VIRTUALBOY BLOG
One of the new capabilities of Windows Server 2008 R2, and Windows 7, is Core Parking.
As we move closer to the release of Windows Server 2008 R2, and combine this with the innovation in CPU technologies that Intel and AMD are producing, it’s clear that CPU’s are getting more powerful, more capable, more-core-heavy (6 and 8 core chips are either here, or imminent) and more efficient, yet we need to ensure that the OS that is running on these CPU’s can take advantage of these CPU improvements. I’m pleased to say that the R2/7 wave of technologies are right on the money.
As you can see from the diagram, it shows a single quad-core processor, but this could equally be a 6 or 8 core processor. You can see that 3 of the 4 cores on this chip are ‘inactive’, or ‘parked’. This means that those cores aren’t using as much power, and thus, the overall power consumption of that server is decreased. Now, I’m not going to say that this is going to change the world and cut your electricity bills buy 2/3rd’s, but every little helps in a very green-focused climate. Best thing about it is, it’s just there, in the OS, managed by the Kernel.
So, as an example, if I’ve got a number of VMs, running on a host, and those VM’s resource needs, from a CPU perspective, could easily be handled by fewer cores than I have in my box, the OS will ‘move’ processing for those VMs onto the fewer cores, and ‘park’ the spare ones, giving some power back to the server. Again, the amount won’t be huge, but over a period of months, it will add up to be a noticeable difference I’m sure, however at the end of the day, it’s an OS capability, so you’re not buying the power related features.
Don’t worry though – resuming from ‘parked’ isn’t like resuming your laptop from sleep – this is practically instant resuming!
It doesn’t just stop there – Windows Server 2008 R2 also provides reduced processor power consumption by adjusting processor speed:
Windows Server 2008 R2 has the ability to adjust the ACPI “P-states” of processors and subsequently adjust server power consumption. ACPI “P-states” are the processor performance states within the ACPI specification. Depending on the processor architecture, Windows Server 2008 R2 can adjust the “P-states” of individual processors and provide very fine control over power consumption.
There’s a nice example of that here.
How can you see Core Parking? Well, if you open up Resource Monitor, click on the CPU tab, adjust your view appropriately, and viola, check out the parking on that!
This is a 16 core (4 x quad core) box, running 5 (pretty idle) VMs, but still, only 2 of my 16 cores are active at this point in time, and you have to have a minimum of 1, so that isn’t bad! It might not always be the same 2 – it will fluctuate from time to time, but still, it gives you a good idea that something is happening, and that little something will save you money, and improve your green credentials, and, like a hell of a lot of other things, it’s just out of the box with Windows Server 2008 R2, regardless of Hyper-V virtualisation.
Top stuff.
Hot on the heels of my post a couple of days back, about Core Parking, I’d like to add to that, some information around some of the other CPU-related enhancements that are coming in the R2 wave.
The first, is the increase in the number of logical cores supported in the Hyper-V R2 parent OS. In Hyper-V V1, which shipped as part of Windows Server 2008, we supported 16 cores in the physical server, however that quickly grew to 24 cores, with the addition of a hotfix. The plan of action for R2 would be to extend this to 32 cores, however the Hyper-V team have extended well past that mark, and when R2 ships, we’ll actually support 64 logical processors in the physical server, which is a great example of high scalability and puts us on a par with the upcoming vSphere release, from VMware. It’s also important to note, that Windows Server 2008 R2 as on OS, without enabling the Hyper-V role, will actually support up to 256 cores. At the current levels of cores per CPU, that’s still a hell of a lot of CPUs!
So what does this mean in terms of the number of VMs I can run?
Well, how about:
In all honesty, when you get to numbers this high, it starts to get a bit crazy, and it’s unlikely that anyone will run that number of VMs on a single server on any virtual platform in production for a good while yet – think how much memory you’d potentially need! Obviously not much if you wanted to run 384 x 8mb RAM VMs :-) In fact, you’d only need about 100GB RAM to get 384 x 256mb RAM VMs running, so perhaps not as unrealistic as I first thought! Still, I don’t have 64 cores do I!!!
Processor Compatibility
With the inclusion of Live Migration, the matching of hardware, particularly CPUs, within your Hyper-V cluster becomes even more important. You wouldn’t want to Live Migrate a VM from a newer CPU Host, to an older CPU Host, to find that it results in a failure because a feature of the CPU on the newer host, wasn’t present on the older CPU host. Ideally, a check should be performed to ensure that migrations only take place if a level of compatibility is in place between your hosts CPU’s, but that suggests that if the check is performed, and the move is deemed incompatible, that there would be no way to migrate the VM between that particular source and target. That isn’t good enough. Thankfully, it isn’t an issue.
Processor Compatibility mode is a feature that is set on a VM by VM basis, and basically, ensures that if you want to migrate a VM between hosts that have slightly different CPU feature sets, Hyper-V only exposes the common CPU features to the VM, ensuring migration compatibility. It’s not magic, so it won’t provide an Intel –> AMD migration, or vice versa, but it will for example, allow scenarios like:
This is one of the test cases that has been performed in Redmond, but not just a few times. They’ve been migrating VMs, every 15 seconds, between these different hosts on a 1GB iSCSI SAN backend, for a week now, topping over 110,000 Live Migrations between these different CPU architectures.
This capability, like I said, is enabled on a VM per VM basis, and isn’t to be confused by similar, hardware based approaches from AMD (Extended Migration) and Intel (Flex Migration) which are enabled per host, in the BIOS.
You can read more about Processor Capability on the Virtualisation blog, and also hear about it in one of my upcoming demo video’s.
As virtualisation becomes more and more popular, one of the common questions I’m being asked is around USB and getting USB devices into the virtual machines. This isn’t such a problem in the Desktop Virtualisation space – most technologies now allow for USB redirection of some sort, but in the server space, it’s still a little scarce. I blogged about USB over IP a while back, and referenced a technology provided by a company called Fabulatech.
This video, is your chance to see Fabulatech and Hyper-V R2 in action. Together.
You can view the video straight in this webpage, or head on over to the VirtualboyTV to see it in higher definition. For the best experience, you can always download it from this page too.
Feedback welcome, and stay tuned for future videos!
Application Virtualisation (App-V) has been around a fair while now. It started as SoftGrid, and was owned by Softricity, which was then acquired by Microsoft, and became App-V, with the current release being 4.5. Version 4.6, which brings in things like x64 support, is currently in the Alpha stage. The thing I love about App-V, is the power is gives the IT Admin to deploy applications, without going through a huge regression testing period, as each application exists within it’s own little sandbox, isolated from other applications that would normally conflict. This means that you overcome app to app incompatibilities, but you also gain an extra level of control over the OS, as it can be quickly replaced, yet when the user logs in to their new OS, the same applications are streamed down to them, on demand.
That’s the client side covered, but what about the server side? Can we not stream to servers? Not yet, but in the future, it could well be reality. Check out this video if you don’t believe me:
In the video, Bill Morein uses an updated version of System Center Virtual Machine Manager, which allows applications to be stored in the VMM library as well as regular VHDs etc. In the video, he clicks on a running VM, and selects ‘Deploy Application’ and chooses the right app from the library. In this example, he deploys an Enterprise Search application, so we’re not talking Office, or Adobe Reader here.
Bill shows that the app is now running in the VM, and in all honesty, it’s pretty seamless and very cool. There’s more stuff in the video, and it’s definitely worth watching. It won’t be here for a good while, but it’ll cause big ripples in the traditional way you deploy server applications in your infrastructure of the future.
OK, now the fun starts! The previous 2 ‘R2’ videos (Initial Installation & Configuration, and Dynamic Storage) gently introduced you to some of the new capabilities in Hyper-V R2, but now we’re going to step it up a notch, and start looking at the clustering and migration capabilities in R2. Before we start looking at the features around clustering, we need to build it first! Thankfully, this video walks you through that!
The video is just over 16 minutes long, but is definitely worthwhile if you want to understand the validating and building phases of cluster construction.
Feedback welcome, and stay tuned for part 2 of this mini cluster-series!
The 3rd video in this cluster-tastic mini-series focuses on introducing a completely ‘new-for-R2’ technology called Cluster Shared Volumes. This new technology brings 2 major benefits. The first, is it allows you to drop multiple VMs per LUN, and still retain the ability to individually migrate and failover the workloads. Secondly, it provides you with I/O Redirection, which means, should you lose connection to the SAN from a node, or a network from a node, the VMs will be unaffected, as CSV will redirect the I/O via another available path. You can read all about CSV here.
So, the video…
In this video, I not only enable CSV, but go on to quickly create a couple of HA VMs, using CSV as their backend. Sounds simple and straightforward? That’s because it is!
Feedback welcome, and stay tuned for the final part of this mini cluster-series!
SQL Server, both 2005, and 2008, have both been supported in virtual environments, Microsoft or otherwise, for some time now. This means that, providing you followed the specific guidelines for SQL and virtualisation, you could happily virtualise SQL, as a standalone instance, on a virtual platform. This was acceptable for a number of smaller organisations - the fact that you could virtualise a SQL instance, and make it highly available on a virtualisation host cluster, gave them a greater level of resiliency than they had before! Win Win!
However, what about slightly larger, or more complex SQL deployments? In the physical world, no problems, clustering at the SQL level is supported – happy days. What about in virtual machines? Can I cluster SQL 2005, or 2008 at the Guest OS level, i.e. from a Windows VM, to a Windows VM, running on a virtual platform? Well, up until a few days back, you could do it, but you wouldn’t be supported. I’m now happy to say, you are supported when clustering SQL at the Guest OS level.
You can grab all the info you need, over at this SQL KB Article.
I hinted in one of my final posts before holiday, that there would be some big announcements taking place in and around the virtualisation space. Well, firstly, Windows 7 Release Candidate launched, and there is a tonne of information about that, here:
Whilst Windows 7 is great, and my use of the Release Candidate has been fantastic, it’s the virtualisation that I’m interested in, which brings us on to Windows XP Mode and Windows Virtual PC. Let’s face it, whilst Vista was good, Windows 7 is better, and many organisations see Windows 7 as their next logical port of call, when it comes to their desktop OS. This is all well and good, but it’s quite a jump from an OS that is years old, to a brand spanking new one, and whilst most stuff will work fine, what about the niggling applications that you can’t take across to natively run on Windows 7? Virtualise them, with App-V? Yep, you could do, but if the application isn’t designed to run on Windows 7, chances are, virtualising it won’t help. It needs to run in it’s native environment – XP. Bring in Windows XP Mode for Windows 7.
What is it?
Well, taken from the previous link, “XP Mode consists of the Virtual PC-based virtual environment and a fully licensed copy of Windows XP with Service Pack 3 (SP3). It will be made available, for free, to users of Windows 7 Professional, Enterprise, and Ultimate editions via a download from the Microsoft web site. (That is, it will not be included in the box with Windows 7, but is considered an out-of-band update, like Windows Live Essentials.) XPM works much like today's Virtual PC products, but with one important exception: As with the enterprise-based MED-V (Microsoft Enterprise Desktop Virtualization) product, XPM does not require you to run the virtual environment as a separate Windows desktop. Instead, as you install applications inside the virtual XP environment, they are published to the host (Windows 7) OS as well. (With shortcuts placed in the Start Menu.) That way, users can run Windows XP-based applications (like IE 6) alongside Windows 7 applications under a single desktop.”
As you can see from this image below, courtesy of WinSuperSite:
Seamless to the Windows 7 desktop, is a localised application, Word 2007, with all it’s Aero Glass loveliness, yet also, an XP-based Word 2003 instance running side by side. This Word 2003 instance is actually running in the XP Virtual Machine, running on Windows Virtual PC (Beta), in the background to the main OS. The apps that are installed within the XP VM, can also be exposed into the Start Menu of the Windows 7 OS, so again, it’s all seamless to the end user, which is the aim of the game.
Rafael, over at WithinWindows, has a detailed overview, from a technical perspective, of XP Mode, split into 2 parts:
James has also been busy and posted some good stuff about XP Mode.
So, this XP Mode relies on Windows Virtual PC, which is currently in Beta. What’s new in Windows Virtual PC over the old Virtual PC 2007? This table summarises it pretty well:
I think you’ll agree, there’s some nice features in there, but there is a reliance on having the relevant AMD or Intel virtualisation technologies on the chips.
How does this ‘invisible’ XP VM running on Windows 7 compare with Microsoft Enterprise Desktop Virtualisation (MED-V)?
This is important. MED-V, for those of you not familiar, is very much like XP Mode, in the sense that the end user just sees the XP apps presented to them into their Vista Start Menu, yet that ‘invisible’ XP VM has been distributed centrally, in a managed fashion. The end result to the user is very similar, but there are a number of differences to be aware of, which James, and Scott have both done great jobs in weighing in with their views.
In a nutshell, if you’re a smaller business and you need app compatibility for Windows 7, you’ll use the free XP Mode. If you’re a midsize or larger organisation and you want to centrally manage your XP VMs that are running on your users’ devices, you’d use MED-V. Both great technologies, and both meet the needs of certain scenarios.
That’s enough about the client stuff – give me the news on the server side!
Well, first up, we’ve released the Release Candidate of Windows Server 2008 R2, which with it, brings the next release of Hyper-V. No ‘within 180 days’ launch stuff this time – Hyper-V R2 will launch with Windows Server 2008 R2. No waiting around :-)
If you don’t know what’s in Hyper-V R2, and would like to know more, I’ve provided a detailed overview video over on the Virtualboy TV site.
With the release of Server 2008 R2, also comes the relevant Windows 7 client tools (the Remote Server Admin Tools, RSAT) to download, so you can manage your R2 servers from Windows 7 client. You can grab those bits here.
As you can imagine, the Price Comparisons are flying about all over the place, but that’s not a discussion I want to get into on my blog. There’s some great technologies out there now, each with varying feature sets, and after all, competition drives innovation.
The final bit of news is that Hyper-V Server 2008 R2, which I discussed in detail here, has also been released to the web, in it’s Release Candidate form. You can grab that here. The guys at the virtualisation blog have also weighed in with their views, and the obligatory cost comparison.
Hopefully that’s given you enough to be going on with – there’s a lot to take in there, but I’d urge you to start testing R2 and Windows 7. I’ve just signed up for our internal testing of DirectAccess, which is a feature which removes the need for VPN access to access your internal systems and file shares etc. I no longer need VPN, wherever I am in the world! It’s great! That’s just one example of where Windows 7 combined with R2 is making my life easier. Go on, give them a try.
For ease, I’ve grouped the best download links below:
If you get some spare time today, or over the next few days, and you want to see some pretty cool demos, then I’d encourage you to take a look at the TechEd Keynote, which was originally delivered on Monday of this week. In the Keynote, Bill Veghte, Senior Vice President of the Windows Business, and a number of others, including Mark Russinovich and Iain McDonald, take you through a number of key development areas in Windows Server 2008 R2, Windows 7 Client, Exchange 2010, Office 2010 and more. There’s even a bit of SCVMM 2008 R2 thrown in for good measure. If you want to understand just a few of the key features and capabilities of the next wave of Microsoft technology, this could be a great starting point for you. I watched it this morning and I was suitably impressed.
You can view the keynote here, but it’s a bit small, so you may want to use the download link on that page to get a slightly bigger version.
On the back of the TechEd kickoff, and actually within the keynote, Bill announced that Windows 7 and Windows Server 2008 R2 will ship before the holiday season. Now, when I think Holiday season, I think Jingle Bells, which would be December, however The Register, seems to think it will be sooner than that, citing an August-ish release. That would be nice!
In terms of specific new features and capabilities that were announced, the list includes up to 64 logical processor (cores) support for Windows Server 2008 R2 Hyper-V, which is a massive jump from both the 24-core limit in 2008 (with the hotfix), and the original proposal of 32 cores for R2. This post, over at the Windows Server blog, not only highlights the increase in the number of cores’ support in R2 Hyper-V, but also highlights the improvements that have been made to Hyper-V to ensure Live Migrations are compatible between different CPUs of the same family. This capability wasn’t in the beta, so the development team have been working hard to make this kind of feature a reality, in a short space of time. This allows scenarios such as adding nodes to a cluster, that have slightly newer (same family, Intel or AMD) CPU’s than the other nodes in the cluster. Can I not migrate my VMs to those nodes? Yes, with this capability, you can, and it’s built in, out of the box. AMD are the first to provide a real world example video of this in action, so you can read the details, or watch the video.
Again, on the back of TechEd, the SCVMM team took the opportunity to announce the key new features that will make it into the Release Candidate of SCVMM 2008 R2, that weren't in the Beta. Before you ask, RC should be late May. What are the extra features?
Storage Migration is an interesting one. This basically allows you to move a VHD of a running VM, from one place to another. I guess, as Vishwa points on out the SCVMM blog, it will aid the migration from 1 VM per LUN scenarios on 2008 Hyper-V, to multiple VMs per LUN with CSV in 2008 R2. The queuing of Live Migrations is a useful little tool. Without SCVMM, if you try and migrate a VM from Node A to Node B, whilst a VM is already being migrated from A to B, the migration will fail, as only 1 Live Migration can take place at a time. Not the end of the world, but a bit more manual than I’d like. This doesn’t stop you performing multiple migrations between different sources and targets. With VMM R2 however, it will form an orderly queue of Live Migrations and perform them 1 by 1.
Rapid provisioning is a nifty little feature, that really allows you to start taking advantage of the investments you’ve made in your storage backend. Deploying a VM from SCVMM’s library takes time, as typically, it’s transferred via BITS, but what if you’ve already got a LUN replication capability within your SAN, which would allow you to quickly duplicate the contents of LUN A, to make a new LUN, B. LUN B would subsequently contain the VHD file (much quicker than transferring with BITS from the library), so all you need from VMM is some sort of config file template type thing, to make it into a VM, rather than just a Virtual Hard Disk. More details on this, and all the other improvements listed above, over on the SCVMM blog.
Bring on the end of May / Early June, when I can add SCVMM 2008 R2 to my Hyper-V R2 environment! Happy days!
It’s been a long time coming, (and it’s still not quite here), but the details have finally been confirmed. The Microsoft Partner Virtualisation Competency assures that you, as a Microsoft Partner, have the skills to build your virtualisation practice based on solutions like Windows Server 2008 Hyper-V technology, Microsoft System Center Virtual Machine Manager 2008, App-V, MED-V, or Microsoft VDI.
The exams that make up the competency have been around for a while, such as the Hyper-V Exam, the SCVMM 208 Exam etc, however, when it all comes together to form an official competency, Partners will be instantly recognisable as having these skills.
After you’ve gained the competency, Partners will also have access to the following:
You can grab all the details, including Exam information, Reference Requirements, Recommended Resources and more, over at the Microsoft Partner Portal Page.
So, best advice is, get ready. Start working with Hyper-V and SCVMM, App-V and TS, and start to get certified, so when the competency launches, you can be one of the first, official, Microsoft Virtualisation Partners.