Hadoop offered on Windows Azure as a PaaS, Windows Azure HDInsight provides a copy of the Apache Hadoop data processing platform and associated tools. HDInsight will use the Hortonworks Data Platform (HDP), which is the flagship Hadoop distribution offered by Hortonworks.
Microsoft is not the first to offer Hadoop as a cloud service. Amazon Web Services offers Hadoop and Rackspace plans to offer the HDP as a service soon. IBM’s SoftLayer hosting service also announced this week it will offer the ability quickly spin up copies of Cloudera’s Hadoop distribution across multiple bare-metal servers.
But Microsoft is hoping to carve out a competitive advantage in this growing space by simplifying the process of deploying, then using Hadoop.
For instance, the service works well with Microsoft data analysis tools. Users of the Excel spreadsheet can deploy a feature of the spreadsheet, called power BI, to ingest, then analyze and visualize data delivered by Hadoop MapReduce.
Microsoft first announced plans for this service in 2011. Originally, the company intended to develop its own version of Hadoop, along with Hortonworks, which would be configured to run on Windows Server.
“We realized pretty quickly there would be a level of redundancy between what Hortonworks would provide and what we were going to provide. So we decided to align on one core offering, HDP on Windows,” Kelly said. Microsoft contributed over 16,000 lines of code to Apache Hadoop, an open source project, and related software.
Windows Azure HDInsight will run a stock version of HDP, allowing users to seamlessly move their workloads between Azure and other non-Azure HDP deployments, Kelly said. After Hortonworks releases its next version of HDP, version 2.0 is due next month, Azure’s version of HDP will be automatically upgraded.
The company has actually been running Windows Azure HDInsight in full production mode, at least for select clients, for a couple of months, Kelly said.