Re: [MPTCP] Call for ideas for a presentation about MPTCP Upstream project at NetDev 0x12 in July

[Christoph Paasch <cpaasch at apple.com>]

[-- Attachment #1: Type: text/plain, Size: 20195 bytes --]

Hello,

On 30/04/18 - 09:05:41, Mat Martineau wrote:
> 
> On Mon, 30 Apr 2018, Matthieu Baerts wrote:
> 
> > Hi Mat,
> > 
> > On Mon, Apr 30, 2018 at 5:37 PM, Mat Martineau <mathew.j.martineau(a)linux.intel.com> wrote:
> > 
> >       On Sun, 29 Apr 2018, Matthieu Baerts wrote:
> > 
> >             Hi Mat,
> > 
> >             Thank you for your review and input!
> > 
> >             On Sat, Apr 28, 2018 at 2:44 AM, Mat Martineau <mathew.j.martineau(a)linux.intel.com> wrote:
> > 
> >                   Hi Matthieu -
> > 
> >                   On Fri, 27 Apr 2018, Matthieu Baerts wrote:
> > 
> >                         Hello,
> > 
> >                         Following yesterday's discussion about NetDev 0x12, here is a proposition of mail to send to NetDev
> >             committee. I
> >                         already put some comments. Note that I have followed the submission guidelines from:
> >                         https://www.netdevconf.org/0x12/submit-proposal.html#proposals
> > 
> >                         Please comment it before Monday morning. Sorry for the rush but the deadline is the 1st of May :)
> > 
> >                               * Name(s) of the submitter(s): Christoph Paasch (Apple), Mat Martineau
> > 
> >                         (Intel), Peter Krystad (Intel), Ossama Othman (Intel) and myself, Matthieu Baerts (Tessares)
> > 
> >                         I wrote down the names of people who participated in the discussion in the ML and during the weekly
> >             meetings. I can
> >                         add more people if more people would like to join the preparation and presentation of this tutorial.
> >                         @Christoph, Mat, Peter, Ossama: can I write your names there?
> > 
> > 
> >                   Ossama won't be able to attend. It looks like most sessions have 1 or 2 presenters, so I don't know if 4 is too
> >             many. I'm sure
> >                   they'll give us some feedback if they want to limit the number of presenters.
> > 
> > 
> >             OK thank you!
> >             Yes indeed, I am sure they will say something if we are too many.
> > 
> >                               * Title of the submission: MPTCP: from the basic to an upstreamable base
> > 
> > 
> >                         We certainly need a better title, please comment!
> > 
> > 
> >                   "Multipath TCP: Present Use Cases and an Upstream Future" ?
> > 
> > 
> >             I was trying to find something that catch the attention but it is maybe not needed for these kind of presentation.
> > 
> >                         > * Label (one of moonshot, nuts'n bolts, hands-on): hands-on
> > 
> >                         "hands-on" seems to be the correct one according to the Submission Types.
> >                         https://www.netdevconf.org/0x12/submit-proposal.html#types
> > 
> > 
> >                   It's a little bit of "nuts & bolts" too, but I agree that hands-on is the better match.
> > 
> > 
> >             Thank you!
> > 
> >                               * Submission type (one of talk, presentation, workshop): tutorial,
> > 
> >                         instructor-led sessions (minimum 1 hour long and not to exceed 1.5 hours. The instructor will go over the
> >             technology
> >                         either through code review or execution and interact with the attendees.)
> > 
> >                         I guess there is a typo here: presentation should be replaced by tutorial in the guidelines.
> > 
> >                               Estimate of length of time for presentation: 1h
> > 
> > 
> >                         That's what we agreed yesterday but I can change.
> > 
> >                               Affiliations of submitters (needed for conflict of interest check): Apple,
> > 
> >                         Intel, Tessares
> > 
> >                               Description of proposal:
> >                                   A project to add an implementation of the MultiPath TCP protocol to the
> > 
> >                         Linux kernel is in progress by a small community. The goal of this tutorial is to discover what is this
> >             TCP extension
> >                         (RFC 6824), what are the different use-cases already in production by some companies and what are the
> >             challenges to
> >                         upstream MPTCP. We hope having interactive discussions and getting feedback from experienced developers
> >             will help us
> >                         in this task of easily bringing MPTCP to all Linux users, a technology already used by millions of
> >             people.
> > 
> > 
> >                                   In a bit more detail, we will start with a basic introduction of MPTCP.
> > 
> >                         A few use-cases will be presented with a demo to explain how useful this protocol is in today's Internet
> >             and how it
> >                         can be extended with API's like Netlink and BPF.
> > 
> > 
> >                   I recognize how Netlink is associated with a userspace path manager, but what's the BPF extensibility you're
> >             referring to?
> > 
> > 
> >             It was only to mention that it would be possible to extend MPTCP with eBPF.
> >             Even if it is not already available in the current Open-Source implementation, there is already the possibility to
> >             get a version with a
> >             programmable scheduler: https://progmp.net
> > 
> > 
> > If we don't have something specific to our effort with BPF, I'm not sure it belongs in the summary. But I don't have a strong opinion
> > about it.
> > 
> > 
> > We are saying that "the [Open-Source] implementation can currently be extended with netlink and BPF". It can come later :)
> > 
> >                         Then we will have some explanations about how MPTCP is currently implemented. This current implementation
> >             is quite
> >                         intrusive and that is certainly not something we would like to have upstream. We would like to express
> >             what we have
> >                         in mind to change that, with some samples and initiate discussions.
> > 
> > 
> >                   My edit of the above:
> > 
> >                   """
> > 
> >                   A community project is underway to add Multipath TCP to the upstream Linux kernel. This tutorial will introduce
> >             the audience to
> >                   this TCP extension (RFC 6824), show some use cases already in production, and discuss the challenges in
> >             converging on an upstream
> >                   MPTCP implementation.
> > 
> > 
> >             Should we insist on the "discussion" part? Like saying: "discuss with the audience"? "have interactive discussions"?
> >             It is maybe not needed but it was mainly to express the interactivity, it is a tutorial, not just a "simple"
> >             presentation.
> > 
> > 
> >       I think it's ok to mention that - while I left the word "discuss" in there for that reason, maybe that's not obvious or clear
> >       enough.
> > 
> > 
> > It was maybe me, maybe no need to insist :-)
> > 
> >                   We will use the current MPTCP implementation to demonstrate the utility of the protocol on today's internet,
> >             and to show how this
> >                   implementation can currently be extended with netlink and BPF. This not only has practical application for
> >             deploying MPTCP now,
> >                   but also illustrates how the APIs and code will need to evolve in order to properly coexist with the optimized
> >             Linux TCP core we
> >                   all rely on. We will discuss our ideas for bringing MPTCP to the upstream kernel so the technology is available
> >             to all Linux
> >                   users.
> > 
> > 
> >             To "attract" people, should we mention that the current implementation is already used by millions of users?
> >             I like how your improve the last bit :-)
> > 
> > 
> >       Sure. I liked that in your text, but ran out of time figuring out how to fit it in to the last sentence :)
> > 
> > 
> > Could we simply say:
> > 
> > We will use the current MPTCP implementation to demonstrate the utility of the protocol already used by millions of users on today's Internet
> 
> How about "used by millions of devices"? I think it flows a little better
> than "used by ... users".

+1 on the "used by millions of devices".


Christoph

> 
> 
> Thanks,
> 
> Mat
> 
> > 
> > or:
> > 
> > and to show how this implementation - already used by millions of users - can currently be extended with Netlink and BPF.
> > 
> > 
> >                   """
> > 
> >                   Feel free to edit/merge/expand/discard as needed :)
> > 
> > 
> >             Thank you for your edit, it is indeed cleaner!
> > 
> >                                   For more information about this project:
> > 
> >                         https://github.com/multipath-tcp/mptcp_net-next/wiki
> > 
> > 
> >             I guess I can keep this, right?
> > 
> > 
> >       Yes, that was my intent.
> > 
> > 
> > Will do!
> > 
> > Thank you!
> > 
> > Matthieu
> > 
> > 
> > 
> > 
> >       Thanks,
> > 
> >       Mat
> > 
> > 
> > 
> >                         Please feel free to comment this as well. We are still far from the max 350 words limit we found last
> >             time. But on
> >                         the other hand, I can no longer find this limit on their website :)
> > 
> > 
> >                   Hope the above is helpful. Thanks again for your work on this proposal.
> > 
> > 
> >             Yes it is, thank you for your help!
> > 
> >             Matthieu
> > 
> > 
> > 
> >                   Mat
> > 
> > 
> > 
> >                         Thank you for your help!
> > 
> >                         Have a good day/evening,
> >                         Matthieu
> > 
> >                         On 25/04/2018 21:58, Mat Martineau wrote:
> > 
> >                               Hi Matthieu -
> > 
> >                               On Fri, 20 Apr 2018, Matthieu Baerts wrote:
> > 
> >                                     Hello,
> > 
> > 
> >                                     NetDev 0x12 is coming to Montréal this summer: July 11th to 13th, 2018.
> > 
> > 
> >                                     We already talked about this event on this ML and at our weekly meetings but here is a
> >                                     summary of the discussions we had:
> > 
> >                                      - we would like to have a presentation there mainly to get feedback and advice from other
> >                                     kernel developers
> >                                      - a presentation would clearly indicate that this MPTCP Upstream project exists and we could
> >                                     get help from more developers
> >                                      - we would like to indicate that having MPTCP upstream is asked by different companies, some
> >                                     are even ready to contribute ; it is then important to have MPTCP upstream
> > 
> > 
> >                                     Also note that:
> > 
> >                                      - David Miller will not be present in Montréal [1] but other main contributors should be
> >                                     there (we don't have a list)
> > 
> > 
> >                               Side note, in the past day David reiterated his statement about not attending or supporting the
> >                               conference:
> > 
> >                               https://marc.info/?l=linux-netdev&m=152466827203301&w=2
> > 
> >                                      - A presentation by Octavian Purdila about "MPTCP Upstreaming" has already been given in
> >                                     2015 (NetDev 0.1) [2]
> >                                      - 3 types of presentation are available: talks, tutorials and workshops [3]
> >                                      - Call for Presentation Proposals closes on May 1st, 2018.
> > 
> > 
> >                                     The current idea we briefly discussed during our weekly meetings is to give a tutorial:
> > 
> >                                      - It is not useful to give almost the same presentation as the one of Octavian
> >                                      - It will allow us more flexibility somehow to explain what is MPTCP, the different
> >                                     use-cases, why it is important to have it upstream and what problems we are currently facing.
> >                                      - David Miller and many other kernel developers will go to LPC in November: a good place to
> >                                     give a talk this time.
> > 
> > 
> >                                     Do you have any ideas on what we could show in this tutorial?
> > 
> > 
> >                                     I recently discussed with my colleague Olivier Bonaventure who has a lot of experiences in
> >                                     giving different introductions and more about MPTCP and here is what he suggests:
> > 
> >                                      - A first part about a basic introduction of MPTCP
> >                                      - Indicate different use-cases -- if possible with a "closed demo" to be sure it is working
> >                                     -- asking people to setup something is not easy in 1h, max 1h30.
> > 
> > 
> >                                From the description at [3], either "instructor-led" 60-90 minute tutorials or
> >             "student-participation"
> >                               2-3 hour sessions are possible. The closed demo maps well to their "instructor-led" category.
> >             Looking at
> >                               the schedule, the past two Netdev Conferences have had one tutorial each, of 60-70 minutes. I think
> >             it
> >                               helps to be closer to an hour in length to hold the audience's attention.
> > 
> >                                      - Then trying to have interactive discussions or explanations about how MPTCP is currently
> >                                     implemented or should be implemented if it goes upstream, e.g.: for MPTCP, we need to have
> >                                     extra TCP Options, we need to support middleboxes, we need to link subflows of the same
> >                                     connection together, we need a scheduler, a PM, etc.
> > 
> > 
> >                               It's difficult to predict how interactive an audience will be. I've only attended one Netdev
> >             Conference,
> >                               and it seemed like there were a lot more people with expertise and interest in drivers, lower
> >             layers
> >                               (XDP, BPF, TC, netfilter), and network topology/simulation. Discussion around middlebox support and
> >             the
> >                               userspace API might have more audience interaction. If we want to drive a discussion, we could try
> >             to
> >                               strike a balance between topics for the broader audience and those with more knowledge of TCP
> >             internals.
> >                               (Hopefully some TCP internals people are still planning to attend)
> > 
> >                                      - Of course, we should focus our discussions on the upstreaming aspect, e.g. reducing the
> >                                     footprint of MPTCP in the current TCP stack: what are we allow to do, what not. It is linked
> >                                     to many previous discussions we had on this ML, e.g. why we need more indirect function calls
> >                                     and how to reduce the impact, etc.
> > 
> > 
> >                               The previous talk ([2]) had a section like this. I haven't watched it recently, I should look at it
> >             again
> >                               to see what kind of questions the audience was asking. As you mentioned above we should be careful
> >             to
> >                               have new content compared to the previous session.
> > 
> >                                      - If we have time, we could discuss about how users could interact with MPTCP: enable it per
> >                                     connection, control the path manager, maybe the scheduler, etc.
> > 
> > 
> >                                     What do you think about this? Feel free to comment and even propose completely different
> >                                     ideas!
> > 
> > 
> >                               Thank you for outlining these ideas. I see that this topic is on our meeting agenda so it will be
> >             good to
> >                               discuss the tutorial there.
> > 
> > 
> >                                     [1] https://lists.01.org/pipermail/mptcp/2018-March/000379.html
> >                                     [2] https://www.youtube.com/watch?v=wftz2cU5SZs
> >                                     [3] https://www.netdevconf.org/0x12/submit-proposal.html
> > 
> > 
> >       --
> >       Mat Martineau
> >       Intel OTC
> > 
> > 
> > 
> > 
> > --
> > Tessares SA
> > Matthieu Baerts | R&D Engineer
> > matthieu.baerts(a)tessares.net
> > Tessares SA | Hybrid Access Solutions
> > www.tessares.net 1 Avenue Jean Monnet, 1348 Louvain-la-Neuve, Belgium
> > 
> > 
> > _____________________________________________________________________________________________________________________________________________
> > DISCLAIMER.
> > This email and any files transmitted with it are confidential and intended solely for the use of the individual or entity to whom they are
> > addressed. If you have received this email in error please notify the system manager. This message contains confidential information and is
> > intended only for the individual named. If you are not the named addressee you should not disseminate, distribute or copy this e-mail. Please
> > notify the sender immediately by e-mail if you have received this e-mail by mistake and delete this e-mail from your system. If you are not
> > the intended recipient you are notified that disclosing, copying, distributing or taking any action in reliance on the contents of this
> > information is strictly prohibited.
> > 
> 
> --
> Mat Martineau
> Intel OTC