DevOps Professional · Module 2: Reliability, capacity and resilience · Lab
Measure a Bottleneck and Refute a Wrong Scaling Proposal
50 min hands-on · Core
A local machine with a fixture service (app plus pool-limited dependency); a small load generator. Small load only; numbers are local findings.
Local guide: run the steps below on your own machine in order, then check the validation list.
Objectives
- Profile the fixture under small load to the real bottleneck
- Refute a given wrong scaling proposal with measured numbers
- Size the correct fix with headroom and state the limit
Step 1
Load small and profile
Drive the fixture with small controlled load while watching app CPU, pool exhaustion, queue depth and downstream latency. Name the bottleneck with quoted numbers.
Step 2
Refute in writing
Take the provided wrong proposal (double the app fleet) and refute it in a one-page brief: the numbers showing the constraint elsewhere, the cost of the wrong fix, the cost of the right one.
Step 3
Size the right fix
Size the correct fix from measured throughput with stated headroom. Add the honesty paragraph: what small-load measurement proves and what it does not.
How to confirm it worked
- Bottleneck named with quoted profile numbers
- One-page refutation brief with numbers and costs
- Right fix sized with headroom stated
- Honesty paragraph present