Latency could create huge issues, like  bottle necks, slowing down of applications. 

To keep that from happening, we can use timeouts for our requests.

As a general rule, bigger timeouts are implemented at the edge or gateways for example and shorter timeouts are implemented as traffic gets closer to the running application.

Below is a simple example of how istio implements timeouts in its virtualService that takes care of routing requests to our service - service1.ns1.svc.cluster.local - and its endpoints - pods/VMs -: