|
Description
How to dynamically change the ramp-up rate of a LoadRunner Scenario to minimise ramp-up time.
Solution
It is often a time-consuming process to get to a pre-determined number of users as a starting point for a load test, or to determine the maximum number of users that a large system can support before it fails. This is especially the case in complex web systems that are sensitive to the ramp-up rate of new users connecting and using the system if a slow ramp-up rate is used for the entire scenario. If you only have a few days in a test lab, every minute of test execution time counts, and the following suggestion can save hours by rapidly ramping up the first users in a test, while the system is able to support a high ramp-up rate.
One way of dynamically managing the ramp-up rate of a test is to set up multiple groups (or sets of groups), each of which have different ramp up rates, all starting at the beginning of the scenario.
eg
group_5ps where 5 users are started each second
group_2ps where 2 users are started each second
group_1ps where 1 user is started each second
group_1p2s where 1 users is started each 2 seconds
Setup each group with sufficient numbers of users and start the test. When the response times start to degrade significantly, remove all un-started users from the group that is ramping up at the fastest rate. In this example, stopping new users from being added from the first group slows the ramp-up rate from 8.5 per second to 3.5 per second. The ramp-up rate can again be slowed to 1.5 per second at the point where 3.5 per second is excessive. Note the numbers of users in each group, for use in subsequent tests.
If there were 500 users in each of the above groups, then it would take 1000 seconds (< 17 minutes) to ramp up to 2000 users, compared with 4000 seconds (1 hour, 7 minutes) at the slowest ramp-up rate.
Note that the slowest ramp-up rate should not be lower than the minimum ramp-up rate expected to be experienced in production.
Paul McLean
www.loadtest.com.au |
|