Re: Virtualization project idea

["Pasi Kärkkäinen" <pasik@iki.fi>]

On Fri, Aug 27, 2010 at 07:05:11PM +0530, Dhananjay Goel wrote:
>    If you don't want to do it over-the-network, but somehow 'through' the
>    hypervisor,
>    then you'd need to build some kind of special filesystem protocol that is
>    able
>    to do the client-server communication over the hypervisor specific paths
>    (xenbus etc).
>    And have drivers for it in the hypervisor-host, and in the VMs.
>    --So, is it possible to build a filesystem protocol to share USB devices?
>    Is my project feasible?
> 

Yes, you can build some filesystem protocol and share ANY filesystem, 
be it USB, SATA, SAS, RAID, whatnot.

If it's filesystem protocol it doesn't matter at all if it's USB or not.

-- Pasi

>    Thanks,
>    Dhananjay
> 
>    On Fri, Aug 27, 2010 at 7:00 PM, Pasi K*rkk*inen <[1]pasik@iki.fi> wrote:
> 
>      On Fri, Aug 27, 2010 at 03:17:39PM +0200, Michal Novotny wrote:
>      > On 08/27/2010 03:08 PM, Pasi K*rkk*inen wrote:
>      >> On Fri, Aug 27, 2010 at 06:33:35PM +0530, Dhananjay Goel wrote:
>      >>
>      >>> * * Yes, exactly. So, we wanted to know if it is possible to *share
>      USB*
>      >>> * * across VMs.
>      >>>
>      >>>
>      >> I don't think USB protocol has been designed for *sharing*.
>      >> I'm pretty certain only one computer/device/VM can use USB device at
>      a time.
>      >>
>      >> -- Pasi
>      >>
>      >>
>      >
>      > Pasi, I agree. I think the think here is that Dhananjay confused the
>      USB
>      > device sharing with the file system sharing. I guess the USB protocol
>      > was not designed for sharing nevertheless sharing the filesystem on a
>      > USB stick is a completely different think.
>      >
>      > Dhananajay, you need to plug in the USB stick onto one computer (and
>      > it's impossible to plug it into multiple computer at one time, of
>      > course) and then setup the sharing. Everybody here is talking about
>      the
>      > hardware abstraction and virtualization and what you wrote is a
>      > completely different thing - it's software-related and this has
>      nothing
>      > to do with the hardware emulation/abstraction what-so-ever.
>      >
>      > Considering the NFS and all the sharing protocols there was something
>      > why it doesn't corrupt the data. I'm no expert on this subject but I
>      > think this is because they run in the server-client mode. All the
>      > clients are talking to the server and the server itself is one
>      computer
>      > that's having the just one operating system working with this
>      particular
>      > device - no matter what the underlaying device is - it may be
>      everything
>      > - USB stick, IDE/SCSI/SAS drive or just a relay workstation to save
>      all
>      > the data into one remote media (e.g. for replication). What I mean is
>      > that the basic thing is that it's running on only one operating system
>      > (because of it's connected to this one machine *only*) so it takes
>      care
>      > of everything and it's aware of the write-cache and data operations
>      > being done to this media.
>      >
> 
>      Yep.
> 
>      If you want to share files from the hypervisor-host (from USB stick or
>      from actual disk)
>      to the VMs you *can* do it today over-the-network (nfs,cifs,webdav,ftp)
>      using the standard
>      client-server tools that have been used for over 20 years.
> 
>      If you don't want to do it over-the-network, but somehow 'through' the
>      hypervisor,
>      then you'd need to build some kind of special filesystem protocol that
>      is able
>      to do the client-server communication over the hypervisor specific paths
>      (xenbus etc).
>      And have drivers for it in the hypervisor-host, and in the VMs.
>      -- Pasi
> 
> References
> 
>    Visible links
>    1. mailto:pasik@iki.fi