Re: [RFC] vhost-blk implementation

[Chris Wright <chrisw@sous-sol.org>]

* Badari Pulavarty (pbadari@us.ibm.com) wrote:
> I modified my vhost-blk implementation to offload work to
> work_queues instead of doing synchronously. Infact, I tried
> to spread the work across all the CPUs. But to my surprise,
> this did not improve the performance compared to virtio-blk.
> 
> I see vhost-blk taking more interrupts and context switches
> compared to virtio-blk. What is virtio-blk doing which I
> am not able to from vhost-blk ???

Your io wait time is twice as long and your throughput is about half.
I think the qmeu block submission does an extra attempt at merging
requests.  Does blktrace tell you anything interesting?

> procs -----------memory---------- ---swap-- -----io---- --system-- -----cpu-----
>  r  b   swpd   free   buff  cache   si   so    bi    bo   in   cs us sy id wa st
>  3  1   8920  56076  20760 5603556    0  104   196 79826 17164 13912  0  5 65 30  0
>  2  4   9488  57216  20744 5605616    0  114   195 81120 17397 13824  0  5 65 30  0
>  2  2  10028  68476  20728 5594764    0  108   206 80318 17162 13845  0  5 65 30  0
>  0  4  10560  70856  20708 5593088    0  106   205 82363 17402 13904  0  5 65 30  0
>  1  3  10948  80380  20672 5584452    0   78   178 79714 17113 13875  0  5 66 29  0
> 
> qemu virtio-blk:
> 
> procs -----------memory---------- ---swap-- -----io---- --system-- -----cpu-----
>  r  b   swpd   free   buff  cache   si   so    bi    bo   in   cs us sy id wa st
>  0  1  14124  57456   5144 4924060    0    0   139 142546 11287 9312  1  4 80 15  0
>  0  2  14124  56736   5148 4927396    0    0   146 142968 11283 9248  1  4 80 15  0
>  0  1  14124  56712   5384 4927020    0    0    74 150738 11182 9327  1  4 80 16  0
>  1  1  14124  55496   5392 4927904    0    0     2 159902 11172 9401  1  3 79 17  0
>  0  1  14124  55968   5408 4927232    0    0     0 159202 11212 9325  1  3 80 16  0