Question about seperating request and response ring in PV network

Question about seperating request and response ring in PV network

[openlui]

[-- Attachment #1.1: Type: text/plain, Size: 768 bytes --]

Hi, all:
    I am trying to improve the performance of netfront/netback, and  I found that there were some discussion about PV network performance improvement in devel mailing list ([1]). The proposals mentioned  in [1] are helpful, such as multipage ring,  multiqueue, etc, and some of them have already merged into upstream.
    However, I am wondering if we have already used multipage ring,  why "seperate request and response ring" listed in [1] is also helpful for performance improvement? I have read the implementation of vmxnet3 driver, it does have separated request and response rings, but I can not understand what the advantage of this method is.
    Thanks.

[1] http://lists.xen.org/archives/html/xen-devel/2013-05/msg01904.html
----
Best Regards
Lui



[-- Attachment #1.2: Type: text/html, Size: 1197 bytes --]

[-- Attachment #2: Type: text/plain, Size: 126 bytes --]

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel

Re: Question about seperating request and response ring in PV network

[Wei Liu]

On Fri, Jul 10, 2015 at 09:41:10AM +0800, openlui wrote:
> Hi, all:
>
>     I am trying to improve the performance of netfront/netback, and  I
>     found that there were some discussion about PV network performance
>     improvement in devel mailing list ([1]). The proposals mentioned
>     in [1] are helpful, such as multipage ring,  multiqueue, etc, and
>     some of them have already merged into upstream.
>
>     However, I am wondering if we have already used multipage ring,
>     why "seperate request and response ring" listed in [1] is also
>     helpful for performance improvement? I have read the
>     implementation of vmxnet3 driver, it does have separated request
>     and response rings, but I can not understand what the advantage of
>     this method is.

Correction: network driver doesn't use multi-page ring and I don't think
we have plan to do that.

That email contains mostly brainstormed ideas that we thought might help
improve performance. Some ideas were solid (and upstreamed), some were
not. I implemented several prototype back in the day but I couldn't
remember what I did.

As for separating rings, there are two possibilities: 1) I didn't
implement that; 2) I implemented that but there was no visible
improvement. So there is no such feature in upstream driver.

Wei.

>     Thanks.
> 
> [1] http://lists.xen.org/archives/html/xen-devel/2013-05/msg01904.html
> ----
> Best Regards
> Lui
> 
> 

> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xen.org
> http://lists.xen.org/xen-devel