OpenStack Data processing service¶

The Data processing service for OpenStack (sahara) aims to provide users with a simple means to provision data processing (Hadoop, Spark) clusters by specifying several parameters like Hadoop version, cluster topology, node hardware details and a few more. After a user fills in all the parameters, the Data processing service deploys the cluster in a few minutes. Sahara also provides a means to scale already provisioned clusters by adding or removing worker nodes on demand.

The solution addresses the following use cases:

• Fast provisioning of Hadoop clusters on OpenStack for development and QA.
• Utilization of unused compute power from general purpose OpenStack IaaS cloud.
• Analytics-as-a-Service for ad-hoc or bursty analytic workloads.

Key features are:

• Designed as an OpenStack component.
• Managed through REST API with UI available as part of OpenStack dashboard.
• Support for different Hadoop distributions:
  • Pluggable system of Hadoop installation engines.
  • Integration with vendor specific management tools, such as Apache Ambari or Cloudera Management Console.
• Predefined templates of Hadoop configurations with the ability to modify parameters.
• User-friendly UI for ad-hoc analytics queries based on Hive or Pig.