User Tools

Site Tools


Session Start: Wed Sep 16 22:01:33 2015 Session Ident: #dpacc 03[22:01] * Now talking in #dpacc [22:01] <@varunfsl> hi Lingli 01[22:01] <Lingli_Deng> Hi Varun [22:02] <@varunfsl> I am not able to join gotomeeting 01[22:02] <Lingli_Deng> Yes, I just notice the problem and advise to change to chatting here. [22:02] <@varunfsl> I get an error message saying that the meeting id is invalid 01[22:02] <Lingli_Deng> So did I. I will check with it later. [22:02] <@varunfsl> ok 01[22:02] <Lingli_Deng> Hope everybody sees my note in time. 03[22:03] * olli_ (d98c608c@gateway/web/freenode/ip. has joined #dpacc 01[22:03] <Lingli_Deng> Hi, who is this? 03[22:03] * Keith_ (c0373725@gateway/web/freenode/ip. has joined #dpacc 01[22:03] <Lingli_Deng> Hello Keith. [22:03] <Keith_> Hi Lingli 01[22:04] <Lingli_Deng> Let's wait a minute for everybody to come in. 03[22:04] * Rob (d98c608c@gateway/web/freenode/ip. has joined #dpacc 03[22:04] * Magnus_ (86868b46@gateway/web/freenode/ip. has joined #dpacc 03[22:05] * jfender ( has joined #dpacc [22:05] <Magnus_> Good morning everybody [22:05] <Rob> Morning Magnus! 03[22:05] * subha-fsl (6bdfd219@gateway/web/freenode/ip. has joined #dpacc 03[22:05] * picnet ( has joined #dpacc [22:05] <Magnus_> Hi Rob. How are you? [22:05] <picnet> Mike Rooke(Nokia) Here. 01[22:06] <Lingli_Deng> Hello. Would everybody identify yourself and affiliation when you coming in. 01[22:06] <Lingli_Deng> Thank you Mike. [22:06] <Rob> Good thanks Magnus hope you're well [22:06] <Keith_> Keith Wiles (Intel) [22:06] <Rob> This is Rob Dimond from ARM 03[22:06] * jfender is now known as jfender-altera- [22:06] <olli_> Ola Liljedahl from ARM [22:06] <subha-fsl> Subha (Freescale) [22:06] <Magnus_> Magnus Karlsson from Intel 03[22:06] * Kin-Yip (62d2713f@gateway/web/freenode/ip. has joined #dpacc [22:07] <jfender-altera→ Josh Fender (Altera) 03[22:07] * picnet is now known as MikeR 01[22:08] <Lingli_Deng> Ola, is Bob coming today? 01[22:08] <Lingli_Deng> shall we wait for him? [22:08] <olli_> I suppose so [22:08] <MikeR> is there an associated voice call or are we chat only today? [22:08] <olli_> I will check with Bob [22:08] <Keith_> Lets wait and see a few minutes [22:09] <Keith_> Today it is chat only :-( 01[22:09] <Lingli_Deng> I am sorry I am having problems with GTM, and don't have any backup for now. 03[22:10] * Bose (3ba3cafe@gateway/web/freenode/ip. has joined #dpacc 03[22:10] * mikeH ( has joined #dpacc [22:10] <olli_> And I liked GoToMeeting! Better than WebEx 03[22:11] * bob_ (d98c674b@gateway/web/freenode/ip. has joined #dpacc 01[22:11] <Lingli_Deng> I think we should get started now. [22:11] <olli_> On IRC, nobody can hear you scream [22:11] <Keith_> lol [22:11] <bob_> ok…I am on 03[22:11] * Raj_ (7aa683ac@gateway/web/freenode/ip. has joined #dpacc [22:11] <Keith_> or is it haha now 01[22:11] <Lingli_Deng> Hi Bob? [22:11] <bob_> yes…I am here 01[22:12] <Lingli_Deng> Great. Sorry for the inconvenience. 01[22:12] <Lingli_Deng> Let us get started. 01[22:12] <Lingli_Deng> Before the big discussion over gAPI requirements. I have a few slides to share and discuss about. 01[22:12] <Lingli_Deng> Would everybody go to the wiki page and open the slides that I put there 01[22:13] <Lingli_Deng> 03[22:13] * BillFischofer (181c46ef@gateway/web/freenode/ip. has joined #dpacc 01[22:13] <Lingli_Deng> on page 2, you can see the schedule of B-release 01[22:14] <Lingli_Deng> which is proposed by Debra of release project, and approved by TSC 02[22:14] * Bose (3ba3cafe@gateway/web/freenode/ip. Quit (Ping timeout: 246 seconds) 01[22:14] <Lingli_Deng> As discussed last week, we applied to join in B-release project, for dpacc documentation. 01[22:15] <Lingli_Deng> To keep up the pace, we need to do two things in the following weeks as highlighted on this page. 01[22:15] <Lingli_Deng> AR1: identify work items and dependencies in Jira 01[22:15] <Lingli_Deng> AR2: upstream requirements published 02[22:15] * bob_ (d98c674b@gateway/web/freenode/ip. Quit (Ping timeout: 246 seconds) 01[22:15] <Lingli_Deng> which are detailed in the following two pages 01[22:16] <Lingli_Deng> any questions on Page 2? [22:16] <Keith_> None for me 01[22:16] <Lingli_Deng> I will give you a minute to read the table before we move to next page. 03[22:16] * Bose (3ba3cafe@gateway/web/freenode/ip. has joined #dpacc 01[22:17] <Lingli_Deng> Ok, next page. 01[22:17] <Lingli_Deng> for this week, we identify work items and dependencies in Jira 02[22:18] * Raj_ (7aa683ac@gateway/web/freenode/ip. Quit (Ping timeout: 246 seconds) 03[22:18] * bob_ (d98c674b@gateway/web/freenode/ip. has joined #dpacc [22:18] <bob_> hello? [22:18] <bob_> ok…my echo is working again 02[22:18] * BillFischofer (181c46ef@gateway/web/freenode/ip. Quit (Quit: Page closed) 01[22:19] <Lingli_Deng> for the current proposed 4 work items, please feedback you comments on the current proopsed dependencies. 01[22:19] <Lingli_Deng> @Bob, I will copy the previous notes to you privately. [22:19] <bob_> i am back it seems 02[22:20] * Lingli_Deng (~LIngli@ Quit (Excess Flood) Session Close: Wed Sep 16 22:20:05 2015

Session Start: Wed Sep 16 22:20:05 2015 Session Ident: #dpacc 02[22:20] * Attempting to rejoin channel #dpacc 03[22:20] * Rejoined channel #dpacc [22:20] <bob_> oh, i see the N/A 02[22:21] * Bose (3ba3cafe@gateway/web/freenode/ip. Quit (Ping timeout: 246 seconds) [22:21] <MikeR> high level requirements for various APIs as shown #1 slide 4. 01[22:21] <Lingli_Deng> sorry, I lost connection, back again. [22:21] <Keith_> It looks like #3 is the only one that applies to us, right? 03[22:21] * netoptimizer ( has joined #dpacc [22:21] <bob_> I am not familiar with the IPSec benchmarkig in QTip [22:21] <MikeR> slide 5 rather. 01[22:22] <Lingli_Deng> Bose, would you explain? [22:22] <Keith_> QTip is just a test framework, it could apply if we have something to test [22:22] <bob_> right, but is there a specific IPSec test that is being recommended? [22:23] <bob_> as for traffic profiles…last I saw it was just a shell and no tests actually exists 01[22:23] <Lingli_Deng> That is to be worked on via cooperation between us and them. [22:23] <bob_> but I have not seen a message since Ray P commneted the above [22:23] <Keith_> I think they have a set of tests they use for IPSec, but you can always write a new one if I am not mistaken [22:23] <Rob> The gap analysis would be against the requirements identified on the etherpad? [22:24] <bob_> Rob..that is my assumption 01[22:24] <Lingli_Deng> it seems we are crossing the lines. 01[22:24] <Lingli_Deng> I am trying to soliciting feedback on Page 3. 01[22:24] <Lingli_Deng> Is it suggested that we remove dependencies from Functest and QTip? [22:25] <Keith_> #3 applies and #4 applies if we have something to test [22:25] <Keith_> I do not see it as a dependency at this time [22:25] <subha-fsl> @Lingli, what is functest? [22:25] <subha-fsl> Is there any document around that? [22:25] <olli_> If we are providing something, it should be testable? 01[22:26] <Lingli_Deng> It is another OPNFV project, which would deliver a set of traffic profiles as output. [22:26] <subha-fsl> It is possible to identify test cases for IPsec and give that input for Qtip [22:26] <olli_> Or how do we know our delivery is fulfilling the requirements? 01[22:26] <Lingli_Deng> I am thinking of using their output, e.g. mobile traffic profile, for testing later. [22:27] <Rob> I am not familiar with the details of the projects but it makes sense for there to be testing infrastructure that can show the benefit of acceleration [22:27] <Keith_> The output from Qtip is the scripts and profiles, correct? 01[22:28] <Lingli_Deng> @Subha I don't think so, they are currently work-in-progress. [22:28] <olli_> I would like to see one deliverable from DPACC being a VM guest/host interface for accessing accelerated crypto/IPSec [22:28] <Keith_> I do not think we have a dependency on QTip only that we use the framework for any testing [22:29] <Rob> @Keith: makes sense 03[22:29] * Bose (3ba3cafe@gateway/web/freenode/ip. has joined #dpacc 01[22:29] <Lingli_Deng> OK. I did not see Bose around. I will check with him later and come back. 03[22:29] * Mario_ (df3ed80d@gateway/web/freenode/ip. has joined #dpacc 01[22:30] <Lingli_Deng> Hi Bose, Keith is wondering if QTip outputs scripts and profiles? 01[22:30] <Lingli_Deng> woudl you help clarifying [22:30] <Rob> I am glad to see virtio there I think this is an area where we can really make progress [22:30] <bob_> Yes…if we can produce test cases for sample implementations, seems Qtip could useful facilitate testing. 03[22:31] * julien_ZTE (~zhang@ has joined #dpacc [22:31] <Keith_> Yes, virtio testing is a must and one of our main goals IMO [22:31] <bob_> AGreed, Rob. We should propose to work on a sample virtio like crypto and test? [22:31] <julien_ZTE> #info Julien [22:31] <Bose> Currently QTIP is will be based on Ansible scripts [22:32] <Magnus_> Agree that virtio-crypto is really interesting [22:32] <bob_> Bose…does that mean custom tests can be built to use that scripting language to create tests of target software?> [22:32] <Magnus_> Think there was a draft proposal from Freescale around this [22:32] <Magnus_> a while back [22:32] <bob_> I believ so, Magnus [22:33] <subha-fsl> yes, there is on virtio-ipsec [22:33] <olli_> yes. I think that is really the only meaningful deliverable we have as of now [22:33] <Bose> Yes custom scripts can be used, but QTIP focuses only on performance measurements [22:33] <Magnus_> Why not use that as a starting point [22:33] <Magnus_> The goal should be for this to be upstreamed into Linux and Qemu [22:33] <Magnus_> Linux/KVM [22:34] <Magnus_> The proposal was on the IPSec level, but maybe we should start on the lower crypto level [22:34] <olli_> So upstreaming is one acceptance test 01[22:34] <Lingli_Deng> Let us mark a AR here, gap analysis for virtio-ipsec. I need a volunteer to start and take care of the work. [22:34] <bob_> Magnus could be right [22:34] <Magnus_> ? Should have been a question :-) [22:35] <Magnus_> Would like some help though [22:35] <bob_> I am no tthe expert on the levels but perhaps lower cyrpto is a starting point [22:35] <Magnus_> But my proposal is to start with virtio-crypto and then build IPSec on top of that [22:35] <subha-fsl> Some of the virtio-ipsec messages can be re-used for virtio-crypto as well [22:35] <Rob> I can see value in complete protocol offload (stateful) and stateless. I think that is Magnus' point? [22:36] <Magnus_> Agree with Subha [22:36] <bob_> Subha? FF? [22:36] <subha-fsl> yes, agree. we can re-use the messages from virtio-ipsec [22:36] <subha-fsl> we can add additional messages that may be needed for virtio-crypto [22:36] <bob_> I seem to recall tht both FF and Freescale had proposed prototype concepts [22:36] <subha-fsl> freescale proposal addresses lookaside accelerator [22:37] <Magnus_> Where can I get FF's proposal? [22:37] <olli_> pure crypto acceleration has to be lookaside [22:37] <subha-fsl> agree with @rob complete offload would be an interesting to work 01[22:37] <Lingli_Deng> Yes, FF also indicated agreement with Freescale's proposal after the July hackfest. [22:38] <bob_> I believe the proposals are archived, LingLi? [22:38] <bob_> previous preso [22:38] <Magnus_> But let us start to build this from the ground up 01[22:38] <Lingli_Deng> Great guys. looking forward to your contribution. Magnus and Subha. [22:38] <Magnus_> Complete offload is interesting, but is that not step two? [22:38] <olli_> "complete offload", do you mean IPsec protocol processing or inline IPsec [22:38] <Magnus_> Virtio today is very basic. See it hard to get virtio-ipsec in there before something more basic like virtio-crypto [22:39] <Rob> @Magnus: Yes I think we all agree 01[22:39] <Lingli_Deng> agreed. [22:39] <Magnus_> Former [22:39] <Mario_> Agreed [22:39] <olli_> so step 1: plain virtio-crypto [22:39] <subha-fsl> recommend we keep virtio-ipsec as baseline; re-use relevant messages and update virtio-ipsec as needed [22:40] <Keith_> agreed 01[22:40] <Lingli_Deng> Let us keep the detailed discussion of today's meeting. [22:40] <Magnus_> Agree with Subha 01[22:40] <Lingli_Deng> move on to next page, guys. [22:41] <bob_> @subha, not sure what that means as baseline v doing the virtio-crypto as step 1 [22:41] <Keith_> Page #$? [22:41] <Keith_> 4? [22:41] <bob_> yes 01[22:41] <Lingli_Deng> Yes, as you can see, 01[22:42] <Lingli_Deng> no other comments on the proposed editor/care-taker for the docs? [22:42] <bob_> I agree with the separation. high level reqs do not list out steps and do not dictate an implementations 01[22:42] <Lingli_Deng> Howard Huang for Openstack? 01[22:42] <Lingli_Deng> Keith for DPDK? 01[22:42] <Lingli_Deng> Bob for ODP? [22:42] <bob_> gap analysis for proposed solution x and Y can enumrate gaps and etc [22:42] <bob_> Yes…I will work with my team to assess ODP gaps 01[22:43] <Lingli_Deng> and Magnus for Virtio [22:43] <bob_> we assume, againsts requirements? 01[22:43] <Lingli_Deng> Thanks Bob 01[22:43] <Lingli_Deng> Yes, that brings us to the last page. 01[22:43] <Lingli_Deng> Page 5. 01[22:44] <Lingli_Deng> there are questions about the dependencies to QTip and Functest. I will take an AR for further clarification. [22:44] <bob_> work items? for which? deadline tomorrow seems tight 02[22:44] * Bose (3ba3cafe@gateway/web/freenode/ip. Quit (Ping timeout: 246 seconds) 01[22:45] <Lingli_Deng> to move forward with gap analysis, we need to push the high level requirements ASAP 01[22:45] <Lingli_Deng> the proposed time points, any comments? [22:46] <bob_> I am not clear on work items deadline [22:46] <bob_> what work items? [22:46] <Keith_> The times look fine, just not sure how to meet them today [22:46] <bob_> AR1 [22:46] <olli_> @Magnus, could we help you with virtio-crypto? 01[22:46] <Lingli_Deng> @Bob back to page 2, according to the schedule of B-release [22:47] <Magnus_> Yes please. 01[22:47] <Lingli_Deng> we need to fulfill three tasks the following two weeks 01[22:47] <Lingli_Deng> AR1 by this week, preferably. [22:47] <Magnus_> I do not have any time this week to devote to this. [22:47] <olli_> Good. Let's talk offline. 01[22:47] <Lingli_Deng> AR2 by next week (high level requirements), for the best. [22:47] <olli_> Next week? I am on California… [22:47] <Magnus_> Meeting with you guys on Friday plus travel. [22:48] <Magnus_> I leave California on Sunday :-( 02[22:48] * Mario_ (df3ed80d@gateway/web/freenode/ip. Quit (Ping timeout: 246 seconds) [22:48] <bob_> Many of us are in a week long conference next week and an all day meeting on Friday. [22:49] <Magnus_> Linaro Connect next week? [22:49] <bob_> tough deadlines and I am still not clear on what needs tobe created [22:49] <bob_> for AR 1 ,,Jira cards? [22:49] <bob_> I think we can clean up requirements by next week but htne need committer reveiw 01[22:49] <Lingli_Deng> I will create them based on our consensus. [22:50] <olli_> @Magnus. Yes. Let's continue over email. [22:50] <bob_> OK…let's just try to communicate and review on mailing list as much as possible [22:50] <Magnus_> OK [22:50] <bob_> That to LingLi 01[22:51] <Lingli_Deng> OK 01[22:51] <Lingli_Deng> Any other comments to the slides? [22:51] <bob_> I do not think so. [22:51] <Keith_> Am good with the slides, but it going to be hard to meet the deadlines 01[22:52] <Lingli_Deng> If not, I will leave it to Keith and Bob for gAPI requirements. [22:52] <bob_> @Keith..agreed [22:52] <bob_> Ok, so Bill, Rob, Keith and I need to try to clean up requirements by Monday [22:53] <bob_> But we will want to remove implementations from requriements [22:53] <@varunfsl> I am not clear on the comparison between the virtio-crypto and virtio-ipsec. Virtio-ipsec is stateful, whereas a virtio-crypto device could be stateless [22:53] <bob_> then committers can review and ratify 03[22:54] * Magnus (86868b46@gateway/web/freenode/ip. has joined #dpacc [22:54] <olli_> So virtio-crypto could be simpler 03[22:54] * Keith (86868b4a@gateway/web/freenode/ip. has joined #dpacc 02[22:54] * Magnus_ (86868b46@gateway/web/freenode/ip. Quit (Ping timeout: 246 seconds) [22:54] <Magnus> @olli got kicked out so missed your conversation [22:54] <bob_> I will repeat my statement re requirements [22:55] <olli_> Although we might opt to make it very similar to virtio-ipsec anyway [22:55] <bob_> Keith, Bill, Rob D and I need to try and close requirements by Monday ish [22:55] <Keith> OK, back. @bob please update the gapis and we can discuss those changes later [22:55] <bob_> But we want to be clear that we want all implmentation spoecifications/dictates removed from requirements. [22:55] <bob_> This is requirements 101 [22:55] <olli_> I responded to Varun's comment on virtio-crypto could be stateless. Which is good because we could make the interface simpler [22:56] <bob_> sounds good Keith [22:56] <@varunfsl> but don't agree with the point about pushing virtio crypto upstream as a base of virtio-ipsec 02[22:56] * Keith_ (c0373725@gateway/web/freenode/ip. Quit (Ping timeout: 246 seconds) 03[22:56] * Mario_ (7be452ec@gateway/web/freenode/ip. has joined #dpacc [22:56] <@varunfsl> upstream [22:56] <bob_> OK…we lsot Keith it seems…We will be discussing this Friday Lingli [22:56] <bob_> I need to run [22:56] <@varunfsl> they can exist in parallel [22:56] <Magnus> Anything we do with Linux or other open source has to go upstream 01[22:57] <Lingli_Deng> thanks Bob. 02[22:57] * bob_ (d98c674b@gateway/web/freenode/ip. Quit (Quit: Page closed) [22:57] <olli_> as a base doesn't have to mean as a technical abse where virtio-ipsec builds upon virtio-crypto. [22:58] <olli_> virtio-crypto could just be a base for our efforts of extending virtio, building credibility and simplify future extensions. [22:58] <subha-fsl> @varun, yes they can exist in parallel. @Magnus: agree we need to make it available upstream [22:58] <olli_> But start with something simple 01[22:58] <Lingli_Deng> Keith? [22:58] <@varunfsl> we have started with POC for virtio-ipsec [22:59] <olli_> @varunfsl good [22:59] <olli_> But a POC does not have to be based on upstreamable code. 01[23:00] <Lingli_Deng> @subha any progress with usecase doc? [23:00] <Rob> @varunfsl always useful to learn from a POC. You should work with Ola and Magnus 02[23:01] * Keith (86868b4a@gateway/web/freenode/ip. Quit (Ping timeout: 246 seconds) [23:01] <@varunfsl> ok [23:02] <subha-fsl> @Lingli: I should have a version out today or tomorrow [23:02] <Magnus> This time I got kicked out of my physical room :-) [23:02] <subha-fsl> I will upload to the DPACC wiki [23:02] <Magnus> Back again 01[23:04] <Lingli_Deng> thanks, Subha 01[23:05] <Lingli_Deng> Time is up for today. 01[23:05] <Lingli_Deng> Thanks everyone for joining. 03[23:06] * jfender-altera- ( has left #dpacc 02[23:06] * MikeR ( Quit (Quit: ircII EPIC4-2.10.5 – Are we there yet?) 01[23:06] <Lingli_Deng> Bye.

dpacc/meeting_log_0916.txt · Last modified: 2015/09/17 06:47 by Lingli Deng