Load Generators
A load generator is a machine that runs JMeter and produces load. Atlas copies your test plan to each one and starts them together.
Most teams share a pool that is already set up — in that case you can skip this section entirely and just pick machines when you run a test.
Read on if you need to add machines of your own.
Two kinds
| Servers | Fixed machines that exist whether or not a test is running. Cloud VMs Atlas can power on and off, or your own hardware. |
| Clusters | Elastic capacity. Units are created when a test starts and destroyed when it ends. |
Both appear under Fleet in the sidebar, on separate tabs.
Why run your own?
The usual reason is that your application isn't reachable from the internet. A load generator inside your own network can reach it, and only needs an outbound connection to Atlas — no inbound access, no firewall changes, no VPN.
The other reason is geography: generating load from where your users actually are gives you a more honest number.
Power state
Machines Atlas can control are powered on when a test needs them and switched off when it ends, so you don't pay for idle capacity. This is why a switched-off machine can still be selected on the Run Test tab.
Machines Atlas can't power — your own hardware — are assumed to be always on and are never powered off.
Status
The Fleet page shows two independent things per machine:
- whether the machine is on or off
- whether its connection to Atlas is live
A machine can be on but not connected, which usually means the agent isn't running. See Troubleshooting.