Databases Reference
In-Depth Information
•
The ramp-up period
: This parameter controls the period within which
the load testing tool will create a maximum number of concurrent threads
configured in the script. For example, if a portal is expected to have 2000
concurrent users during pick time, all 2000 users will not start accessing the
portal at the same time. They will start accessing the portal one by one and
at a certain point in time all 2000 users will be accessing the Portal. So, here
the time interval within which the number concurrent users reach from zero
to 2000 is the ramp-up period. If it is configured incorrectly, it may give
unrealistic load test results.
•
The think time
: This parameter controls the wait time between two
operations of the load testing script. Let's take an example of a user
accessing a portal. He/she will first log in to the portal, then access page
X
and then access page
Y
, and so on. The user will take some time after
logging in and before accessing page
X
. This time is called the
think time
.
This parameter should be configured properly to generate a realistic load
on the system. If it is configured to a very low value, the portal will be
flooded with a number of requests. And it will not be a realistic scenario.
Apart from the aforementioned parameters, there are some of the Liferay-specific
best practices that should be followed while writing test scripts. They are as follows:
• Login is one of the costliest operations in Liferay Portal. It is recommended
to write realistic test scenarios where the user first signs in and then performs
various operations and finally signs out.
• JMeter provides a way to read CSV- or XLS-based inputs to perform any
operations. It is recommended to create a number of dummy users in Liferay
Portal and perform a load test with those dummy users. If load tests are
performed with a single user, it will not generate a realistic load.
• JMeter instances should be configured on the machine that is in the same
network as Portal. If JMeter instances are not part of same network as Portal,
the load test may give inaccurate results because of varying Internet speed.
The purpose of load testing is not to determine the bottleneck of a network
between the user and server. Hence, it is recommended to run load tests
from the same network.
