Mobile Monitoring: Filter out network errors by default so mean latency is correct

Background and context

The mean latency information in mobile monitoring is currently very skewed by that some calls show extremely high latencies. This is due to that Instana measures some corner cases that is not really interesting from performance point of view (but might be for other purposes).

These are

• Various network communication errors such as timeouts causing Instana to not be able to capture a correct HTTP response. These are reported with status code '-1' and often have huge latenices along with them in the range of hours or even days.

• Cases where the user closes the app or change to another app before the mobile application has finished communication with the backend. These traces are mapped to the view 'Background' by the Instana SDK and also this category contains calls with very high latencies, often several minutes and/or hours.

From performance tracking point of view it is not interesting to include these erroneous cases for which we as app developers anyway don't have any control over. We're only interested in measuring performance of successful communication between the frontend and the backend.

Current workaround in analytics and custom dashboards

A workaround we use currently is to filter out these two categories by the Instana filters

• HTTP Status code != -1

• View != Background

Proposed solution

A possible way for Instana to handle this in a way so that mean latency is correct by default when viewing a set of calls from mobile monitoring is to make it configurable whether these two categories should be a part of the monitored dataset, perhaps in a similar way as in the backend monitoring where it's configurable whether to show synthetics calls or not, where the default is that these calls are excluded from the dataset.

Example:

[x] Show network errors and timeouts

[x] Show calls from Background view