linux 磁盘队列深度nr_requests 和 queue_depth

nr_requests
Specifies the maximum number of read and write requests that can be queued at one time. 
The default value is 128, which means that 128 read requests and 128 write requests can be queued before the next process to request a read or write is put to sleep.
For latency-sensitive applications, lower the value of this parameter and limit the command queue depth on the storage so that write-back I/O cannot fill the device queue with write requests.

When the device queue fills, other processes attempting to perform I/O operations are put to sleep until queue space becomes available.

Requests are then allocated in a round-robin manner, which prevents one process from continuously consuming all spots in the queue.
The maximum number of I/O operations within the I/O scheduler is nr_requests*2.

As stated, nr_requests is applied separately for reads and writes.

Note that nr_requests only applies to the I/O operations within the I/O scheduler and not to I/O operations already dispatched to the underlying device.
Therefore, the maximum outstanding limit of I/O operations against a device is (nr_requests*2)+(queue_depth) where queue_depth is /sys/block/sdN/device/queue_depth, sometimes also referred to as the LUN queue depth.

You can see this total outstanding number of I/O operations in, for example, the output of iostat in the avgqu-sz column.
指定一次可以排队的读请求和写请求的最大数目。

默认值是128，这意味着128个读请求和128个写请求可以在请求读或写的下一个进程进入睡眠状态之前排队。
对于对延迟敏感的应用程序，可以降低该参数的值，并限制存储上的命令队列深度，以防止回写I/O用写请求填满设备队列。

当设备队列满时，其他试图执行I/O操作的进程将进入休眠状态，直到队列空间可用。

然后以循环的方式分配请求，这可以防止一个进程持续地消耗队列中的所有位置。
I/O调度器中的最大I/O操作数是nr_requests*2。

如前所述，对于读和写，分别应用nr_requests。

注意，nr_requests仅适用于I/O调度器中的I/O操作，而不适用已经分配到底层设备的I/O操作。
因此，对一个设备的I/O操作的最大未完成限制为(nr_requests*2)+(queue_depth)，其中queue_depth为/sys/block/sdN/device/queue_depth，有时也称为LUN队列深度。

例如，您可以在avgqu-sz列中的iostat输出中看到这个未完成的I/O操作总数。

[root@node-1 ~]# echo "128" >/sys/block/sdj/device/queue_depth
[root@node-1 ~]# lsscsi -l | grep -A 1 sdj
[0:0:16:0]   disk    SEAGATE  ST1000NX0453     NS02  /dev/sdj 
  state=running queue_depth=128 scsi_level=7 type=0 device_blocked=0 timeout=90
[root@node-1 ~]# echo "256" >/sys/block/sdj/device/queue_depth 
[root@node-1 ~]# lsscsi -l | grep -A 1 sdj
[0:0:16:0]   disk    SEAGATE  ST1000NX0453     NS02  /dev/sdj 
  state=running queue_depth=256 scsi_level=7 type=0 device_blocked=0 timeout=90