Insights: Smart Performance Analysis

Smart Performance Analysis is our advanced technique powered by machine learning and data science algorithms. We have built the first iteration of this technology into Insights and we are going to expand the functionality to make Insights more and more powerful.

In short, 
Smart Performance Analysis detects possible performance issues.


What is under the hood?

While the test is running, Smart Performance Analysis will run online algorithms that processes the main metrics:
  • VU ramping
  • Response time
  • Throughput
  • Number of requests being processed by target system
  • Input rate
  • Requests failures
Internally Smart Performance Analysis represents the results of a test run as a collection of time segments, the amount of time segments and their size depends on the system under test's behavior. In fact, Smart Analysis, like a human, can describe system behaviors during a test run. For example:
  • 0:00  - 2:04: VUs are growing, and the system under test can handle that growth well
  • 2:05: breakpoint, system reaches local throughput maximum, response time is increasing
  • 3:17 - 4:90: system scales throughput and handling load well again
  • ...
  • after 8:00: system overloaded with requests, high percent of failures detected
Segments that are identified as bad from a performance perspective are marked by Insights in the main chart. This gives you a hint of where to start when analyzing your test result.


Example

Smart Performance Analysis detects the time segment where system throughput reach its maximum, and a further increase in VUs causes faster growth of the response time. 


Feedback and Knowledge Base