Ok, you have run a test and found some issues with your application performance. Now, how can you fix them?
To help you solve this problem, we have created the predictive performance analytics tool.
The tool is based on a machine learning proprietary algorithm that processes information coming from our test results and data collected by a monitoring agent. The algorithm will then provide a visual interpretation of the components in your application stack that have shown the highest variations as the performance of your application were degrading.
The data will help identify the components that are the most likely bottlenecks of your application.
All you need is the following:
- have a New Relic agent installed in your application under test. If this is not your production environment, you should add the agents on this environment. Nouvola has a partner agreement with New Relic and you can get free extra agents on a partner standard plan.
- add your New Relic API key to your DiveCloud account information. Click your Account link, and select the API & Keys tab. Insert the New Relic API key as shown:
- run your test, making sure that it is created with the following criteria:
- Duration: min 10 minutes
- Traffic model: Linear Ramp. Flat traffic will not be accepted by the algorithm
- Number of users: this has to be at least 20 VU
- in your result table, you will see a DiveData column.
- If your tier includes Predictive Performance Analytics, you should be able to click on the "View" link, and this will show you the analytics page, as in the image below
Each sector maps to one or more New Relic categories.
|Server & Software||RubyVM, Middleware, HttpDispatcher, Memcache, Mongrel|
|Application Stack||View, Controller, ActiveRecord, ActiveMerchant, ActiveJob|
|Database||RemoteService, Database, Datastore|
|Browser||Browser, EndUser, Supportability, WebFrontend, Derived, AJAX|
The metrics from the monitoring tool are correlated to the metrics collected by DiveCloud during the test.
The yellow areas warns that some methods within the category have shown higher performance numbers as the overall application performance decreased, but they are within the threshold area.
The red ares represents values that are above the threshold.
If you hover on these areas, you can see the methods that show the highest correlation to the slowest page load response times.