleneck, because it is too complex to develop, too expensive to operate, and takes too long to execute.

By most accounts, 80 percent of the development effort in a big data project goes into data integration and only 20 percent goes toward data analysis. Furthermore, a traditional EDW platform can cost upwards of USD 60K per terabyte. Analyzing one petabyte—the amount of data Google processes in 1 hour—would cost USD 60M. Clearly “more of the same” is not a big data strategy that any CIO can afford. So, enter Apache Hadoop.

Apache Hadoop for Big Data


When Yahoo, Google, Facebook, and other companies extended their services to web-scale, the amount of data they collected routinely from user interactions online would have overwhelmed the capabilities of traditional IT architectures. So they built their own. In the interest of advancing the development of core infrastructure components rapidly, they published papers and released code for many of the components into open source. Of these components, Apache Hadoop has rapidly emerged as the de facto standard for managing large volumes of unstructured data.


Apache Hadoop is an open source distributed software platform for storing and processing data. Written in Java, it runs on a cluster of industry-standard servers configured with direct-attached storage. Using Hadoop, you can store petabytes of data reliably on tens of thousands of servers while scaling performance cost-effectively by merely adding inexpensive nodes to the cluster.


Central to the scalability of Apache Hadoop is the distributed processing framework known as MapReduce (Figure 1). MapReduce helps programmers solve data-parallel problems for which the data set can be sub-divided into small parts and processed independently. MapReduce is an important advance because it allows ordinary developers, not just those skilled in high-performance computing, to use parallel programming constructs without worrying about the complex details of intra-cluster communication, task monitoring, and failure handling. MapReduce simplifies all that.


The system splits the input data-set into multiple chunks, each of which is assigned a map task that can process the data in parallel. Each map task reads the input as a set of (key, value) pairs and produces a transformed set of (key, value) pairs as the output. The framework shuffles and sorts outputs of the map tasks, sending the intermediate (key, value) pairs to the reduce tasks, which group them into final results. MapReduce uses JobTracker and TaskTracker mechanisms to schedule tasks, monitor them, and restart any that fail.


The Apache Hadoop platform also includes the Hadoop Distributed File System (HDFS), which is designed for scalability and fault-tolerance. HDFS stores large files by dividing them into blocks (usually 64 or 128 MB) and replicating the blocks on three or more servers. HDFS provides APIs for MapReduce applications to read and write data in parallel. Capacity and performance can be scaled by adding Data Nodes, and a single NameNode mechanism manages data placement and monitors server availability. HDFS clusters in production use today reliably hold petabytes of data on thousands of nodes.


Click Here! to Check out the full Pdf.

Axiom Members

Follow us

Our tweets:

Axiom Groupe Save the date! Don't miss our next EHS event the 19th and 20 November 2020 ! Link: https://t.co/smu2VZ6DFE… https://t.co/DWlx12KmSD
Axiom Groupe @AxiomGroupe is very proud to present one of our speakers at our EHS towards 4.0 event, Bas Vennix, the director of… https://t.co/9r7bSt68xR

Upcoming Events