[dm-devel] [PATCH 5/5] dm-mpath: improve I/O schedule

[Bart Van Assche]

On Sat, 2017-09-16 at 00:44 +0800, Ming Lei wrote:
> ---------------------------------------
>           |v4.13+         |v4.13+
>           |+scsi_mq_perf  |+scsi_mq_perf+patches
> -----------------------------------------
>  IOPS(K)  |MQ-DEADLINE    |MQ-DEADLINE
> ------------------------------------------
> read      |       30.71   |      343.91
> -----------------------------------------
> randread  |       22.98   |       17.17
> ------------------------------------------
> write     |       16.45   |      390.88
> ------------------------------------------
> randwrite |       16.21   |       16.09
> ---------------------------------------

What is the reason for the random I/O performance regressions? Users whose
workload triggers a random I/O pattern will be really unhappy with a 33% IOPS
regression for random reads. Does that regression get worse for lower latency
transports, e.g. SRP?

Bart.