How to Monitor Linux System Performance: CPU, Memory, Disk I/O & Network

Linux System Monitoring

1. CPU

Load average : Use uptime to view the 1‑, 5‑, and 15‑minute load averages. A server is considered overloaded when the load exceeds the number of logical CPUs (e.g., ≥ 3 for a single‑core system, ≥ 12 for a four‑core system).

Logical CPU count : grep -c 'processor' /proc/cpuinfo returns the number of logical processors.

Top CPU consumers : Run top and press P to sort by %CPU. Example output:

PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
720 root 20 0 473464 22060 19128 S 1.7 0.2 119:55.14 sunloginclient
3590 emqx 20 0 3483576 201492 74484 S 1.3 2.0 39.37 beam.smp
984 root 20 0 2754328 50696 5192 S 1.0 0.0 32.96 taosd

Note that on a multi‑core system a process can exceed 100 % CPU (e.g., up to 400 % on a 4‑core CPU).

Per‑CPU utilization : In top, press 1 to display a line for each CPU, e.g.:

%Cpu0 : 0.0 us, 3.4 sy, 0.0 ni, 96.6 id, 0.0 wa, 0.0 hi, 0.0 si, 0.0 st
%Cpu1 : 0.0 us, 0.0 sy, 0.0 ni,100.0 id, 0.0 wa, 0.0 hi, 0.0 si, 0.0 st

High load together with high per‑CPU usage and a run‑queue > 3 indicates a busy CPU.

2. Memory

Total and used memory : free -h provides a human‑readable summary.

total   used   free  shared  buff/cache  available
Mem:          9.7Gi  2.1Gi  5.4Gi   49Mi     2.3Gi       7.3Gi
Swap:          0B    0B     0B

Key fields: total, used, free, shared, buff/cache, available. Swap is used only when physical RAM is exhausted.

Processes using most memory : In top, press M to sort by %MEM and e to toggle units (M, G, T). Example output:

PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
1377 mysql 20 0 4.8g 0.8g 0.0g S 0.7 7.8 67:01.24 mysqld
977 root 20 0 7.7g 0.3g 0.0g S 0.0 3.4 32:34.83 java
3590 emqx 20 0 3.3g 0.2g 0.1g S 0.7 2.0 64:48.04 beam.smp

3. Disk I/O

Device read/write activity : iostat -dkp shows per‑device statistics (tps, kB_read/s, kB_wrtn/s, total kB read/write). Example:

Device    tps  kB_read/s  kB_wrtn/s  kB_read  kB_wrtn
sda       0.64   16.17      58.90   487458  1775844
sda6      0.63   15.79      58.90   476129  1775748

Processes causing the most I/O : iotop -o lists only processes performing I/O. Example:

Total DISK READ : 5.79 M/s | Total DISK WRITE : 2.28 M/s
TID PRIO USER DISK READ DISK WRITE SWAPIN IO% COMMAND
17238 be/4 tom 5.79 M/s 0.00 B/s 0.00 % 64.19 % find /

If CPU and memory appear idle while the system feels sluggish, the bottleneck is often disk I/O; upgrading to enterprise‑grade SSDs can alleviate it.

4. Network

Overall bandwidth : nload -u h displays incoming and outgoing traffic in human‑readable units.

Device eth0 [10.0.0.18]
Incoming: Curr: 558.33 kBit/s  Avg: 297.12 kBit/s  Min: 1.39 kBit/s  Max: 894.77 kBit/s  Ttl: 58.23 MByte
Outgoing: Curr: 16.16 kBit/s  Avg: 15.37 kBit/s  Min: 4.52 kBit/s  Max: 32.62 kBit/s  Ttl: 766.79 kByte

Per‑process bandwidth usage : sudo nethogs lists processes with their sent/received rates.

PID USER PROGRAM DEV SENT RECEIVED
1733 root wget eth0 13.352 4317.425 KB/sec
1605 tom sshd: tom@pts/0 eth0 1.025 0.070 KB/sec