When you use the Blaze engine to run mappings, Blaze uses a Grid Manager at run time to allot tasks to various nodes in a Hadoop cluster. The Grid Manager aids in resource allocation.
You can use the same Hadoop cluster to stage your test environment and establish a production environment. To control resource use on the cluster, you can establish a separate Blaze instance for testing and another for production.
Each instance requires a separate Grid Manager. You create an additional Grid Manager by performing a series of steps to create separate infrastructure for each Blaze instance, including a unique namespace and a Hadoop connection for each Blaze instance to use.
Each Blaze instance also requires a separate Hadoop connection.
The following image shows how a separate Data Integration Service on the domain creates a separate Grid Manager on the cluster:
The image shows how separate Data Integration Services use separate Blaze instances. Each instance uses a separate Grid Manager to communicate with the cluster resource manager to balance resources.
Perform the following steps to set up separate Blaze instances:
Step 1. Prepare the Hadoop cluster for the Blaze engine.
Step 2. Configure Data Integration Service properties.