Current Behaviour : Responding time graphs of each URI generated by Performance Plugin includes responses with errors. This current behaviour is described in the resolved issue JENKINS-26988.
From our point of view, this behaviour does not give a good view of the quality of the service and leads to confusion. This behaviour should be customisable by the user
=> [Ignores or includes errors in response time]

Example :
-> The 2 regular criterias for a Service Level Agreement are Response Time and Error Rate.
-> Errors can either be timeout of the 3rd party (leads to extra long response time) or Front end global error (leads to extra short response time).
=> In case of Error Rate increase between 2 builds, the responding time graphs are highly modified and show an unstability of response times, even if the application have super stable response time (success cases).

As attached,

• an example of the current behaviour (responding time including error response)
  => At first view, a high instability of response time !!
• the same result source (.csv) but for not successfull responses we have suffixed the URI by _error
  => High stability of response time (120sec) but an high rate of errors

Best regards,

Quentin