Performance & tracing
Find CPU, memory, storage, and scheduling bottlenecks with evidence.
Begin with a concrete symptom and observation window. Check utilization, saturation, and errors before reaching for deeper tracing. A busy CPU, high load average, or large page cache alone does not identify the cause of a slowdown.
Use the USE method to choose measurements, procfs and ps to understand process evidence, sysstat for intervals and history, and perf or BPF when ordinary counters leave an unanswered question.
No matching references. Try a broader term or clear the filters.
Find CPU, memory, storage, and scheduling bottlenecks with evidence.
-
The USE method
Organize an investigation around utilization, saturation, and errors so measurements answer a specific resource question.
Author's performance methodology · Author guide · Selection reviewed -
procfs: process and system counters
Interpret the interfaces under /proc and follow the detailed pages for the counters a monitoring tool reports.
Linux kernel userspace interface · Project documentation · Selection reviewed -
ps: select process evidence
Select process fields and formatting deliberately when investigating states, threads, elapsed time, or ownership.
Linux procps-ng · Project documentation · Selection reviewed -
sysstat: interval and historical statistics
Choose among sar, iostat, mpstat, and pidstat for interval and historical measurements, and follow project documentation for each tool.
sar, iostat, mpstat and pidstat · Project documentation · Selection reviewed -
perf profiling examples
Choose a counting, sampling, or tracing approach and understand what a profile does and does not measure.
Author's Linux perf examples · Author guide · Selection reviewed -
BPF kernel documentation
Read the kernel's BPF concepts and interfaces when investigating tracing capabilities, verifier behavior, or program types.
Current upstream kernel documentation · Project documentation · Selection reviewed
How this library is selected
References are selected for a concrete operational task, with preference for project and distribution documentation. Annotations describe when to use a source; version labels make its scope explicit.
Selection reviewed means the destination, subject, and version scope were checked. It is not a claim that every upstream command has been tested. Original guides identify their own test environment. Platform filters also include references that apply across distributions.