You can use a grid to increase session and workflow performance. A grid is an alias assigned to a group of nodes that allows you to automate the distribution of workflows and sessions across nodes.
A Load Balancer distributes tasks to nodes without overloading any node.
When you use a grid, the Integration Service distributes workflow tasks and session threads across multiple nodes. A Load Balancer distributes tasks to nodes without overloading any node. Running workflows and sessions on the nodes of a grid provides the following performance gains:
Balances the Integration Service workload.
Processes concurrent sessions faster.
Processes partitions faster.
The Integration Service requires CPU resources for parsing input data and formatting the output data. A grid can improve performance when you have a performance bottleneck in the extract and load steps of a session.
A grid can improve performance when memory or temporary storage is a performance bottleneck. When a PowerCenter mapping contains a transformation that has cache memory, deploying adequate memory and separate disk storage for each cache instance improves performance.
Running a session on a grid can improve throughput because the grid provides more resources to run the session. Performance improves when you run a few sessions on the grid at a time. Running a session on a grid is more efficient than running a workflow over a grid if the number of concurrent session partitions is less than the number of nodes.
When you run multiple sessions on a grid, session subtasks share node resources with subtasks of other concurrent sessions. Running a session on a grid requires coordination between processes running on different nodes. For some mappings, running a session on a grid requires additional overhead to move data from one node to another node. In addition to loading the memory and CPU resources on each node, running multiple sessions on a grid adds to network traffic.
When you run a workflow on a grid, the Integration Service loads memory and CPU resources on nodes without requiring coordination between the nodes.