[Ovirt-devel] I'm all for it

Tue Dec 29 07:39:23 UTC 2009

> I'm not 100% sure I understand what Mr. Fortaine is getting at from the
> emails he has sent out.  (i.e. I'm not sure if he is driving at a
> particular usage model of oVirt Node or if he is advocating a BIOS HV
> implementation)
>    

We are advocating a BIOS HV implementation.This design proposal come from 3 engineering concerns :

a) Performance

Faulty firmware (BIOS or UEFI) implementations prevent efficient I/O
Virtualization. To quote David Woodhouse, lead Intel embedded software
developer for the Linux Kernel and principal maintainer of the file
drivers/pci/intel-iommu.c  [3] ( Intel's I/O Virtualization, to
provide high performance I/O inside the virtual machines ) [4] :

"Well done, Dell and HP -- although I didn't think it was possible,
you have _further_ lowered my already-unprintable opinion of closed
source BIOSes and BIOS engineers.

We _really_ need open source firmware.

Or at _least_ firmware written by competent engineers -- but I think
we've all fairly much given up on that happening by now?"

Because many commits of the file intel-iommu.c are related to BIOS
bugs [5] [6] [7] [8] [9] [10].

And it is evident that Intel will not be able to test each firmware
version on each IOMMU capable Hardware.

By the way, we would greatly appreciate to invite you to a further
reading of this Phoronix article entitled : "Intel Core i7
Virtualization Performance" [11] on a X58 chipset (with IOMMU support,
according to the Xen wiki) [12] :

"Linux's Kernel-based Virtual Machine performed quite well in a number
of tests when compared to the host OS speed and VirtualBox, but
particularly when it came to the disk-related tests KVM did not fair
well at all."

The engineering target is to achieve 1:1 parity in terms of
performance between real-hardware and a virtual machine.

By ensuring a full use of hardware capabilities inside the Virtual
Machines, through a custom Virtualization Firmware, it lowers the TCO
(Total Cost of Ownership), thus requiring less frequent hardware
upgrades that can be substantial saves on a large scale volume or a
critical requirement for budget aware customers.

b) Out-of-the-box support

By compiling a custom Linux Kernel inside the Firmware, we can ensure
full Hardware Support, skipping painful devices incompatibilities,
thus providing the best out-of-the-box experience.

To quote [13]:

"Hello all-

First off I'm not a dev, just a small business owner trying to start up.
I don't have a budget for much of anything & wind up doing most
everything myself, which for the last year includes evaluating virtual
infrastructure offerings.

It's quite discouraging to see just how much doesn't "just work" for
someone like me- unfortunately oVirt included (no firmware package to
initialize Broadcom NetXtreme II's on the standalone iso). I usually
move on when "build" is mentioned & have no aspirations to learn git &
thus, testing oVirt's on pause until a new iso's released.

Not to mention how it feels to hit a feature paywall after tinkering a
few days with something.

I'd very much like to try something like what Mr. Fortaine's speaking
of, & I hope to see it realized one day, in a way that wouldn't require
someone like me to hire a consultant to implement.

Just my 2 cents.

-Chris Bartels

  President / CEO
  Nocturnal Feast"

c) Cost

By leveraging Open Source Technology, we could provide a solution at
the fraction of the cost than the current ones, especially VMWare ESX
Server (2250 USD) and Microsoft Server 2008 R2 with Hyper-V (2500 USD)
[14]

Indeed, two Volume x86 servers with a BIOS HV will cost the equivalent
of a licence for 1 VMWare/Microsoft Hypervisor.

Best Regards,

Guillaume FORTAINE

[3] 
http://git.infradead.org/iommu-2.6.git/blob/e0fc7e0b4b5e69616f10a894ab9afff3c64be74e:/drivers/pci/intel-iommu.c
[4] http://patchwork.kernel.org/patch/42841/
[5] 
http://git.infradead.org/iommu-2.6.git/commit/e0fc7e0b4b5e69616f10a894ab9afff3c64be74e
[6] 
http://git.infradead.org/iommu-2.6.git/commit/2ff729f5445cc47d1910386c36e53fc6b1c5e47a
[7] 
http://git.infradead.org/iommu-2.6.git/commit/6ecbf01c7ce4c0f4c3bdfa0e64ac6258328fda6c
[8] 
http://git.infradead.org/iommu-2.6.git/commit/5854d9c8d18359b1fc2f23c0ef2d51dd53281bd6
[9] 
http://git.infradead.org/iommu-2.6.git/commit/86cf898e1d0fca245173980e3897580db38569a8
[10] 
http://git.infradead.org/iommu-2.6.git/commit/0815565adfe3f4c369110c57d8ffe83caefeed68
[11] 
http://www.phoronix.com/scan.php?page=article&item=intel_corei7_virt&num=6
[12] http://wiki.xensource.com/xenwiki/VTdHowTo
[13] https://www.redhat.com/archives/ovirt-devel/2009-December/msg00063.html
[14] 
http://www.virtualization.info/2009/11/red-hat-releases-enterprise.html 
<http://www.virtualization.info/2009/11/red-hat-releases-enterprise.html>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listman.redhat.com/archives/ovirt-devel/attachments/20091229/b29982f0/attachment.htm>