Load one resource deliberately and prove which one the tools blame
Task
Create load on one resource at a time in the disposable guest, and record what each looks like in Task Manager and Resource Monitor. The exercise establishes the signatures, so that a real machine's symptoms can be recognised rather than guessed at.
Steps
- Record the guest's idle figures in
lab/perf/baseline.csvwith headerresource,idle_value,unitfor processor, memory, disk active time and network. - Load the processor -- a compression job over a large directory works -- and record the same four figures in
lab/perf/cpu.csvusing the same header. - Load the disk -- copy a multi-gigabyte file, or read the whole disk -- and record the four figures in
lab/perf/disk.csv, adding disk queue length and response time in milliseconds. - Reduce the guest's memory until it pages, reproduce ordinary work, and record hard faults per second in
lab/perf/memory.csvalongside the four figures. - Write
lab/perf/signatures.mdgiving, for each of the four resources, the one figure that identifies it and the user-visible symptom that goes with it -- including why a memory shortage presents as a disk problem.
Verify
for f in baseline cpu disk memory; do awk -F, -v f=$f 'NR>1 && NF>=3 {n++} END {print f": "n" reading(s)"}' lab/perf/$f.csv; done
grep -Eic 'hard fault' lab/perf/memory.csv lab/perf/signatures.md
grep -Eic 'queue|response' lab/perf/disk.csv
grep -Ec '^[-*#0-9]' lab/perf/signatures.md
Four readings in each file, hard faults recorded, and the signatures note explaining the memory-looks-like-disk case. If the four resource figures are identical across all four files, the load did not land where you thought.
Notes
The memory run is the one worth doing carefully. It is the case that produces the wrong recommendation most often, because the disk is pinned and the obvious conclusion is a new disk.
This is an independent study companion for CompTIA A+ Core 2 220-1202 and is not produced by or endorsed by CompTIA.