# iostat — Report CPU and Disk I/O Statistics > Practical guide to iostat — measure CPU usage and disk I/O, spot bottlenecks via await, %util and aqu-sz. Part of the sysstat package. Source: https://www.jpkc.com/db/en/cheatsheets/shell-system/iostat/ iostat gives you a single-glance view of how hard your CPU and storage devices are working – perfect for spotting I/O bottlenecks before they slow your applications down. The tool ships with the `sysstat` package and often has to be installed first. One thing to remember: the very first report is always an average since the last boot, so for current values you measure with an interval, e.g. `iostat -xz 2`. This guide takes you from the basics to the extended metrics such as `%util`, `await` and `aqu-sz`. ## Basic Usage `iostat` — Show CPU and disk I/O statistics since boot. ```bash iostat ``` `iostat INTERVAL` — Show statistics every INTERVAL seconds, continuously. ```bash iostat 2 ``` `iostat INTERVAL COUNT` — Show statistics every INTERVAL seconds, COUNT times. ```bash iostat 2 10 ``` `iostat -h` — Human-readable output with unit suffixes. ```bash iostat -h ``` ## CPU Statistics `iostat -c` — Show only CPU statistics. ```bash iostat -c 1 ``` `iostat -c 1 5` — Quick CPU utilization snapshot (five reports, one per second). ```bash iostat -c 1 5 ``` ## Disk Statistics `iostat -d` — Show only disk I/O statistics. ```bash iostat -d ``` `iostat -d DEVICE` — Show statistics for a specific device. ```bash iostat -d sda ``` `iostat -d DEVICE1 DEVICE2` — Show statistics for multiple devices. ```bash iostat -d sda nvme0n1 ``` `iostat -p DEVICE` — Show statistics for a device and all its partitions. ```bash iostat -p sda ``` `iostat -p ALL` — Show statistics for all devices and partitions. ```bash iostat -p ALL ``` ## Extended Statistics `iostat -x` — Show extended statistics: `await`, `%util`, queue sizes and more. ```bash iostat -x ``` `iostat -xd INTERVAL` — Extended disk statistics at regular intervals. Most common usage. ```bash iostat -xd 1 ``` `iostat -xd DEVICE INTERVAL` — Extended stats for a specific device. ```bash iostat -xd sda 2 ``` `iostat -x -t` — Extended stats with timestamps. ```bash iostat -x -t 1 ``` ## Output Options `iostat -k` — Show statistics in kilobytes per second. ```bash iostat -k ``` `iostat -m` — Show statistics in megabytes per second. ```bash iostat -m ``` `iostat -t` — Print timestamps with each report. ```bash iostat -t 1 ``` `iostat -y INTERVAL` — Skip the first (since-boot) report. Show only interval data. ```bash iostat -y 1 ``` `iostat -z` — Omit inactive devices from output. ```bash iostat -z 1 ``` `iostat -o JSON` — Output in JSON format. ```bash iostat -o JSON 1 3 ``` `iostat -N` — Show device mapper names for LVM volumes. ```bash iostat -N ``` ## Key Metrics Explained `tps` — Transfers per second (I/O requests). Combined read + write operations. `kB_read/s, kB_wrtn/s` — Kilobytes read/written per second. `r/s, w/s` — Read/write requests per second (extended mode). `rkB/s, wkB/s` — Kilobytes read/written per second (extended mode). `await` — Average time (ms) for I/O requests. Includes queue time + service time. `r_await, w_await` — Average time (ms) for read/write requests separately. `aqu-sz` — Average queue size. High values indicate I/O saturation. `%util` — Percentage of time the device was busy serving I/O requests. 100% = saturated. `rareq-sz, wareq-sz` — Average read/write request size in kilobytes. ## Common Patterns `iostat -xdm 1` — Real-time extended disk monitoring in MB/s. The go-to command for I/O analysis. ```bash iostat -xdm 1 ``` `iostat -xd 1 | awk '$NF > 80'` — Watch for disks with utilization above 80%. ```bash iostat -xd 1 | awk '$NF > 80' ``` `iostat -xdm -y 1 60 > iostat.log` — Log 60 seconds of I/O stats (skip boot average). ```bash iostat -xdm -y 1 60 > iostat.log ``` ## Conclusion iostat is a read-only command – it changes nothing on your system, so you can run it safely on production servers. For everyday work, the one pattern worth memorising is `iostat -xz 2`: it hides inactive devices, prints the extended metrics and measures real interval values instead of the since-boot average. Watch `%util` (device saturation), `await` (response time including queueing) plus `r/s`/`w/s` and `aqu-sz`. Note that on modern SSDs and NVMe drives a high `%util` alone doesn't necessarily mean a bottleneck – they handle many requests in parallel, so always read `await` and `aqu-sz` alongside it. ## Further Reading - [iostat(1) man page](https://man7.org/linux/man-pages/man1/iostat.1.html) – official reference for all options and metrics - [sysstat project](https://github.com/sysstat/sysstat) – source and documentation for the sysstat package ## Related Commands - [vmstat](https://www.jpkc.com/db/en/cheatsheets/shell-system/vmstat/) – overview of memory, process and I/O statistics - [top](https://www.jpkc.com/db/en/cheatsheets/shell-system/top/) – watch processes and system load in real time - [free](https://www.jpkc.com/db/en/cheatsheets/shell-system/free/) – show free and used memory