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

1. CPU

Introduce the CPU as the most important component; use top to observe performance.

1.1 top command

Run top and press 1 to see per‑core details.

Key metrics include:

us user‑mode CPU percentage.

sy kernel‑mode CPU percentage.

ni nice‑priority CPU percentage.

wa I/O wait percentage.

hi hardware interrupt percentage.

si software interrupt percentage.

st stolen time for virtualized environments.

id idle CPU percentage.

Load average shows the queue length; values depend on CPU core count.

1.2 Load average

Explain load values for single‑, dual‑, and quad‑core systems.

1.3 vmstat

Use vmstat to see CPU busy state; important columns are b (blocked processes), cs (context switches), and si/so (swap activity).

$ vmstat 1
procs ---------memory---------- ---swap-- -----io---- -system-- ------cpu-----
 r  b swpd   free   buff  cache   si   so    bi    bo   in   cs us sy id wa st
34  0    0 200889792  73708 591828    0    0     0     5    6   10 96  1  3  0  0
...

2. Memory

2.1 Observation commands

Use top to view VIRT, RES, and SHR columns; RES shows actual memory used by a process.

2.2 CPU cache

CPU caches reduce the speed gap between CPU and memory; false sharing occurs when multiple threads modify variables on the same cache line.

Cache line size can be read with:

cat /sys/devices/system/cpu/cpu0/cache/index0/coherency_line_size

Cache sizes for each level can be queried:

cat /sys/devices/system/cpu/cpu0/cache/index1/size   # 32K
cat /sys/devices/system/cpu/cpu0/cache/index2/size   # 256K
cat /sys/devices/system/cpu/cpu0/cache/index3/size   # 20480K

2.3 HugePage

HugePages increase page size (default 4 KB) to reduce TLB pressure on large memories.

2.4 Pre‑touch

JVM option -XX:+AlwaysPreTouch forces memory allocation at startup, improving runtime performance.

3. I/O

3.1 Observation commands

Disk I/O is often the bottleneck; iostat shows %util, avgqu‑sz, await, svctm, etc.

%util near 100 % means the device is saturated.

await should be < 5 ms; > 10 ms indicates problems.

3.2 Zero copy

Zero‑copy (e.g., sendfile) eliminates copying between kernel and user space, reducing CPU load.

4. Network

Use sar to monitor network traffic and TCP statistics.

$ sar -n DEV 1
...

5. Conclusion

These metrics give a quick view of Linux performance, but deeper analysis may require advanced tools such as eBPF‑based BCC.