hesela.devRandom I/O
Random I/O is an access pattern in which successive requests do not consistently follow adjacent address ranges.
Meaning: Common usage describes a nonsequential workload, not necessarily a statistically random or uniform address distribution [1].
Benchmark setup: fio selects random reads or writes with rw=randread or rw=randwrite. Its random_distribution option defaults to uniform and also supports skewed distributions such as Zipf and Pareto [2]. The word random alone does not specify request size, read/write mix, accessed range or achieved queue depth [2].
Coverage: In fio 3.39, the default random map avoids revisiting blocks within a coverage pass. norandommap removes that tracking; it does not guarantee independent sampling, because the chosen generator still matters. The lfsr generator avoids repeat offsets and requires a single block size. Selecting a nonuniform random_distribution disables the map [3].
Relations
- related
- Sequential I/O, I/O request size, Queue depth, IOPS, Storage throughput, Solid-state drive (SSD), Hard disk drive (HDD), POSIX_FADV_RANDOM
Citations
- SNIA Dictionary, 28 April 2026: random I/O and sequential I/O, pp. 244-245 and 270.
- fio documentation: rw, blocksize, random_distribution, iodepth and serialize_overlap.
- fio 3.39 HOWTO: norandommap, random_generator and random_distribution.
Sources & caveats
- The definitions are original paraphrases, not reproductions of SNIA Dictionary entries.
- The fio online manual was reviewed on 24 October 2026; option semantics describe a workload generator, not measured device performance.