Comparing Hadoop Distributions
You’ll find that the Hadoop ecosystem has many component parts, all of which exist as their own Apache projects. Because Hadoop has grown considerably, and faces some significant further changes, different versions of these open source community components might not be fully compatible with other components. This poses considerable difficulties for people looking to get an independent start with Hadoop by downloading and compiling projects directly from Apache.
Red Hat is, for many people, the model of how to successfully make money in the open source software market. What Red Hat has done is to take Linux (an open source operating system), bundle all its required components, build a simple installer, and provide paid support to any customers.
In the same way that Red Hat has provided a handy packaging for Linux, a number of companies have bundled Hadoop and some related technologies into their own Hadoop distributions. This list describes the more prominent ones:
Cloudera: Perhaps the best-known player in the field, Cloudera is able to claim Doug Cutting, Hadoop’s co-founder, as its chief architect. Cloudera is seen by many people as the market leader in the Hadoop space because it released the first commercial Hadoop distribution and it is a highly active contributor of code to the Hadoop ecosystem.
Cloudera Enterprise, a product positioned by Cloudera at the center of what it calls the “Enterprise Data Hub,” includes the Cloudera Distribution for Hadoop (CDH), an open-source-based distribution of Hadoop and its related projects as well as its proprietary Cloudera Manager. Also included is a technical support subscription for the core components of CDH.
Cloudera’s primary business model has long been based on its ability to leverage its popular CDH distribution and provide paid services and support. In the fall of 2013, Cloudera formally announced that it is focusing on adding proprietary value-added components on top of open source Hadoop to act as a differentiator.
Also, Cloudera has made it a common practice to accelerate the adoption of alpha- and beta-level open source code for the newer Hadoop releases. Its approach is to take components it deems to be mature and retrofit them into the existing production-ready open source libraries that are included in its distribution.
EMC: Pivotal HD, the Apache Hadoop distribution from EMC, natively integrates EMC’s massively parallel processing (MPP) database technology (formerly known as Greenplum, and now known as HAWQ) with Apache Hadoop. The result is a high-performance Hadoop distribution with true SQL processing for Hadoop. SQL-based queries and other business intelligence tools can be used to analyze data that is stored in HDFS.
Hortonworks: Another major player in the Hadoop market, Hortonworks has the largest number of committers and code contributors for the Hadoop ecosystem components. (Committers are the gatekeepers of Apache projects and have the power to approve code changes.)
Hortonworks is a spin-off from Yahoo!, which was the original corporate driver of the Hadoop project because it needed a large-scale platform to support its search engine business. Of all the Hadoop distribution vendors, Hortonworks is the most committed to the open source movement, based on the sheer volume of the development work it contributes to the community, and because all its development efforts are (eventually) folded into the open source codebase.
The Hortonworks business model is based on its ability to leverage its popular HDP distribution and provide paid services and support. However, it does not sell proprietary software. Rather, the company enthusiastically supports the idea of working within the open source community to develop solutions that address enterprise feature requirements (for example, faster query processing with Hive).
Hortonworks has forged a number of relationships with established companies in the data management industry: Teradata, Microsoft, Informatica, and SAS, for example. Though these companies don’t have their own, in-house Hadoop offerings, they collaborate with Hortonworks to provide integrated Hadoop solutions with their own product sets.
The Hortonworks Hadoop offering is the Hortonworks Data Platform (HDP), which includes Hadoop as well as related tooling and projects. Also unlike Cloudera, Hortonworks releases only HDP versions with production-level code from the open source community.
IBM: Big Blue offers a range of Hadoop offerings, with the focus around value added on top of the open source Hadoop stack.
Intel: The Intel Distribution for Apache Hadoop (Intel Distribution) provides distributed processing and data management for enterprise applications that analyze big data.
Key features include excellent performance with optimizations for Intel Xeon processors, Intel SSD storage, and Intel 10GbE networking; data security via encryption and decryption in HDFS, and role-based access control with cell-level granularity in HBase; improved Hive query performance; support for statistical analysis with a connector for R, the popular open source statistical package; and analytical graphics through Intel Graph Builder.
MapR: For a complete distribution for Apache Hadoop and related projects that’s independent of the Apache Software Foundation, look no further than MapR. Boasting no Java dependencies or reliance on the Linux file system, MapR is being promoted as the only Hadoop distribution that provides full data protection, no single points of failure, and significant ease-of-use advantages.
Three MapR editions are available: M3, M5, and M7. The M3 Edition is free and available for unlimited production use; MapR M5 is an intermediate-level subscription software offering; and MapR M7 is a complete distribution for Apache Hadoop and HBase that includes Pig, Hive, Sqoop, and much more.