iops is an IO benchmark tool that performs random reads on block devices. If an exact block size is not specified using -b, the the size starts with the physical sector size (defaulting to 4k) and doubles every iteration of the loop. You can switch the read pattern using -p toggle.
iops [-n|--num-threads threads] [-t|--time time] [-m|--machine-readable]
[-b|--block-size size] [-p|--pattern random|sequential] <device>
num-threads := number of concurrent io threads, default 32
time := time in seconds, default 2
machine-readable := used to switch off conversion into MiB and other SI units
block-size := block size (should be a multiple of 512)
pattern := random|sequential
device := some block device, like /dev/sda or \\\\.\\PhysicalDrive0
$ iops /dev/sda
/dev/sda, 3.00 T, sectorsize=512B, #threads=32, pattern=random:
512 B blocks: 79.6 IO/s, 40.8 kB/s (326.2 kbit/s)
1 kB blocks: 62.3 IO/s, 63.8 kB/s (510.1 kbit/s)
2 kB blocks: 58.6 IO/s, 120.0 kB/s (959.8 kbit/s)
4 kB blocks: 49.0 IO/s, 200.6 kB/s ( 1.6 Mbit/s)
8 kB blocks: 55.5 IO/s, 454.3 kB/s ( 3.6 Mbit/s)
16 kB blocks: 59.9 IO/s, 981.3 kB/s ( 7.9 Mbit/s)
32 kB blocks: 60.7 IO/s, 2.0 MB/s ( 15.9 Mbit/s)
65 kB blocks: 53.4 IO/s, 3.5 MB/s ( 28.0 Mbit/s)
131 kB blocks: 43.7 IO/s, 5.7 MB/s ( 45.8 Mbit/s)
262 kB blocks: 45.5 IO/s, 11.9 MB/s ( 95.5 Mbit/s)
524 kB blocks: 28.5 IO/s, 14.9 MB/s (119.5 Mbit/s)
1 MB blocks: 22.1 IO/s, 23.2 MB/s (185.2 Mbit/s)
2 MB blocks: 18.4 IO/s, 38.7 MB/s (309.3 Mbit/s)
4 MB blocks: 9.2 IO/s, 38.5 MB/s (308.2 Mbit/s)
8 MB blocks: 6.6 IO/s, 55.1 MB/s (440.7 Mbit/s)
$ iops /dev/vda
/dev/vda, 34.36 GB, 32 threads:
512 B blocks: 374.1 IO/s, 187.0 KiB/s ( 1.5 Mbit/s)
1 KiB blocks: 322.6 IO/s, 322.6 KiB/s ( 2.6 Mbit/s)
2 KiB blocks: 285.6 IO/s, 571.2 KiB/s ( 4.7 Mbit/s)
4 KiB blocks: 268.3 IO/s, 1.0 MiB/s ( 8.8 Mbit/s)
8 KiB blocks: 270.1 IO/s, 2.1 MiB/s ( 17.7 Mbit/s)
16 KiB blocks: 227.0 IO/s, 3.5 MiB/s ( 29.8 Mbit/s)
32 KiB blocks: 212.0 IO/s, 6.6 MiB/s ( 55.6 Mbit/s)
64 KiB blocks: 157.0 IO/s, 9.8 MiB/s ( 82.3 Mbit/s)
128 KiB blocks: 137.7 IO/s, 17.2 MiB/s (144.4 Mbit/s)
256 KiB blocks: 99.0 IO/s, 24.8 MiB/s (207.7 Mbit/s)
512 KiB blocks: 66.9 IO/s, 33.5 MiB/s (280.8 Mbit/s)
1 MiB blocks: 44.7 IO/s, 44.7 MiB/s (375.2 Mbit/s)
2 MiB blocks: 22.3 IO/s, 44.7 MiB/s (375.0 Mbit/s)
$ iops --num-threads 8 --time 2 /dev/disk0
/dev/disk0, 251.00 G, sectorsize=512B, #threads=8, pattern=random:
512 B blocks: 38917.8 IO/s, 19.9 MB/s (159.4 Mbit/s)
1 kB blocks: 39416.6 IO/s, 40.4 MB/s (322.9 Mbit/s)
2 kB blocks: 40357.8 IO/s, 82.7 MB/s (661.2 Mbit/s)
4 kB blocks: 33269.5 IO/s, 136.3 MB/s ( 1.1 Gbit/s)
8 kB blocks: 24222.6 IO/s, 198.4 MB/s ( 1.6 Gbit/s)
16 kB blocks: 18491.4 IO/s, 303.0 MB/s ( 2.4 Gbit/s)
32 kB blocks: 10398.7 IO/s, 340.7 MB/s ( 2.7 Gbit/s)
65 kB blocks: 5611.3 IO/s, 367.7 MB/s ( 2.9 Gbit/s)
131 kB blocks: 2884.1 IO/s, 378.0 MB/s ( 3.0 Gbit/s)
262 kB blocks: 1479.3 IO/s, 387.8 MB/s ( 3.1 Gbit/s)
524 kB blocks: 714.0 IO/s, 374.3 MB/s ( 3.0 Gbit/s)
1 MB blocks: 370.3 IO/s, 388.3 MB/s ( 3.1 Gbit/s)
2 MB blocks: 187.4 IO/s, 393.1 MB/s ( 3.1 Gbit/s)
4 MB blocks: 90.2 IO/s, 378.4 MB/s ( 3.0 Gbit/s)
8 MB blocks: 44.4 IO/s, 372.7 MB/s ( 3.0 Gbit/s)
16 MB blocks: 22.1 IO/s, 371.1 MB/s ( 3.0 Gbit/s)
33 MB blocks: 12.4 IO/s, 416.7 MB/s ( 3.3 Gbit/s)
67 MB blocks: 6.1 IO/s, 408.2 MB/s ( 3.3 Gbit/s)
134 MB blocks: 2.9 IO/s, 394.7 MB/s ( 3.2 Gbit/s)
268 MB blocks: 1.3 IO/s, 343.3 MB/s ( 2.7 Gbit/s)
$ iops --time 10 --num-threads 1 --block-size 16768 --pattern sequential /dev/disk0
/dev/disk0, 251.00 G, sectorsize=512B, #threads=1, pattern=sequential:
16 kB blocks: 7815.2 IO/s, 131.0 MB/s ( 1.0 Gbit/s)
There is a Docker image available on Docker Hub. To run from Docker:
$ docker run --rm -it --device=/dev/sda cxcv/iops --num-threads 2 /dev/sda
To make system devices available inside the Docker container, add as many --device switches as needed.
- Grundlagen des I/O-Benchmarking (German)
- Fundamentals of I/O benchmarking (English)