Just another comment on this.
The load test must be appropriate for what is going to go through the system.
I have seen both sides of the coin:
* Throw a gazillion Create Ticket Requests at SDM in a very short period of time, and it fails.
> Yes, it probably will. But will the site ever come even remotely close to having that type of throughput? Almost certainly not.
* Throw a gazillion "things" and SDM and at it works.
> Hoorah! But is this the type of load that you will get? For example, SDM can happily process 10,000 - 100,000 simple Web Services calls per second on an unloaded box. But this is not what kills a system typically - it will be a simple bad Web Services call itself that does the damage, or something completely unrelated to the item under load test, such as inbound email.
So it is important to make sure that the volume and type of load are appropriate for what will be on the environment.
Thanks, Kyle_R.