美好的一天，我有一个问题但找不到答案。我测试了我的硬盘，使用 dd 安装在 /media/data 上。一些简单的脚本：
写：
echo "3" > /proc/sys/vm/drop_caches
for i in {1..10}; do
dd if=/dev/zero of=/media/data/test/testfile-$i bs=4096 count=1310720 conv=fdatasync
done
result:
my-home# bash test1
5368709120 bytes (5,4 GB, 5,0 GiB) copy, 30,7945 s, 174 MB/s
5368709120 bytes (5,4 GB, 5,0 GiB) copy, 30,8114 s, 174 MB/s
5368709120 bytes (5,4 GB, 5,0 GiB) copy, 30,9237 s, 174 MB/s
5368709120 bytes (5,4 GB, 5,0 GiB) copy, 30,9386 s, 174 MB/s
5368709120 bytes (5,4 GB, 5,0 GiB) copy, 30,8568 s, 174 MB/s
iostat:
avg-cpu: %user %nice %system %iowait %steal %idle
2,17 0,00 4,22 21,23 0,00 72,38
Device: tps kB_read/s kB_wrtn/s kB_read kB_wrtn
sda 0,00 0,00 0,00 0 0
sdb 172,00 0,00 176128,00 0 176128
读：
echo "3" > /proc/sys/vm/drop_caches
for i in $(ls /media/data/iso); do
dd if=/media/data/iso/$i of=/dev/null bs=4096
done
result:
my-home# bash test1_read
632291328 bytes (632 MB, 603 MiB) copy, 3,46719 s, 182 MB/s
3994091520 bytes (4,0 GB, 3,7 GiB) copy, 21,4655 s, 186 MB/s
1317427200 bytes (1,3 GB, 1,2 GiB) copy, 7,03572 s, 187 MB/s
4696938496 bytes (4,7 GB, 4,4 GiB) copy, 25,0823 s, 187 MB/s
699400192 bytes (699 MB, 667 MiB) copy, 3,77014 s, 186 MB/s
iostat:
avg-cpu: %user %nice %system %iowait %steal %idle
1,13 0,00 2,89 10,30 0,00 85,68
Device: tps kB_read/s kB_wrtn/s kB_read kB_wrtn
sda 0,00 0,00 0,00 0 0
sdb 1432,00 183296,00 0,00 183296 0
速度很好，但是当我在不同的终端中同时运行这些脚本时。 Write-script 运行速度更快并占用所有硬盘资源：
my-home# bash test1
Sep 29 7:57:57 UTC 2016
5368709120 bytes (5,4 GB, 5,0 GiB) copy, 31,9903 s, 168 MB/s
5368709120 bytes (5,4 GB, 5,0 GiB) copy, 32,0175 s, 168 MB/s
5368709120 bytes (5,4 GB, 5,0 GiB) copy, 32,0138 s, 168 MB/s
5368709120 bytes (5,4 GB, 5,0 GiB) copy, 31,9956 s, 168 MB/s
5368709120 bytes (5,4 GB, 5,0 GiB) copy, 32,0198 s, 168 MB/s
5368709120 bytes (5,4 GB, 5,0 GiB) copy, 32,136 s, 167 MB/s
5368709120 bytes (5,4 GB, 5,0 GiB) copy, 31,9074 s, 168 MB/s
5368709120 bytes (5,4 GB, 5,0 GiB) copy, 32,237 s, 167 MB/s
5368709120 bytes (5,4 GB, 5,0 GiB) copy, 32,0998 s, 167 MB/s
5368709120 bytes (5,4 GB, 5,0 GiB) copy, 32,4163 s, 166 MB/s
my-home# bash test1_read
Sep 29 7:57:57 UTC 2016
632291328 bytes (632 MB, 603 MiB) copy, 3,46899 s, 182 MB/s
3994091520 bytes (4,0 GB, 3,7 GiB) copy, 149,177 s, 26,8 MB/
4696938496 bytes (4,7 GB, 4,4 GiB) copy, 214,307 s, 21,9 MB/s
263127040 bytes (263 MB, 251 MiB) copy, 1,44909 s, 182 MB/s
iostat:
Device: tps MB_read/s MB_wrtn/s MB_read MB_wrtn
29.09.2016 08:03:17 sdb 164,00 1,38 157,03 2 314
29.09.2016 08:03:19 sdb 172,00 1,50 161,52 3 323
29.09.2016 08:03:21 sdb 163,00 1,38 156,52 2 313
29.09.2016 08:03:23 sdb 169,50 0,88 166,00 1 332
29.09.2016 08:03:25 sdb 164,50 0,38 162,01 0 324
29.09.2016 08:03:27 sdb 161,50 1,38 154,52 2 309
29.09.2016 08:03:29 sdb 168,00 0,50 165,51 1 331
29.09.2016 08:03:31 sdb 160,50 0,75 156,02 1 312
29.09.2016 08:03:33 sdb 164,00 0,38 162,00 0 324
29.09.2016 08:03:35 sdb 164,50 0,25 162,01 0 324
29.09.2016 08:03:37 sdb 277,50 0,88 156,52 1 313
29.09.2016 08:03:39 sdb 318,00 0,25 161,00 0 322
29.09.2016 08:03:41 sdb 313,50 1,00 154,35 2 308
29.09.2016 08:03:43 sdb 758,00 80,31 88,19 160 176
29.09.2016 08:03:45 sdb 1398,00 174,63 0,00 349 0
29.09.2016 08:03:47 sdb 1429,00 178,62 0,00 357 0
29.09.2016 08:03:49 sdb 1429,50 178,69 0,00 357 0
29.09.2016 08:03:51 sdb 1427,00 178,31 0,00 356 0
29.09.2016 08:03:53 sdb 1430,50 178,81 0,00 357 0
29.09.2016 08:03:55 sdb 1431,00 178,88 0,00 357 0
29.09.2016 08:03:57 sdb 1431,00 178,88 0,00 357 0
29.09.2016 08:03:59 sdb 1431,00 178,88 0,00 357 0
29.09.2016 08:04:01 sdb 1423,00 177,81 0,00 355 0
29.09.2016 08:04:03 sdb 1431,50 178,94 0,00 357 0
为什么写优先于读？我的调度程序是 Linux Mint 18 的默认调度程序：
my-home# cat /sys/block/sdb/queue/scheduler
noop [deadline] cfq
对于这个调度程序来说，读取的优先级高于写入，但事实并非如此。