DevOps Automation

About the assignment of "Executing Agents" to "User Groups" in a Workload Schedule and its effects on the control of reading data from a Dataset, is there a limitation or recommendation for the cases below?

1. More than one Agent defined in a "User Group"

2. Many "User Groups" using the same agent

it's ok to use more than 1 agent for a group, in fact if 1 agent can't support the load of a specific group, then we would need an additional agent or even more. Usually we know we need another agent if we get a message similar to "the workbench has lost communication with the agent on host xxxx'

Same for Many "User Groups" using the same agent.

Starting with RPT 9.5, datasets aren't allocated on a machine (segmented) basis.

Dataset row requests are allocated on a first come first serve basis in the order any given VU requests a row, assuming the access is sequential.

Please refer to this technote for all the open modes and how they behave:

"Understanding the Various Open Modes for Datasets in Rational Performance Tester"

https://www.ibm.com/support/pages/node/1076889

You will find that the right answer varies and has a "try it and see" component. There is no universal standard for how much "system resource" any given test/schedule will consume. You may find you have one test that doesn't handle much data that runs just fine on a single agent. At the same time you may have another test that opens a lot of connections, processes a lot of data, and requires load distribution across 2-N agents.

User groups and agents are not as strictly tied together as your questions implies. A User Group is a set of emulated users who will be doing the same work - test(s) - across the same set of 0-N agents.

A schedule can have multiple User Groups. Those User Groups can share agents or have distinct agents. It's up to you and your ultimate configuration will be driven by what you find works for your specific schedule. And it may not be the same for your next schedule.