05:30:10 <anil_rao> #startmeeting taas
05:30:11 <openstack> Meeting started Wed Jul 13 05:30:10 2016 UTC and is due to finish in 60 minutes.  The chair is anil_rao. Information about MeetBot at http://wiki.debian.org/MeetBot.
05:30:12 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
05:30:14 <openstack> The meeting name has been set to 'taas'
05:30:35 <soichi> hi
05:30:42 <anil_rao> Hi
05:30:44 <kaz> hi
05:31:15 <anil_rao> #topic Finalize and submit Barcelona Summit Presentation proposal
05:31:54 <reedip> hi
05:32:02 <anil_rao> Hi
05:32:15 <soichi> reedip: hi
05:32:50 <soichi> reedip: thank you for your modification to abstract
05:32:57 <anil_rao> soichi,reedip: I just saw your suggested modifications to the abstract
05:33:22 <yamamot__> hi
05:33:23 <reedip> soichi: it was minor
05:33:29 <reedip> hi yamamot__
05:33:35 <anil_rao> one thing to keep in mind is that we have a 1000 character limit. The one I sent out yesterday was alerady at 995 chars.
05:33:57 <anil_rao> I'll reword it some more tomorrow and incorporate the changes from both of you.
05:34:46 <anil_rao> Hi yamamot__
05:34:55 <soichi> anil_rao: my proposal is minor modification, so you can ignore it
05:35:41 <anil_rao> soichi: I think I can include it because it is a nice point. I'll just changes the words a little bit to make it all fit in 1000 chars. :-)
05:35:55 <soichi> thank you
05:36:29 <soichi> we need to ansewer two questions:
05:36:42 <anil_rao> Yes, let's discuss this a bit.
05:36:43 <reedip> lemme try and reword it
05:36:57 <soichi> 1) What is the problem or use case you're addressing in this session?
05:37:16 <anil_rao> Reedip, I have it done mostly. so I'll make final adjustments just before submitting tomorrow morning. :-)
05:37:24 <reedip> Umm, okay :)
05:37:45 <anil_rao> soichi: We will elaborate that section by listing the following items.
05:38:12 <reedip> anil_rao: I have shared the speaker details on the email (JFYI)
05:38:12 <anil_rao> 1. The plugin/agent/driver architecture which allows platform specific and vendor extension to be implemented.
05:38:28 <anil_rao> reedip: Thanks you. :-)
05:38:58 <anil_rao> 2. The need for isolation between production and mirrored traffic in the underlay network and some techniques to address this.
05:39:16 <anil_rao> 3. Performance analysis and lessons learnt from it. Steps to improve performance going forward.
05:39:36 <anil_rao> Any other thoughts?
05:41:00 <soichi> It looks fine for me.
05:41:26 <reedip> We should keep it light enough. Going to omuch into technicality can overload the users and operators in the audience
05:42:04 <reedip> so I think we can keep the information limited the ones specified above
05:42:17 <anil_rao> reedip: sounds good.
05:43:19 <anil_rao> For the section on what attendees expect to learn, I was thinking about the following:
05:44:00 <anil_rao> 1. Discover that the TaaS architecture lends itself to multiple backend solutions to meet different needs.
05:44:52 <anil_rao> 2. Several service providers had indicated an interest in having a completely separate data path for monitored traffic. So now they can learn about how this may be achieved.
05:45:17 <anil_rao> Others can also learn why separating the production and mirrored traffic can be beneficial for them.
05:45:49 <anil_rao> 3. The performance analysis is long overdue. We have been asked about TaaS performance during both our previous presentations.
05:46:13 <vnyyad> are we going to provide a demo this time? oor a prersentation
05:46:45 <anil_rao> vnyyad: I think we can skip a demo but provide real data from performance experiments.
05:46:57 <vnyyad> ok.
05:47:01 <anil_rao> what do you think?
05:47:13 <soichi> anil_rao: agree
05:47:27 <kaz> sounds good for me
05:47:43 <yamamot__> anil_rao: sounds good
05:47:49 <vnyyad> yes i guess going into more tech detaill of the solution to get more people develop back ends might be good
05:48:34 <anil_rao> vnyyad: Yes will have a section on the plugin/agent/driver architecture showing how it facilitates writing vendor/platform specific backends.
05:48:48 <vnyyad> that will be great
05:49:20 <soichi> +1
05:50:03 <anil_rao> Well, let's hope we get picked. :-)
05:50:15 <soichi> sure
05:50:19 <vnyyad> In the interest of full disclosure, we from Ericsson in partner with a customer have also submitted a tech talk on applying TaaS to monitor in Telco NFV
05:50:46 <soichi> it sounds great!
05:50:50 <vnyyad> but its purely application of taas as a solution
05:51:04 <vnyyad> its in the telco track
05:51:17 <anil_rao> vnyyd: We are submitting in the Network section so having a TaaS presentation in the Telco section is great!
05:51:34 <vnyyad> yes.... that would be great!
05:52:00 <vnyyad> we will have more representation one on the technical siide and the other on application of it
05:52:12 <soichi> i think it is very noce to have a actical use case
05:52:22 <soichi> noce -> nice
05:52:25 <anil_rao> +1
05:53:26 <anil_rao> soichi,kaz: Do you want to discuss the perf results you had posted a few weeks back?
05:54:27 <kaz> excuse me, we can not have a time for prepearaton, so we would like to discuss next week.
05:54:49 <anil_rao> kaz: Sure, no problem. We can cover that topic next week.
05:55:03 <kaz> thank you.
05:56:06 <vnyyad> i guess we should also bring up the agenda of separating data path for Taas in the comming metings as well
05:56:56 <soichi> sure
05:57:47 <soichi> currently, Kaz is working to implement by using flow based tunneling
05:57:47 <vnyyad> thanks!
05:58:01 <vnyyad> that is good
05:58:32 <vnyyad> does that involve any tunneling protocol like GRE or VxLAN
05:59:04 <kaz> VXLAN
05:59:16 <vnyyad> ok thanks Kaz
05:59:47 <vnyyad> so i guess the idea is to setup tunnels out of band to the existing overlay tunnels
06:01:23 <soichi> i'm not sure, we will explain our design in next week IRC
06:01:28 <vnyyad> anil: sorry for hijacking the discussion, please continue the agenda as proposed
06:01:43 <vnyyad> soichi: sure... thanks
06:02:22 <anil_rao> vnyyad: Our remaining agenda for today is Open Discussion so this is appropriate
06:02:30 <anil_rao> #topic Open Discussion
06:03:47 <vnyyad> We will also be working on a TaaS driver for VLAN based neutron network
06:04:25 <vnyyad> so i guess a separate tunnel based datapath would be good to have for that as well
06:04:51 <anil_rao> vnyyad: Yes.
06:07:30 <anil_rao> We need to consider TaaS requirements from two angles:
06:07:48 <anil_rao> 1. Using the existing tunnels or separate tunnels for the overlay
06:08:18 <anil_rao> 2. Using existing interfaces or separate interfaces (and possibly a separate switching fabric) for the underly.
06:08:31 <vnyyad> anil_rao: +1
06:08:35 <kaz> +1
06:08:44 <vnyyad> i guess flexibility to have both is the best
06:08:52 <soichi> +1
06:08:57 <kaz> +1
06:09:04 <anil_rao> In some cases, it might be shared at one level but isolated at the other. In some cases it might be shared in both or even isolated in both.
06:11:05 <vnyyad> +1
06:13:05 <soichi> I have another topic.
06:13:12 <anil_rao> Sure.
06:13:15 <soichi> I think we need to examine the bug(?) report:
06:13:23 <soichi> #link: http://lists.openstack.org/pipermail/openstack-dev/2016-July/098606.html
06:14:36 <anil_rao> soichi: I am not sure if this happens when both VMs are on the same host. I know we have a problem when using a single-node setup and trying to reach a VM from outside the host.
06:15:16 <anil_rao> soichi,kaz: I wanted to ask you this. Since your perf tests are done on a single node setup and you are passing traffic between two VMs on the same host you have it working correctly.
06:18:10 <kaz> we modified a flow of TapFlow to disable "dl_vlan=..."
06:18:37 <anil_rao> kaz: Thanks!
06:18:47 <anil_rao> Let me look into this problem.
06:19:05 <soichi> okay, thank you.
06:19:13 <vnyyad> thanks
06:19:50 <anil_rao> FYI, we are using the vlan check here because we didn't want to assume that mac address are unique across virtual networks that may be sharing a host.
06:20:47 <soichi> thank you for your information
06:21:01 <vnyyad> anil: +1
06:21:30 <anil_rao> I'll check this out and if there is a problem try to get it fixed.
06:22:51 <soichi> we will report if we find something new
06:23:16 <anil_rao> soichi: Thanks
06:23:26 <vnyyad> thanks
06:26:13 <soichi> i have no more topic.
06:26:54 <anil_rao> Look forward to next week's meeting to discuss perf and separate data-path for TaaS
06:27:03 <soichi> sure
06:29:18 <anil_rao> #endmeeting