Last-mile latency is the delay packets incur between the nearest internet backbone and an end-user device. It is one of the two main components of network latency, alongside geographic latency. The final distance may be short enough to imply less than a millisecond of travel time, yet last-mile infrastructure can add tens of milliseconds. section 3.2.1
Separate the two components before changing a deployment. A long round-trip to a distant region is a geographic problem; moving computation and data closer to users can avoid that trip. It does not follow that the entire end-to-end delay will disappear: packets still have to reach the user’s device. section 3.2.1 section 3.2.2
A nearby edge is not the whole path
A user in Tokyo can reach a nearby edge location instead of making a round-trip to a centralized deployment in Virginia. That change targets the long geographic leg. The connection from the network to the user’s device remains part of the latency the user experiences. section 3.2.2 section 3.2.1
Use placement decisions to narrow down the delay:
- Establish a network-latency baseline in the production environment; local tests may hide a costly remote round-trip. section 3.2.1
- Check whether users are reaching a distant central region or a nearby edge location before attributing their delay to the last mile. section 3.2.2 section 3.2.1
- Assess internode placement first: network communication can outweigh application-level latency improvements. section 3.1
