1. Mental Model: Where Time Goes

Every request a client makes crosses several layers, and load testing is really about finding which layer breaks first as concurrency rises. Draw this stack mentally before touching a tool:

Client                                                        Server
  |--DNS lookup-------------------------------------------------|
  |--TCP handshake (SYN, SYN-ACK, ACK)---------------------------|
  |--TLS handshake (ClientHello...Finished)----------------------|
  |--Request sent (serialize + write to socket)-------------------|
  |          [network transit: propagation + queueing delay]     |
  |                                              [server accept()]|
  |                                       [thread/goroutine pickup]|
  |                                         [business logic + I/O]|
  |                                          [DB / cache / RPC]   |
  |                                       [serialize response]    |
  |          [network transit back]                               |
  |--Response read + deserialize----------------------------------|

A load test result of "p99 = 800ms" is meaningless until you know which segment dominates. The rest of this guide gives you the vocabulary and tools to decompose that number.