Master Linux Performance: Essential CPU, Memory, Disk I/O & Network Tools

Performance Tool Map Overview

The diagram maps Linux kernel subsystems to the most common performance analysis tools. The tools are grouped by the resource they observe: CPU, memory, disk I/O, and network.

CPU Performance Tools

Key CPU metrics include overall utilization, per‑core usage, context‑switch rate, interrupt rate, and CPU cache‑miss ratios. Typical command‑line utilities are: top / htop – real‑time CPU usage and process list. mpstat (from sysstat) – per‑CPU statistics and interrupt counts. perf – hardware‑level profiling, cache‑miss analysis, and flame graphs. sar – historical CPU utilization and context‑switch trends. turbostat – Intel‑specific power and frequency metrics.

These tools help correlate high CPU load with specific processes, identify excessive context switches, and locate cache‑related bottlenecks.

Memory Performance Tools

Memory‑related metrics focus on total RAM consumption, per‑process resident set size (RSS), page‑fault rate, swap activity, and kernel slab usage. Common utilities include: free / vmstat – overall memory and swap statistics. smem – proportional set size (PSS) and shared memory breakdown. pmap – memory map of a single process. slabtop – live view of kernel slab allocator. ps_mem – detailed per‑process memory accounting. pcstat – aggregates per‑process memory, swap, and OOM‑killer data. Source code: https://github.com/tobert/pcstat

By examining these metrics you can detect memory leaks, excessive swapping, or inefficient kernel memory caches.

Disk I/O Performance Tools

Important I/O metrics are filesystem utilization, block‑device throughput, average request latency, and queue depth. Representative tools are: iostat (sysstat) – device‑level I/O rates and utilization. iotop – real‑time per‑process I/O bandwidth. blktrace / btrace – low‑level block‑layer tracing. dstat – combined CPU, memory, and I/O statistics. sar -b – historical I/O activity.

These utilities allow you to pinpoint hot devices, identify processes causing high I/O, and measure latency spikes.

Network Performance Tools

Network metrics include throughput, packet loss, round‑trip latency, connection count, and per‑protocol statistics. Typical tools are: ifstat / nload – interface bandwidth usage. netstat / ss – socket state and connection counts. iperf3 – synthetic throughput and latency testing. tcpdump / wireshark – packet capture and protocol analysis. bpftrace – custom kernel‑level network tracing.

The accompanying images illustrate the metric categories and the mapping of each tool to the relevant Linux subsystem.