business / transaction throughput of 3000 users, we can simulate that
throughput by approximately decreasing the think_time in our script by
half. However, what are the potential performance issues we may miss
(e.g. connection configuration issues?) by using this load testing
method for doubling our user load?
Answer :
The method you describe will not accurately predict application
behavior at the 3000 concurrent user mark, which is more a factor of
resource utilization (memory per user session, number of sessions,
database connections, server threads, etc...) than the rate at which
transactions are submitted. You will likely see a slow down of
transactions using the method you describe, as the application queues
transactions at the increased rate. However if this increased
transaction rate is not representative of the rate in production, your
results will not be useful in any way.
If cost is an issue, there are different methods of increasing your
vuser license temporarily such as term licenses or vuser days. Both
methods will allow you to increase your vuser license at a lower cost
than purchasing them outright. I suggest talking to your local HP
Sales representative for more information.
behavior at the 3000 concurrent user mark, which is more a factor of
resource utilization (memory per user session, number of sessions,
database connections, server threads, etc...) than the rate at which
transactions are submitted. You will likely see a slow down of
transactions using the method you describe, as the application queues
transactions at the increased rate. However if this increased
transaction rate is not representative of the rate in production, your
results will not be useful in any way.
If cost is an issue, there are different methods of increasing your
vuser license temporarily such as term licenses or vuser days. Both
methods will allow you to increase your vuser license at a lower cost
than purchasing them outright. I suggest talking to your local HP
Sales representative for more information.
