Although on the Windows virtualization team here, it's heads down working on the next generation of virtual machine support to be built into Windows Longhorn server with a "hypervisor" architecture, we're absolutely not sitting still on the Virtual Server 2005 R2 product which became free a couple of weeks back.
There will be a couple of beta releases of Virtual Server 2005 R2 SP1 - today we are announcing the availabilty of beta 1. You can download it by applying through Microsoft Connect. You'll need a passport account to log on to that site. Scroll down to the bottom and click Apply. It could take up to 24 hours to get an email back with the link for download.
So what's in Beta 1? Functionally, the main change is the support for Intels VT processors which provides hardware assistance for virtualization support. What does that mean? If you're running on a VT enhanced processor, NON-Windows guests will run much faster as we're no longer performing ring compression as part of the emulated environment (note you can choose to turn VT support off). However, Windows guests will run at parity in terms of performance as we already performing optimal tuning through the use of VM Additions. Windows guests will get a significant boost during the install process though, the time before you have an opportunity to install the VM Additions. The host clustering white paper and the associated VB Script I blogged about back in November last year, this is now included in the box.
Moving forward to what's going to be included in Beta 2, we'll add support for hardware assistance in the AMD chipset (aka Pacifica), integration with Active Directory to allow you to identify VM Host machines in a consistent manner, plus one killer feature (IMHO) - VSS support. This will allow you to take a snapshot of a running VM for backup - something I've had so much feedback from people to put into the product. More details to follow on that closer to the time.
Timewise, current schedule dates for Beta2 and released product are last quarter 2006 and first quarter of 2007 respectively.
Once again - the registration link is here.
Any questions, let me know.Cheers,John.
I had applied for participation to Virtual Server 2005 R2 SP1 Beta few days ago but I haven't receive any e-mail with link to download.
When I select Virtual Server 2005 R2 SP1 Beta link on connect.microsoft.com. It show the message "Coming Soon".
This sure sounds great...
Is there some way that Microsoft could allow a free virtual machine license for a guest, equivalent to the licensed OS of the host?
In other words, I would like to use Virtual Server on my XP Pro box, but I can't install XP Pro into Virtual Server because I can't afford another license....
Chong - entry into program only went live a couple of hours ago. It appears that the someone was a little ahead of themselves and posted the link onto connect a few days ago. You should receive an email within 24 hours of 10AM (PST) when it officially opened.
Jace - a million $ question. Understand the predicament. Whereas I won't rule this out entirely, I would certainly say that is unlikely in the near future. There is somewhat of a precedent with Windows Server 2003 EE allowing 4 virtual machine use rights with the server license. I'll pass the feedback on certainly.
FWIW - if this is for dev & test, it may be worth signing up for a TechNet Plus or MSDN subscription. Both have rights for up to 10 instances of XP/2003/... There's usually some type of promotional code available.
From a VS blogger:
So what's in Beta 1? Functionally, the main change is the support for Intels VT processors...
The lack of performance improvement for guest Windows VM when Intel hardware VT support is enabled seems counter-intuitive. I'm not a chip or VM guru, but my general impression was that without hardware support there is always some level of ring 0 emulation and interrupt handling that is inefficient and should be done better in hardware.
So is this wrong, or is the lack of performance improvement indicative of Intel's hardware VT support being lackluster or VS support not being optimized enough to use it and make a difference?
On another note, it seems to me one of the biggest limitations in software VM is the lack of memory image sharing. Assuming I'm running "n" copies of the same guest OS (windows Server 2003, for example) with each guest having 512Mb of ram, I would guess that 80% or more of each VM's ram content is identical code.
It's a shame that either through sw wizardry or hardware VT support for memory mapping all that read-only code pages cannot be globally shared.
Sort of a"single instance store" for VM ram much like Microsoft has done for disk files in server provisioning and the new storage server specialized OS build.
Especially with consolidating lots of low-activity VM's on one physical host, the lack of memory sharing is what limits the density.
If not an elegant solution ,then at least a "committed" and "burst" memory allocation of VM usage like Virtuozzo does would allow some higher density but not the true codepage sharing it should.
You've made some excellent point here. Hopefully I can address them all.
With hardware virtualisation assistance, there is no longer any need for ring compression. All guests VMs run their kernel at ring 0, as would be expected if running on real physical hardware. Hardware virtualisation assistance give you what can be thought of as Ring "-1", a ring below the level in which the guest VMs are running in. Hence, there is no need for ring compression or trapping of non-virtualisable instructions in these circumstances.
I understand your comments about static memory sharing. That is a consideration we're putting into Windows Virtualization, but is not today a capability we're introducing into Virtual Server. Yes, I agree, it would save overall host memory usage where there are many static pages in common between guests, however, our general consensus presently is one of security first and foremost in ensuring that there is no leakage between guests. The model of 100% isolation is generally the more secure option. Hardware assistance unfortunately doesn't provide this capability today - once we get into chipsets support providing true isolation in hardware of memory mapping of device memory space (MMIOU/DMA-r etc.,), it's becomes somewhat a no-brainer as a product feature.
I'm familiar with Virtuosso, but certainly wouldn't consider myself an expert. Both Virtual Server and Windows Virtualization are what can be considered hardware virtualization. Virtuosso on the other hand is what most people would consider "software" virtualization. They address completely different problems in radically different ways. Maybe a little simplistic, but hopefully this make sense.
very interesting stuff.
I´m very excited to see how much performance impact Vanderpool has...
Next week I will test it on a new IBM Dual Core Xeon Server.
This slipped past my RSS reader until the weekend (I had a busy Friday).&nbsp; Service Pack 1 for Virtual...
BTW: On mcseboard.de where I´m a moderator a user have posted a interesting question to this topic.
Is there any common solution how to check if a cpu is vanderpool ready and if it´s activated?
How about using the Intel utility for processor identification:
Same answer I give ;-)
but I pointed to the bootable Version...
I will have to recheck this when I´ll get my Vanderpool-Ready Server.
What would be very interesting is, how can I identify if Virtual Server use Vanderpool.
PingBack from http://blog.jensthebrain.de/archives/2006/05/02/beta-virtual-server-2005-r2-sp1/
Hi Philipp - assuming you mean can a guest tell if it is running on a VT enabled host processor, answer is no.