Re: Prepping for GSOC 2014 [URGENT] - deadline Feb 14 2014

[Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>]

On Wed, Feb 05, 2014 at 02:09:08PM +0000, Ian Campbell wrote:
> On Tue, 2014-01-28 at 13:54 +0000, Lars Kurth wrote:
> > Hi all,
> > I have not gotten any reply to this thread. I saw Wei Lui and Andrés 
> > Lagar-Cavilla make changes to the project list. Please go through the 
> > items below and make changes as suggested. Otherwise, our chances to get 
> > into GSoC 2014 will be relatively slim.
> 
> Going through the list, people listed as technical contacts for projects
> with GSoC == yes (or unknown) are in the To line. Please reiterate your
> interest in mentoring the project(s) and update or remove the entry as
> necessary.
> 
> I skipped things added recently and I skipped "Xen Cloud Platform (XCP)
> and XAPI projects", someone else can pick that up.
> 
> And to reiterate what Lars said:
> 
> > Add new work items : we ought to have a few sexy topics on say 
> > Real-time, mobile and some of the other segments (assuming we can get
> > HW)
> [...]
> > b) Anyone who has some kernel/linux/bsd/distro/qemu work-items, should 
> > get these listed on the respective other programs. And we should link 
> > to these from our project page.
> 
> The list:
> 
> Pasi:
>       * Implement Xen PVSCSI support in xl/libxl toolstack
>       * Implement Xen PVUSB support in xl/libxl toolstack
> 
>         Both have unanswered questions posed by Lars in January 2013.
>         
> Konrad:
>       * Block backend/frontend improvements
> 
>         I suspect a bunch of these are done? (Also CC Roger, who may
>         have done them...)
>         
>         Was in the list twice, they looked identical so I nuked one.
>         

<nods>
>       * Utilize Intel QuickPath on network and block path.
> 
>         No comments etc, but sounds advanced for a GSoC student, plus
>         its unclear when such hardware became available, are they likely
>         to have it? It sounds like it might also be quite high end.
>         
>       * perf working with Xen
> 
>         Done/in progress by Boris I think

<nods>
>         
>       * PAT writecombine fixup
> 
>         Did I see a fix for this go past? GSoC == unknown?

No. Still looking for a victi^H^H^Hvolunteer.

>         
>       * Parallel xenwatch
> 
>         Bit sparse on details, GSOC == unknown

That would still be nice. It came from talking to Matt from Amazon. He
was saying that having only one xenwatch thread slows things down for
a medium to big server with lots of guests. Making multiple
xenwatch threads to process XenBus requests in parallel would be
a good improvement.

> 
>       * Microcode uploader implementation
>         
>         Done I think?

<nods>
> 
>       * Integrating NUMA and Tmem
>         
>         Lists Dan as co-maintainer -- Konrad do you want to propose this
>         to the new tmem guy (I've forgotten his name)

Bob Liu. Yes, lets rope him in.
>         
>       * Performance tools overhaul
> 
>         Bit vague. And has some of this been done?

I can't remember what that is.
> 
>       * "Upstream bugs"
> 
>         There were 4 of these, dating back to 2012, I don't think this
>         list is a good place to track bugs and it seems like at least
>         some of them are now obsolete. So I've nuked the lot. If they
>         are still relevant I think it would be best to get them into the
>         bug tracker.

OK, lets nuke them.

We could also add the:

VCPUOP_register_vcpu_time_memory_area support in Linux upstream kernel.

MSI multi-vector for Linux upstream kernel.

Thought those are mostly just putting pieces together and repost them
so no "new" development.

>         
> Ben:
>       * dom0 kgdb support
> 
>         Is this for GSoC?
> 
> George:
> 
>       * Introducing PowerClamp-like driver for Xen
> 
>         I don't think this has been done?
> 
> Dario:
>       * NUMA effects on inter-VM communication and on multi-VM workloads
> 
>         I think this was under way as part of the GNOME Outreach
>         program. In that case perhaps it needs updating to reflect what
>         has been done and what still needs to be done?
>         
>       * Integrating NUMA and Tmem
> 
>         Lists Dan as co-maintainer -- covered under Konrad's name above.
>         
>       * Is Xen ready for the Real-Time/Embedded World?
>                 
>         Sounds a bit blue sky? Now that there is active interest in this
>         on ARM perhaps a few concrete projects could be proposed to
>         replace it?
> 
> Andy:
> 
>       * IOMMU control for SWIOTLB, to avoid dom0 copy of all >4K DMA
>         allocations
> 
>         Sounds too hard for a GSoC to me. Would need fleshing out in any
>         case.
>         
>       * CPU/RAM/PCI diagram tool
> 
>         Does this not already exist somewhere?
>         
> Paul:
> 
>       * HVM per-event-channel interrupts
> 
>         Might be easier now that Windows PV drivers are opened up?
> 
> Roger:
>       * Refactor Linux hotplug scripts
> 
>         You did some of this I think?
>         
> Ian C:
> 
>       * XL to XCP VM motion
> 
>         Perhaps this could be broadened into VM transport between XL and
>         other things too -- e.g. libvirt?
> 
> Stefano:
> 
>       * VM Snapshots
> 
>         Still a good project I think
>         
> George:
> 
>       * Allowing guests to boot with a passed-through GPU as the primary
>         display
> 
>         This seems like a bit of a rathole for a GSoC student to me...
>         
>       * Advanced Scheduling Parameters
>         
>         Still to do?
> 
> Santosh:
> 
>       * KDD (Windows Debugger Stub) enhancements
>         
> Dave
> 
>       * Create a tiny VM for easy load testing
> 
>         Someone was looking at this I think?
>         
> Ian J:
> 
>       * Testing PV and HVM installs of Debian using debian-installer
>       * Testing NetBSD
> 
>         BSD is done I think, and I'm looking at Debian Installer stuff
>         myself. I've removed these.
> 
> 
> Phew!
> Ian.
> 
> 
> 
> > Lars
> > 
> > On 20/01/2014 09:18, Lars Kurth wrote:
> > > Hi all,
> > >
> > > the GSoC application deadline is coming up : Feb 2014. If we want to 
> > > have any chance of getting accepted this year, we ought to get our 
> > > project list into good shape. The project list and how the project and 
> > > menters present themselves has a bigger impact on whether we get 
> > > accepted than the actual application.
> > >
> > > Also, I would like to add a mentor section this year: a short bio, 
> > > what the mentor cares about and a picture. This will help make the 
> > > project list more real.
> > >
> > > We have *4 weeks* to do this. The bar for GSoC has been getting 
> > > increasingly high. I know, we are tied down with Xen 4.4, but this is 
> > > something you need to do if you want the Xen Project to participate.
> > >
> > > a) Please, update 
> > > http://wiki.xenproject.org/wiki/Xen_Development_Projects urgently 
> > > (these need to be in good shape *before* the application). What I need 
> > > you to do is:
> > > a.1) Remove items that are done
> > > a.2) Add new work items : we ought to have a few sexy topics on say 
> > > Real-time, mobile and some of the other segments (assuming we can get HW)
> > > a.3) All project proposals need to be peer reviewed *and* clear ... 
> > > The peer review process for projects we put in place last year worked 
> > > well, by which we had past mentors sign of project proposals that were 
> > > in good enough state.
> > >
> > > b) Anyone who has some kernel/linux/bsd/distro/qemu work-items, should 
> > > get these listed on the respective other programs. And we should link 
> > > to these from our project page.
> > >
> > > Best Regards
> > > Lars
> > > P.S.: I will also see whether we can participate as Xen Project under 
> > > the LF GSoC program, but last year there was push-back and I don't 
> > > expect this to change
> > >
> > 
> > 
> > _______________________________________________
> > Xen-devel mailing list
> > Xen-devel@lists.xen.org
> > http://lists.xen.org/xen-devel
> 
>