21:00:38 <oneswig> #startmeeting scientific-sig
21:00:39 <openstack> Meeting started Tue Dec 10 21:00:38 2019 UTC and is due to finish in 60 minutes.  The chair is oneswig. Information about MeetBot at http://wiki.debian.org/MeetBot.
21:00:40 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
21:00:41 <martial> Hello friends, welcome to the next edition of the Scientific SIG meeting :)
21:00:43 <openstack> The meeting name has been set to 'scientific_sig'
21:00:50 <martial> Hi Stig ;)
21:00:55 <oneswig> greetings
21:00:59 <oneswig> What's new?
21:01:21 <martial> not much, trying to see the rules about sharing a container built with CuDNN
21:01:36 <martial> looks like they finally integrated DNN for OpenCV
21:01:51 <oneswig> Not my wheelhouse.  What's the issue?
21:02:27 <martial> CuDNN is an Nvidia "developper" only download
21:02:34 <martial> but they provide the base container
21:02:57 <oneswig> Ah so it's not something that any dockerfile can pull in?
21:03:13 <martial> https://hub.docker.com/r/nvidia/cuda/
21:03:21 <martial> it is
21:03:31 <martial> well if you start from the official one
21:03:59 <martial> but pulling means that you are following the license terms set on those pages
21:04:00 <janders> g'day all
21:04:07 <janders> sorry for being late
21:04:18 <oneswig> Hi janders, hardly late
21:05:31 <oneswig> martial: so it's a question of pulling in a license that you don't care for, implicitly?
21:06:00 <martial> something like that
21:06:19 <oneswig> To borrow a little from Sherlock Holmes, that is a two-cheesecake problem :-)
21:07:22 <oneswig> At this end, I've been exploring a bit of mlnx network infrastructure I've not used before - VF LAG - anyone tried it previously?
21:07:56 <martial> not here
21:08:25 <oneswig> The idea is that if your hypervisor has a dual-port Mellanox NIC, if you bond those ports you can create virtual functions that somehow pass traffic that is also bonded (but presented as one virtual function)
21:09:22 <martial> so mirror ?
21:10:02 <oneswig> More the sum of the parts than their reflection, I'd hope.  But I'm still piecing it together.
21:10:30 <oneswig> The investigation continues
21:10:46 <janders> I haven't use those yet
21:10:50 <janders> *used
21:11:05 <martial> seems cool that said
21:11:47 <oneswig> I am not convinced how much traffic distribution there is between the underlying physical ports, but I'll report back when I find out.
21:11:48 <janders> agree that the option of having benefits of LAG without having to set up LAG in every instance sounds promising
21:12:32 <oneswig> I hope so.
21:12:39 <oneswig> What's new with you janders?
21:14:45 <oneswig> trandles and jmlowe send their apologies, I think there's a lot of winding up for end-of-year going on
21:17:17 <oneswig> I don't think we have more agenda items to cover - anything from you martial?
21:17:27 <martial> yep people going on vacation and such
21:18:45 <oneswig> OK, shall we close?  Your dockerfile awaits :-)
21:21:34 <oneswig> #endmeeting