It puts load on server which you are targeting or which is under load test.
Using no. of load generators depends on load your server can take. Suppose your server can take 10000 users/transactions then that amount of load is not possible to generate on single load generator with average resources. (core 2 duo cpu, 2-4 gb ram etc.)
Generally average machine/load generator can generate 500-1000 virtual users load. This will hit max. capacity of client without any other applications on it.
Here question is how do you know 10000 is capacity of my server?
Answer is you should have an rough idea in advance of your server. some sample test runs and mathematical calculations using Little's Law you can find out theoretical limit of your server (which is generally higher than practical limit)
Based on those calculations, setup no. of load generators and perform load testing of server in steps.