Re: [Publicity] Media Opp: Sean Michael Kerner - Xen Project 4.4

[George Dunlap <george.dunlap@xxxxxxxxxxxxx>]

On 03/08/2014 02:20 AM, Konrad Rzeszutek Wilk wrote:
> Q3:
>
> Xen has an ParaVirtualization mode (PV) for guest wherein there are no emulated 
> devices present. It eliminates extra overhead and minimizes the attack vector. 
> The PVH takes PV mode and instead of the hypervisor doing most of the kernel 
> bookkeeping the hardware extensions (EPT, NPT) take care of it. The end result 
> is that the overhead of doing this in an HVM (Hardware Virtual Machine) 
> container is much lower providing lower latency and higher throughput for 
> guests.
Xen's major contribution to the field of virtualization was 
paravirtualization (PV): the idea of making virtualization easier by 
replacing a hardware interface with a software interface.  Xen was 
designed before hardware virtualization (HVM) extensions were available 
for x86, so the default mode, PV mode, not only paravirtualizes the boot 
process and emulated hardware devices, but also privileged CPU 
operations, like pagetable updates, IDT tables, and so on.  Getting rid 
of emulated devices still makes sense, but with HVM extensions now 
available on basically all modern processors, paravirtualizing CPU 
operations no longer makes sense. PVH (ParaVirtualized with HVM 
extensions) attempts to make the best of both worlds -- it uses HVM 
extensions when that makes the most sense, and paravirtualization when 
it makes the most sense.
> Q1: ARM had been marked as experimental and now that it is a stable support 
> along with a set in stone ABI. This means that going forward kernels are 
> guaranteed that the communication to the hypervisor will be supported in this 
> fashion for years to come.
I would probably say:

The ARM platform has been maturing rapidly.  Previously it was marked 
"tech preview", meaning that the ABI for the guests might change in 
response to development needs.  Now the Xen Project team has committed 
to a stable ABI -- meaning operating systems designed to run on Xen 4.4 
will be supported for all future versions of Xen. A number of new ARM 
platforms are now supported, and the addition of new platforms is easier 
than ever.  There have also been a large number of functional, 
performance, stability, and maintainability improvements since the last 
release.
> Q2: The event channel changes (also known as FIFO events) revamp how events 
> (aka interrupts but much more lightweight) are delivered and processed. The end 
> result is that we can easily launch thousands (and more) of guests on one host. 
> That translates to higher density with faster response time.
>
> Q6: Xen 4.5 yes and the bugfix branches of Xen 4.4.x, Xen 4.3.x and Xen 4.2.x.
>
> Q5:  Put more pressure to make the release on time and with more community 
> helping in testing and reporting bugs.
On a day-to-day basis, I don't think LF has changed much the way we do 
development.  We need to be more predictable about releases, it's true.  
I think the main thing that has changed is having the XenProject 
advisory board.  Because companies have more of an investment in Xen, 
they are more likely to contribute developers.
Q4:

XenProject is a community developed open-source project, which means 
that anyone who is willing to invest the time to code can change the 
direction of the project.  The SPICE support is entirely at the 
initiative of a very dedicated user who wanted to use SPICE support 
himself.
Like a lot of KVM features, SPICE is actually in qemu -- and qemu is 
actually shared between KVM and Xen.  A number of features in KVM could 
be used by Xen currently by people willing to do get their hands dirty 
typing arcane runes into their config files.  All it took to make SPICE 
a properly supported feature was someone willing to add the "plumbing" 
to make the SPICE interfaces in qemu useable through the officially 
supported Xen interface without the arcane runes.
This demonstrates two very powerful advantages of open-source: sharing 
of code, and the ability of any motivated person or organization to 
affect the direction of the project.
 -George

_______________________________________________
Publicity mailing list
Publicity@xxxxxxxxxxxxxxxxxxxx
http://lists.xenproject.org/cgi-bin/mailman/listinfo/publicity