Open Stack

A talk before summit is a good idea.

--jyh

From: Ian Wells [mailto:ijw.ubuntu at cack.org.uk]
Sent: Saturday, March 30, 2013 6:42 PM
To: Itzik Brown; Ian Wells
Cc: OpenStack Development Mailing List
Subject: Re: [openstack-dev] PCI-passthrough dev ...

I see the Nova session is now http://summit.openstack.org/cfp/details/243 .  I don't want to pre-empt it - it's important that the discussion happens in front of as many people as possible - but a 40 minute session won't get us very far if we're all explaining what we've already got.

I can schedule a webex or two before the summit.  I'm not sure where the majority of people are located, but maybe Wednesday 3rd April, 10am PST / 7pm CET / 8pm IDT would suit?  Get in contact with me directly if you're interested, and particularly if you have a solution you'd like to talk about or demonstrate, or bitch on the list if it's not a good time (or alternatively if you just disapprove of the idea).

--
Ian.

On 28 March 2013 22:22, Itzik Brown <itzikb at dev.mellanox.co.il<mailto:itzikb at dev.mellanox.co.il>> wrote:
Ian,

I'm going to send a basic VIF Driver and a config for a network configuration using macvtap as part of Mellanox Quantum Plugin.
We can schedule an online meeting to discuss some of the key points we need to address before the summit.

Itzik


On 3/28/2013 7:29 PM, Ian Wells wrote:
Chuck's is not the only summit session:

http://summit.openstack.org/cfp/details/81 (the Quantum side of things if you're mapping network devices)

... and I know Mellanox have also been working on that particular side of things too.  Perhaps we could all get together at the summit before the sessions for a show and tell?  We could equally work it all out in the sessions, but it seems it would be hard to lead a session like that without having some knowledge up front of what other people have done.

To be fair, our code is basically Vladimir Popovski's (Zardara Storage's) work, tidied up and with a scheduler check to make sure there are actually available SRIOV functions on the node before scheduling.  It's there, it's actually a patch against Folsom at the moment, and it works, but I wouldn't lay claim to it necessarily being the One True Way to implement this (and I don't think it has tests enough to pass a review).
--
Ian.


On 25 March 2013 16:23, Russell Bryant <rbryant at redhat.com<mailto:rbryant at redhat.com>> wrote:
On 03/24/2013 01:44 PM, Ian Wells wrote:
> Yep, we've got code in test at the moment.
This is (at least) the second instance that I've heard of where PCI
passthrough has been implemented, but code hasn't surfaced yet.  It's
really unfortunate to see the duplication of effort happening.

Chuck Short also proposed a design summit session on it, presumably to
discuss implementing it yet another time:

    http://summit.openstack.org/cfp/details/29

It would be really nice to get some code out in the open for this.  :-)

--
Russell Bryant

_______________________________________________
OpenStack-dev mailing list
OpenStack-dev at lists.openstack.org<mailto:OpenStack-dev at lists.openstack.org>
http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev



_______________________________________________

OpenStack-dev mailing list

OpenStack-dev at lists.openstack.org<mailto:OpenStack-dev at lists.openstack.org>

http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-dev/attachments/20130401/ef85938f/attachment.html>