...
Many people have suggested that writing 1 byte is not efficient and writing more bytes should be faster. I tried it both with write-caching on and off using the same string of random a-z characters for each write. The code used in this benchmark is slightly modified to allow specifying the write size (up to 8KB), see: https://gist.github.com/ongardie/9218611 (oops, this code had a bug, it only called posix_fallocate for the first 1000 KBs, so most of the 4KB and 8KB areas were not pre-allocated. I should re-run those).
host | disk | write-caching | 1 byte | 1024 bytes | 4096 bytes | 8192 bytes |
---|---|---|---|---|---|---|
rcmonster | 520 | on | 440us | 570us | 1.0ms | 1.0ms |
rcmonster | 530 | on | 9.7ms | 12.0ms | 19.6ms | 19.8ms |
rc66 | M4 | on | 3.7ms | 5.1ms | 9.2ms | 10.4ms |
rc66 | 530 | on | 9.8ms | 12.5ms | 21.0ms | 21.2ms |
flygecko | X25-M | on | 210us | 370us | 910us | 950us |
flygecko | 530 | on | 9.7ms | 12.1ms | 19.6ms | 19.7ms |
rcmonster | 520 | off | 100us | 160us | 400us | 420us |
rcmonster | 530 | off | 10.5ms | 17.5ms | 55.9ms | 68.9ms |
rc66 | M4 | off | 1.5ms | 2.6ms | 5.6ms | 7.2ms |
rc66 | 530 | off | 1.9ms | 3.0ms | 6.9ms | 8.6ms |
flygecko | X25-M | off | 1.0ms | 1.7ms | 3.5ms | 3.7ms |
flygecko | 530 | off | 1.7ms | 2.4ms | 4.9ms | 5.0ms |
...