2022-10-28 Project Stratos Meeting notes


Oct 28, 2022


  • @Alex Bennée - OOF

  • @Mike Holmes

  • @Mathieu Poirier

  • @Anmar Oueja

  • @Randy Linnell

  • @Bill Mills

  • @Vincent Guittot

  • @Leonardo Garcia

  • @Don Harbin


  • Feedback for the next weeks TSC retrospective

    • What went well

      • rust-vmm daemons?

      • bare metal backend with limited memory sharing

    • What didn’t got well

      • engagement?

 Discussion topics

  • MH: We had a lot of grand plans which we didn’t get to but what was done was a success

  • MH: We need to pick something that feeds into a wider Blueprint

    • MP: difficult to get engaged without an obvious finish line

    • MP: couldn’t do everything as we were short of resources

    • RL: I agree focus on tangible, focus on specifics

    • MH: ST have a proposal to connect remoteproc with VirtIO, ARM and EWOL VirtIO stuff

    • BM: lots of places where the tech can be extended, find an actual use-case. GPU would be a big effort if no one was going to use it

    • AO: meeting time may have been a factor, we should show leadership picking a particular use case

    • BM: hypervisor-less VirtIO is an ambiguous name (Different things to different people)

    • BM: Take GPIO/I2C or Block backends on an R5 for example

    • BM: is it virtualisation or VirtIO? Not so sure about the hypervisor independence

    • AO: portability is a goal but maybe only 20% is portable atm

    • LG: AGL want VirtIO for hypervisorless for emulation (cloud native vs deploy to HW)

    • BM: 80% portability is a worthwhile goal

    • AO: what does it actually take to enable

    • LG: AGL is far from its goal, not progressing fast enough - focus on more meaningful use cases

    • MP: we have lots of drivers in kernel, bus driven devices tend to be interoperable

    • MP: how do we pick the winners from the herd of problems and use cases

    • AO: I don’t think there are bigger fish solving these VirtIO problems

    • BM: Stratos was chasing the shiny problems but plenty of basic stuff to solve

    • BM: was hoping we would have a better use case definitions from Arm by now

    • RL: members looking for defined timelines for projects, pick and be definitive

    • MP: but the picking is the problem

    • MH: and a division between cloud native and HPP

    • AO: and the remaining foundational work blocking other stuff?

    • RL: can we show progress and attract members

    • MH: maybe we should spend a bit of time on Bills end

    • AG: Cloud Native not really happening

    • BM: look for a 2 person project rather than a 5 person one where we only work with member engineers

    • MH: should we push a fresh pitch

    • AO: agree - lets push on the BPs rather than under the umbrella of Stratos

    • MP: we can extend on the Stratos work even if we don’t work under that

  • What next - discussion became Linaro only internal, and the results will be published when this bottoms out

  • Broadly

    • Pivot Stratos towards possibly Arm-sponsored external contribution of virtio work to meta-eawol / TRS

    • Take LEDGE member lead BP from ST, attractive to Xilinx, Arm and NXP and look at HPP usecases