I’d like to add some comments on the LaaS side. I have been frustrated about how difficult it is to run simple data plane benchmarks on any OPNFV openstack deployment. There is a large diversity of hardware, wiring blueprints, installers, with different lab teams that makes it difficult to setup and maintain. As a result I can’t run any NFVbench automation because of this lack of stable stack that can handle data plane properly.
I have recently been working on packet.net and I can see how much more easy it can get to bring up any complex stack, with proper arrangement with the packet.net infra team. Currently NFVbench runs on packet.net and can generate traffic to any commissioned server that can do L2 forwarding or L3 routing – and more importantly this can be automated pretty easily because the L2 data plane is programmable (which is a part missing in Pharos labs). This setup will be used for a benchmarking demo for CNCF Kubecon next week (I can provide more info if interested). The hard part was to make sure the L2 data plane in packet.net works properly and I think we have a good foundation at the moment (still in discussion with packet.net infra to make it even better). If we could have OPNFV deployers work on something similar to packet.net I think that would be awesome for a lot of OPNFV projects. Regarding devops CI/CD and use of deployers in production, I think it is important to clarify the difference between current OPNFV deployers, commercial deployers and real production. There is a huge gap between what any OPNFV deployer (or any off the shelf installer for that matter) can do today and the kind of deployments that end up running in production. In addition to the usual stability in feature and functionality verification which are covered in OPNFV, the critical additional pieces are * the interfacing between the virtualization layer and the final bare metal layer, which are simplified in a lab setting but cannot be simplified for real productions, * operational considerations (integrated monitoring/alarming/non-service impacting HW+SW updates/scale/performance and capacity planning…) * and the overall data plane blueprint (i.e. how to bring traffic into and out of potentially hundreds of VNFs) These areas are complex, cannot be trivialized and are currently not well covered by OPNFV but still have to be addressed by somebody. So the goal of coming up with a generic deployer, deploy it and run in production-like setting is not realistic today (even less the idea of doing live CI/CD for production clouds). You could get something that sort of works but you will run very quickly into all sorts of limitations/caveats. And this applies btw independently to which cloud OS is used (k8s or openstack) and of the type of deployment (edge or the traditional NFV private cloud). Best Regards, Alec From: <opnfv-tech-discuss@lists.opnfv.org> on behalf of "Frank Brockners via Lists.Opnfv.Org" <fbrockne=cisco....@lists.opnfv.org> Reply-To: "Frank Brockners (fbrockne)" <fbroc...@cisco.com> Date: Friday, November 30, 2018 at 3:45 AM To: Georg Kunz <georg.k...@ericsson.com>, "HU, BIN" <bh5...@att.com>, Tim Irnich <tim.irn...@suse.com>, Trevor Bramwell <tbramw...@linuxfoundation.org> Cc: "opnfv-tech-discuss@lists.opnfv.org" <opnfv-tech-discuss@lists.opnfv.org> Subject: Re: [opnfv-tech-discuss] [opnfv-tsc] Discussion of OPNFV Strategic Plan In the TSC meeting, Manuel voiced a pretty important ask that might help the discussion moving forward: "Cloud we create a table that compares OPNFV today with the proposed future", assuming that we'd evolve along the path that Bin started to articulate. The table format is to make things more concrete and could help us in a second step to articulate where we’d want to focus on moving forward. This includes which additional work areas we’d like to inspire projects to work on (or even new projects to be created for). I've tried to capture the current discussion and proposal into a table below (see also attached in case groups.io messes up the formatting and color) – feel free to add/evolve/change per your understanding – the below is just my reading of Bin’s deck. Thoughts? Thanks, Frank OPNFV today Potential Evolution (changes in blue) Design – Outline requirements; Design NFV components and solution stacks Design – Outline requirements; Design NFV components and solution stacks; Focus on cloud native and edge use cases. Create – Create new and/or enhance components (most often by working upstream) to meet the design/requirements Create – Create new and/or enhance components (most often by working upstream) to meet the design/requirements. Compose – Follow the design and create a running system from a set of components Compose – Follow the design and create a running system from a set of components Deploy – Install and run the composed system on a set of labs worldwide. This includes enhancement and creation of specific test and deployment tools (installers, XCI,..). Deploy – Install and run the composed system on a set of labs worldwide. This includes enhancement and creation of specific test and deployment tools (installers, XCI,..). Test – Test the installation, i.e. running system that represents an NFV solution stack; as well as test specific aspects of a system. This includes creation and enhancement of specific test and deployment tools. In addition, this includes tooling for verification (OVP/CVP). Test – Test the installation, i.e. running system that represents an NFV solution stack; as well as test specific aspects of a system. This includes creation and enhancement of specific test and deployment tools. In addition, this includes tooling for verification (OVP/CVP). Tool suite: Compose individual tools into a tool suite. Iterate/Automate – Create guidelines and tooling for automated deployment, testing, and test-results reporting. Iterate/Automate – Create guidelines and tooling for automated deployment, testing, and test-results reporting. Create a “DevOps” platform: Tooling to automatically compose the entire DevOps workflow using cloud services. (Require migration from DIY hosted labs/git/gerrit/Jenkins to cloud services like packet.net, github, circleCI, ..) Operate – Operate a set of servers/labs and services (git, gerrit, jenkins,..). Offer Lab-as-a-Service (LaaS). Operate – Operate a set of servers/labs and services (git, gerrit, jenkins,..). Offer Lab-as-a-Service (LaaS). OPNFV Release artifacts: * Scenarios – installable NFV solution stacks (NFVI); “NFVI Platform” * Tools (mostly test/operations tools) * OVP/CVP solution OPNFV Release artifacts: * Scenarios – installable NFV solution stacks (NFVI); streamlined, i.e. fewer scenarios; increased focus on CN and edge; “NFVI Platform” * Packaged testing/operations tool suite * OVP/CVP solution * “DevOps Platform” -----Original Message----- From: opnfv-...@lists.opnfv.org <opnfv-...@lists.opnfv.org> On Behalf Of Georg Kunz Sent: Dienstag, 27. November 2018 18:46 To: HU, BIN <bh5...@att.com>; Tim Irnich <tim.irn...@suse.com>; Trevor Bramwell <tbramw...@linuxfoundation.org> Cc: AshYoung <a...@cachengo.com>; opnfv-tech-discuss@lists.opnfv.org; opnfv-...@lists.opnfv.org; Manuel Buil <mb...@suse.com> Subject: Re: [opnfv-tsc] Discussion of OPNFV Strategic Plan Hi Bin, Hi all, Due to the lively discussions during today's TSC call, the IRC minutes are a little light [1]. However, I have to voice my concern that I cannot agree with the following items: [.] 14:41:31 <bh526r> #info Vote for strategy on Tuesday Dec 4 14:42:01 <bh526r> #info Hopefully everyone will agree 14:43:12 <bh526r> #info We need a decision on Dec 4 in order to trigger following actions 14:43:35 <dmcbride> #topic budget discussion 14:43:45 <bh526r> #info Stalemate is not an option [.] I don't understand why "we need a decision by Dec 4 in order to trigger actions". I seriously appreciate your ambition to move this forward quickly as the main intention is to strengthen OPNFV's position. However, I also don't see why concrete actions are being blocked if there is no decision on Dec 4. A core value of open source communities is that those who are interested in a particular topic, naturally tend to form a group which jointly works towards a common goal. In our concrete scenario, we could i) form a devops working group which works on fleshing out the details of the proposal, and/or ii) find a group of interested people prototyping some of the "cloud-based devops methodologies. None of such activities would be considered a stalemate. The results of such _community-driven_ activities would help to convince the entire community. A very successful example in this regard is XCI, which was driven by a small group of people. Certainly, it is the job of all TSC members to actively participate in the strategy definition and discussion and I urge everybody to do so. An open source community works best if it is driven by personal motivation. For sure it does not work well if deadlines for decisions about unclear directions are put on a community without a clear understanding why. That said, my current view on the proposal is the following: it broadens the scope of the community (by a currently undefined amount), i.e., it adds on top of what we are currently doing. I do not think that this is the right approach given shrinking amounts of resources in the community - both in terms of developers and funding. I believe we need to instead discuss, as an alternative, if we should and can focus on a very specific, well-defined and sought-after contribution to the ecosystem. I mentioned this in a previous email already: based on input from stakeholders, I would argue for strengthening the reference platform (as defined through comprehensive tests) and the corresponding compliance program. This is my perspective for sure - others might disagree and I'd love to discuss better proposals. [1] http://ircbot.wl.linuxfoundation.org/meetings/opnfv-meeting/2018/opnfv-meeting.2018-11-27-13.54.log.txt Best regards Georg -----Original Message----- From: HU, BIN <bh5...@att.com<mailto:bh5...@att.com>> Sent: Tuesday, November 27, 2018 2:22 PM To: Tim Irnich <tim.irn...@suse.com<mailto:tim.irn...@suse.com>>; Trevor Bramwell <tbramw...@linuxfoundation.org<mailto:tbramw...@linuxfoundation.org>> Cc: AshYoung <a...@cachengo.com<mailto:a...@cachengo.com>>; Georg Kunz <georg.k...@ericsson.com<mailto:georg.k...@ericsson.com>>; opnfv-tech-discuss@lists.opnfv.org<mailto:opnfv-tech-discuss@lists.opnfv.org>; opnfv-...@lists.opnfv.org<mailto:opnfv-...@lists.opnfv.org>; Manuel Buil <mb...@suse.com<mailto:mb...@suse.com>> Subject: RE: [opnfv-tsc] Discussion of OPNFV Strategic Plan Thank you for pointing out one possibility based on the assumption that the same resources will do both work. The assumption itself may not be true because there will be different resources to do different work in different projects (which is the reality today). So the resource availability is a key factor to consider when we approve the new projects subsequently after we plan the product portfolio. When we have dedicated resources to do each job, such possibility will be unlikely to happen. Thanks Bin -----Original Message----- From: opnfv-...@lists.opnfv.org<mailto:opnfv-...@lists.opnfv.org> <opnfv-...@lists.opnfv.org<mailto:opnfv-...@lists.opnfv.org>> On Behalf Of Tim Irnich Sent: Monday, November 26, 2018 11:59 PM To: HU, BIN <bh5...@att.com<mailto:bh5...@att.com>>; Trevor Bramwell <tbramw...@linuxfoundation.org<mailto:tbramw...@linuxfoundation.org>> Cc: AshYoung <a...@cachengo.com<mailto:a...@cachengo.com>>; Georg Kunz <georg.k...@ericsson.com<mailto:georg.k...@ericsson.com>>; opnfv-tech-discuss@lists.opnfv.org<mailto:opnfv-tech-discuss@lists.opnfv.org>; opnfv-...@lists.opnfv.org<mailto:opnfv-...@lists.opnfv.org>; Manuel Buil <mb...@suse.com<mailto:mb...@suse.com>> Subject: Re: [opnfv-tsc] Discussion of OPNFV Strategic Plan The way I understand Trevor's concern is that if we start spending more time on packaging tools and supporting their usage downstream, there will be less time for doing integration work and driving upstream production readiness. Which is something I'm concerned about too. Pretending that this problem doesn't exist isn't helpful IMHO. Tim On 11/27/18 2:11 AM, HU, BIN wrote: > Trevor, > > Thank you for you clarifying it. > > The integration work is explicitly mentioned to be continued in 3rd bullet on > slide #13 of v0.8. I am attaching it again just in case you missed it. That > work will continue as usual. All related bug fixes and new features in > upstream will continue as usual too. So I am not sure why it is a concern > here. > > Regarding the concern of spending our time to help people use our tools, > isn't it the usual business we are supposed to do today? For example, after > we release Gambia, we are supposed to help people use it, right? There is a > "opnfv-user" mailing list for this purpose. There isn't much traffic though. > It means either everyone is an expert or no one is interested in using our > release. I wish it was because everyone is an expert, though the reality > might be opposite. > > Recently, someone asked me how to run Yardstick on Dovetail. Thanks Georg for > sharing the docs. I was really excited because finally someone is interested > in using our tool. So getting user to use our tools is exactly what we want, > right? Without users, I don't know how to show others our value, frankly. > > So IMHO, spending our time to help user isn't a concern at all. It is what we > need. And there is no difference of supporting users, e.g. use OpenStack by > OpenStack community, use ODL by ODL community. Etc. > > If there is no user to support, we are in trouble because our deliverables > has no value. > > Let me know what you think, and if you still have concerns. > > Thank you > Bin > > -----Original Message----- > From: opnfv-...@lists.opnfv.org<mailto:opnfv-...@lists.opnfv.org> > <opnfv-...@lists.opnfv.org<mailto:opnfv-...@lists.opnfv.org>> On Behalf > Of Trevor Bramwell > Sent: Monday, November 26, 2018 4:54 PM > To: HU, BIN <bh5...@att.com<mailto:bh5...@att.com>> > Cc: Tim Irnich <tim.irn...@suse.com<mailto:tim.irn...@suse.com>>; AshYoung > <a...@cachengo.com<mailto:a...@cachengo.com>>; > Georg Kunz <georg.k...@ericsson.com<mailto:georg.k...@ericsson.com>>; Manuel > Buil <mb...@suse.com<mailto:mb...@suse.com>>; > opnfv-tech-discuss@lists.opnfv.org<mailto:opnfv-tech-discuss@lists.opnfv.org>; > opnfv-...@lists.opnfv.org<mailto:opnfv-...@lists.opnfv.org> > Subject: Re: [opnfv-tsc] Discussion of OPNFV Strategic Plan > > Hi Bin, > > Perhaps 'integrated' is a better word here than 'supported'. A lot of the > work in OPNFV involves integrating many of these upstream components which in > turn exposes bugs, or creates features that enable an NFV use case. > > I'm quite terrible with examples, but I'm sure others from the community have > time. > > Regards, > Trevor Bramwell > > On Tue, Nov 27, 2018 at 12:26:33AM +0000, HU, BIN wrote: >> Trevor, >> >> Thank you for your question. >> >> Can you give more details and examples of "doing what we're best at, which >> is getting NFV supported by upstream projects."? >> >> Thank you >> Bin >> >> -----Original Message----- >> From: Trevor Bramwell >> <tbramw...@linuxfoundation.org<mailto:tbramw...@linuxfoundation.org>> >> Sent: Monday, November 26, 2018 4:17 PM >> To: HU, BIN <bh5...@att.com<mailto:bh5...@att.com>> >> Cc: Tim Irnich <tim.irn...@suse.com<mailto:tim.irn...@suse.com>>; AshYoung >> <a...@cachengo.com<mailto:a...@cachengo.com>>; >> Georg Kunz <georg.k...@ericsson.com<mailto:georg.k...@ericsson.com>>; Manuel >> Buil <mb...@suse.com<mailto:mb...@suse.com>>; >> opnfv-tech-discuss@lists.opnfv.org<mailto:opnfv-tech-discuss@lists.opnfv.org>; >> opnfv-...@lists.opnfv.org<mailto:opnfv-...@lists.opnfv.org> >> Subject: Re: [opnfv-tsc] Discussion of OPNFV Strategic Plan >> >> Hi Bin, >> >> I'm still unclear on the first point: "Enabling and automating stakeholders' >> business transformation into DevOps organization" >> >> From what I've read it seems like the suggestion is to package up everything >> that makes up OPNFV (Platform, CI/CD piplines, testing / verification / >> certification tools, etc.) and turn that into something that can be deployed >> by a company internally. >> >> Is that what is being suggested here, or something else? And if so I'd be >> concerned that we'd actually be reducing companies incentive to be involved, >> or more of our time would be spent trying to support people using the tool >> then doing what we're best at, which is getting NFV supported by upstream >> projects. >> >> Regards, >> Trevor Bramwell >> >> On Mon, Nov 26, 2018 at 09:04:57PM +0000, HU, BIN wrote: >>> Tim, >>> >>> Not sure if you get a chance to follow the most recent discussion. >>> >>> The ask is merely to agree on a strategy (i.e. the vision and direction) >>> outlined on Slide #13, supported by the steps of actions summarized on >>> slide #16. See attached the most recent update v0.8. >>> >>> Please let me know if there is anything unclear here. >>> >>> Thanks >>> Bin >>> >>> -----Original Message----- >>> From: opnfv-...@lists.opnfv.org<mailto:opnfv-...@lists.opnfv.org> >>> <opnfv-...@lists.opnfv.org<mailto:opnfv-...@lists.opnfv.org>> On >>> Behalf Of Tim Irnich >>> Sent: Monday, November 26, 2018 12:57 PM >>> To: HU, BIN <bh5...@att.com<mailto:bh5...@att.com>>; AshYoung >>> <a...@cachengo.com<mailto:a...@cachengo.com>>; Georg >>> Kunz <georg.k...@ericsson.com<mailto:georg.k...@ericsson.com>>; Manuel Buil >>> <mb...@suse.com<mailto:mb...@suse.com>> >>> Cc: >>> opnfv-tech-discuss@lists.opnfv.org<mailto:opnfv-tech-discuss@lists.opnfv.org>; >>> opnfv-...@lists.opnfv.org<mailto:opnfv-...@lists.opnfv.org> >>> Subject: Re: [opnfv-tsc] Discussion of OPNFV Strategic Plan >>> >>> On 11/26/18 4:40 PM, HU, BIN wrote: >>>> >>>> If I understand correctly, Point #1 and #3 are actually the same question, >>>> i.e. what will we do in the next step? >>> >>> No, I'm rather suggesting to make sure our understanding is complete before >>> we proceed. We clearly do not yet sufficiently understand what exactly the >>> decision is you're asking us to take, so we cannot proceed. >>> Let's continue to work on this until we have the required clarity, and then >>> decide. >>> >>> Regards, Tim >>> >> >> >>> -=-=-=-=-=-=-=-=-=-=-=- >>> Links: You receive all messages sent to this group. >>> >>> View/Reply Online (#4856): >>> https://urldefense.proofpoint.com/v2/url?u=https-3A__lists.opnfv.org >>> _g_opnfv-2Dtsc_message_4856&d=DwIF-g&c=LFYZ-o9_HUMeMTSQicvjIg&r=6qPc >>> DOqMgwf1K_r6YIIHhw&m=MsGFeXeJn1GRo-GvLejVXrmPgRHvZKjmVkiBKZgaQdc&s=j >>> 9hLZ3q9g0pHbtq-b6cZkh4PaKLsKtMkaWRRHHHAcqQ&e= >>> Mute This Topic: >>> https://urldefense.proofpoint.com/v2/url?u=https-3A__lists.opnfv.org >>> _mt_27802341_557206&d=DwIF-g&c=LFYZ-o9_HUMeMTSQicvjIg&r=6qPcDOqMgwf1 >>> K_r6YIIHhw&m=MsGFeXeJn1GRo-GvLejVXrmPgRHvZKjmVkiBKZgaQdc&s=fTo_-Z8GU >>> Aaz9sCAJyClb_m_LGWxF3_23Siiy8SJdtY&e= >>> Group Owner: >>> opnfv-tsc+ow...@lists.opnfv.org<mailto:opnfv-tsc+ow...@lists.opnfv.org> >>> Unsubscribe: >>> https://urldefense.proofpoint.com/v2/url?u=https-3A__lists.opnfv.org >>> _g_opnfv-2Dtsc_unsub&d=DwIF-g&c=LFYZ-o9_HUMeMTSQicvjIg&r=6qPcDOqMgwf >>> 1K_r6YIIHhw&m=MsGFeXeJn1GRo-GvLejVXrmPgRHvZKjmVkiBKZgaQdc&s=f8xgHpaw >>> JHb8E2ELrIpuKGYOIZmFHT6fOJf7huGMVHM&e= >>> [tbramw...@linuxfoundation.org] >>> -=-=-=-=-=-=-=-=-=-=-=- >> -- Dr.-Ing. Tim Irnich, Senior Program Manager Developer Engagement E-Mail: tim.irn...@suse.com<mailto:tim.irn...@suse.com> Mobile: +49 172 2791829 SUSE Linux GmbH, GF: Felix Imendörffer, Jane Smithard, Graham Norton, HRB 21284 (AG Nürnberg)
-=-=-=-=-=-=-=-=-=-=-=- Links: You receive all messages sent to this group. View/Reply Online (#22477): https://lists.opnfv.org/g/opnfv-tech-discuss/message/22477 Mute This Topic: https://lists.opnfv.org/mt/28277855/21656 Group Owner: opnfv-tech-discuss+ow...@lists.opnfv.org Unsubscribe: https://lists.opnfv.org/g/opnfv-tech-discuss/unsub [arch...@mail-archive.com] -=-=-=-=-=-=-=-=-=-=-=-
