10 Seconds is all they get!
The average human attention span in 2000 was 12.5 seconds, in 2015 it dropped to a mere 8.25 seconds (less then a goldfish who zones out at 9 seconds).
To stay 'in the zone', I challenge my teams to keep the duration of their full test runs below 10 seconds.
But I still want them to be using actual azure services.
Yes, the real, latency inducing, remote services.
Using local emulators or, even worse, substitution technologies, like swapping out Azure Service Bus for RabbitMQ, is not a good idea at all.
Nothing will behave in the exact same way as the service itself: there are differences in response codes, connection limits, service quota, throttling behavior and not the least there is the impact of latency itself.
So how can you solve these contradictory requirements?
You can't rely solely on unit tests as you wouldn't have covered the service behavior.
You can't rely solely on integration tests, as the latency per test would quickly increase the test run duration.
The secret is to have the right mix and introduce additional practices that ensure the unit tests will also adhere to the same restrictions that the integration tests adhere to.
I will not go into full detail at this point, when you want to learn all the details come to Cloudbrew on December 7th & 8th, where I will uncover all the details.