d. Perform load testing on Cloud applications using simulated loads and describe the benefits of load testing
1. Use Load Impact or Blazemeter 3rd party load testing services in IBM Bluemix PaaS
IBM Bluemix PaaS has services to assist in characterizing how an applications responds under a simulated user load. These tests provide insights on performance and can also show if horizontal scaling is responding when the application is under stress.
2. Creating user scenarios
A virtual user scenario corresponds to a set of actions within the application and can be captured interactively or defined in a script language of the specific load testing tool. These should be based on a typical use-case of a user accessing the application.
3. Defining virtual user load for a test
The virtual user load run against an application can be focused on investigating different aspects of performance.
Measure response time for the application for a specific number of active users. For this type of test, the user workload is ramped to a steady state level and then held for a period time to gather response time statistics.
Determine peak scalability of an application. In this test the workload is increased in steps and held constant, or a series of test are performed each with a higher number of simulated users. A key metric such as the performance of a page providing user login or catalog display is monitored to ensure it does not exceed required levels.
4. Analyze results from load tests
When a load test completes, tools provide graphical and tabular output of information to review. Graphs from load testing will show observed metrics like response time and number of simulated users, graphed as a function of the time into the load test:
Green dots show the virtual user (VU) load, and blue dots the duration of time for a user scenario to complete. In the example shown, there is no strong correlation between the response time and the number of active users. This can be interpreted as the application response time not showing sensitivity to the quantity of simulated users for the duration of the test.
When reviewing results, it is critical to verify that all application responses are successful and not showing error codes. An application generating HTTP 404 or 500 errors may appear to show a quick response time, but it is not operating correctly.
1. Use Load Impact or Blazemeter 3rd party load testing services in IBM Bluemix PaaS
IBM Bluemix PaaS has services to assist in characterizing how an applications responds under a simulated user load. These tests provide insights on performance and can also show if horizontal scaling is responding when the application is under stress.
2. Creating user scenarios
A virtual user scenario corresponds to a set of actions within the application and can be captured interactively or defined in a script language of the specific load testing tool. These should be based on a typical use-case of a user accessing the application.
3. Defining virtual user load for a test
The virtual user load run against an application can be focused on investigating different aspects of performance.
Measure response time for the application for a specific number of active users. For this type of test, the user workload is ramped to a steady state level and then held for a period time to gather response time statistics.
Determine peak scalability of an application. In this test the workload is increased in steps and held constant, or a series of test are performed each with a higher number of simulated users. A key metric such as the performance of a page providing user login or catalog display is monitored to ensure it does not exceed required levels.
4. Analyze results from load tests
When a load test completes, tools provide graphical and tabular output of information to review. Graphs from load testing will show observed metrics like response time and number of simulated users, graphed as a function of the time into the load test:
Green dots show the virtual user (VU) load, and blue dots the duration of time for a user scenario to complete. In the example shown, there is no strong correlation between the response time and the number of active users. This can be interpreted as the application response time not showing sensitivity to the quantity of simulated users for the duration of the test.
When reviewing results, it is critical to verify that all application responses are successful and not showing error codes. An application generating HTTP 404 or 500 errors may appear to show a quick response time, but it is not operating correctly.
0 comments:
Post a Comment