End-to-end performance
The time a person perceives is the sum of network, application, transfer and browser.
How network, application, transfer and browser compose the experience, and why performance needs to be measured by stage and under real conditions.
Time budget
Four parts of the same experience
-
Network Find and connect
DNS, route and negotiation happen before the response.
-
Application Process and start delivery
Code, queries and dependencies form server time.
-
Transfer Deliver the resources
File weight, count and priority influence when content reaches the browser.
-
Browser Download, compose and respond
Visible content and interaction depend on the device and resources.
One overall measurement hides different causes
Time to first byte includes more than application processing: name resolution, connection and negotiation may participate. Transfer and rendering still affect what a person perceives after it.
Without separating the stages, optimization misses the target: a bigger machine does not fix an excessive browser resource, and shrinking an image does not fix a slow query.
Lab and field data answer different questions
Controlled tests make changes easier to reproduce and compare. Field data shows devices, networks and interactions that do not fit one fixed scenario.
A single measurement does not represent the audience: percentiles and distribution separate a consistent improvement from occasional variation.
The bottleneck changes as the application evolves
A new feature usually adds queries, bytes or browser work. The performance budget shows which stage grew since the last measurement.
Additional capacity solves the problem when the measured limit is capacity. In every other case it only postpones the diagnosis.