[dpdk-dev] [PATCH v4 00/17] Wind River Systems AVP PMD vs virtio?
Wiles, Keith
keith.wiles at intel.com
Thu Mar 16 10:51:56 CET 2017
Sent from my iPhone
> On Mar 16, 2017, at 4:53 PM, Francois Ozog <francois.ozog at linaro.org> wrote:
>
> Hi,
>
> Virtio is special in many ways:
> - it is a multi-vendor supported specification
> - it is a multi-vendor opensource implementation in guest OSes
> (Windows, Linux, FreeBSD...)
> - it is a multi-vendor, opensource implementation in hypervisors
>
>
> So, the great benefit of virtio is that with a SINGLE device driver in
> a VM, applications are guaranteed to work in all situations (all
> hypervisors, all backends). The real issue I see with AVP is that it
> would bring uncertainty in virtual environments, breaking the "peace"
> of mind that virtio brings. does the hypervisor supports this vnic?
> does the virtual switch support the vnic?
> Having a single multi-vendor supported specification and
> implementations foster creativity, so I wouldn't be surprised to see
> native virtio support from Smart NICs in a very near future!
>
> *** Bottom line, if there are good ideas in AVP (performance,
> security...), I would rather push them to virtio. ***
>
>
> Lastly, physical PMDs have been accepted based on implicit existence
> of upstream drivers (valid for virtio and vmxnet3). So as a bare
> minimum requirement, I would ask for Qemu, OVS and Linux upstream AVP
> support. Is it the case?
You are missing the point people will vote with there feet having competing solutions just forces a heathy eco system.
Because the code is now open sourced then virtio can just take the improvement ideas and put it in their code.
I do not believe you have convinced me that having another solution hurts the eco system.
You can say we already have a driver for one nic why would we want another one. If someone produces a driver for a nic we already then I would accept it into dpdk.
Maybe is slower or faster, less or more features, easier to maintain, but fills a gap like AVP then we must except that driver. At some point one will win or some will prefer one over the other.
>
> Cordially,
>
> François-Frédéric
>
>
> On 16 March 2017 at 04:18, O'Driscoll, Tim <tim.odriscoll at intel.com> wrote:
>>> From: Vincent JARDIN [mailto:vincent.jardin at 6wind.com]
>>>
>>> Le 15/03/2017 à 11:55, Thomas Monjalon a écrit :
>>>>> I'd suggest that this is a good topic for the next Tech Board
>>> meeting.
>>>> I agree Tim.
>>>> CC'ing techboard to add this item to the agenda of the next meeting.
>>>
>>> Frankly, I disagree, it is missing some discussions on the list.
>>
>> I think the discussion on the mailing list is at an impasse and it won't be resolved there. I think the Tech Board needs to consider several issues:
>> - What are the requirements for a new PMD to be accepted? For example, you're asking for performance data in this case, when this hasn't been a requirement for other PMDs.
>> - Should there be different requirements for PMDs for virtual devices versus physical devices?
>> - Based on these criteria, should the AVP PMD be accepted or not?
>
>
>
> --
> François-Frédéric Ozog | Director Linaro Networking Group
> T: +33.67221.6485
> francois.ozog at linaro.org | Skype: ffozog
More information about the dev
mailing list