Introduction to Hyper-V

Editor's Notes

• #2: This slide deck was written for the Microsoft UK User Groups Community Day on 9 April 2008, where various user groups presented.The full content is included in this deck, including extensive speaker notes; however many of the slides include builds/animations and so it is recommended to watch the presentation in slide show view.
• #3: Hi there!For those who don’t know me, my name is Mark Wilson and I’m an infrastructure architect who has spent the best part of 15 years designing and implementing IT infrastructure solutions within the United Kingdom, continental Europe and Australia – mostly with Microsoft technologies.I often describe myself as being “operating system agnostic” as I hold certifications from Microsoft, VMware and Red Hat – and I’m also a Mac user at home. Despite this, much of my professional life revolves around the Microsoft Windows Server System and I amthe technology lead for Windows Server 2008 and Hyper-V technologies at Fujitsu Services in the UK - http://uk.fujitsu.com/ .I also spend far too much time in my home office playing with technology and blogging about it – please come and visit me at http://www.markwilson.co.uk/blog/ .
• #4: The Windows Server Team is a UK user group based around Windows Server – and with Windows Server 2008 being the latest release, that’s what we’re excited about right now.We’re quite a “young” user group – having only been formed in 2007 and have run a few events to date, with more to come. Unfortunately, we’re in a state of limbo at the moment as our leader (Scotty McLeod, who many of you will know) was involved in a horrendous accident a few weeks back.Even so, we’re pleased to be taking part in this event and hope to work on building a community of people who are interested in Windows Server technologies – come and see us at our website.
• #5: In this session, I’m here to talk about Hyper-V, formerly known as Windows Server Virtualization and codenamed Viridian, which is Microsoft’s virtualisation platform for the enterprise.People here from Microsoft may disagree, but in my opinion (and in the opinion of many of my colleagues and customers) Microsoft has not had a credible enterprise virtualisation product to date. Virtual Server 2005 was good, but the management story was incomplete and VMware, who frankly invented the x86 virtualisation market, were running away with the market.But Microsoft’s story is getting better – with new management tools and a new virtualisation platform – my feeling is that Hyper-V will pick up a significant chunk of the virtualisation market over the next couple of years.And before everyone thinks that I’m here today as a Microsoft mouthpiece, it’s worth noting that I am a VMware Certified Professional too!(Ask the audience who has experience of server virtualisation? Using Microsoft products? Using VMware products? Using something else? In test and development? For production workloads?)
• #6: In this presentation, I’m going to set the scene by taking a look at what virtualisation is about – in terms of the technologies and also the opportunity. I’m English, so I spell virtualisation with an s, but some of the direct quotes in this presentation may use a z!I’ll then spend some time looking at the architecture that Microsoft has used for Hyper-V, before I demonstrate installation of the product. After that, I’d like to spend some time talking about management of Hyper-V.We should have an hour but there is a lot of material to get through. This is also the first time I’ve presented at Microsoft, so if you could go easy on me and hold off with questions until the end I’d really appreciate it (unless I’m saying something really confusing!)(click) Just before we move on, I’ll mention that Microsoft recently made a release candidate available for Hyper-V. I’m running that version here today, but the videos I’m using were recorded with the beta product and some of the information may change before the product is released.
• #7: The demonstration environment that I’m using today consists of a single notebook PC (UK090922LT). This is a reasonably highly specified machine, with a dual-core CPU and 4GB RAM, it’s also joined to my organisation’s Active Directory.(click) The notebook is running a full x64 installation of Windows Server 2008, with the Hyper-V role (RC0). For anybody who thinks that it looks like Vista, it does – I’ve turned on most of the Vista functionality that’s not there by default with Windows Server 2008 and because Hyper-V can run on Windows Server 2008, it means I also have a really good virtualisation platform on my desktop!(click) There are also some virtual machines available - a server core installation (UK090922VM3) that I’ll be using for the server core session this afternoon and that has fully Hyper-V support, (click) and a 32-bit XP installation (UK090922VM2) that I use for VPN access to my office (our corporate VPN client is not supported on 64-bit Windows) and I’ll also be creating another VM as part of the demonstration.
• #8: So, let’s kick off by looking at what virtualisation is – and why it’s such an opportunity.
• #9: I thought, seeing as I’m at Microsoft today, I’d start off by defining virtualisation as the competition see it.Virtualisation lets one computer do the job of multiple computers, transforming hardware into software to create a fully functional virtual machine that can run its own operating system and applications so that you can run several operating systems and applications at the same time on a single computer.(Does anybody disagree with that?)
• #10: The number one reason that people talk about virtualisation is server consolidation. Server consolidation is certainly the driving force behind the technology’s adoption, with the potential to reduce costs, improve management, and reduce the power consumption and data centre footprint of our server estates.For many organisations, it’s pretty easy to justify the purchase of some new servers but there is nowhere to put them. At a national level, we are running out of power!!!By cutting the number of servers, it’s very easy to show cost reductions (and there are obvious environmental benefits too) - 2:1 consolidation is easy! But we should be able to get much higher than that, and (click) using intelligent management to monitor the workload and consolidate further, we can improve utilisation of the server infrastructure, which is generally reckoned to be (on average) around 5-15% CPU utilisation (pretty much undisputed)There will always be some servers (e.g. SQL, Exchange) which are not good virtualisation candidates, but these are the exception rather than the norm.(click) Business continuity is an important factor – virtualisation is no panacea and one downside often stated is the prospect of a single point of failure. This is why virtualisation took off in test and development environments but is only now making it into production. With real-world consolidation values of 16:1, when a physical server falls over, that’s a big problem – not just one workload out of action but sixteen! (click) On the flip side, virtualisation offers the flexibility to move virtual machines around with ease, allowing rapid deployment of servers in minutes (not weeks) – although even that needs to be carefully managed.
• #11: If virtualisation is so great, is everybody using it?Not yet! In fact, we’ve only just scraped the surface and it’s generally recognised that virtualisation adoption is at around 5% for server virtualisation and 1% on the desktop.
• #12: Even by 2010, this is only expected to be 17% - although that’s still a massive number of servers moved from physical to virtual hardware.
• #13: So, what’s the barrier to adoption?Firstly, cost - until recently, enterprise virtualisation was expensive. I worked on a VMware virtual infrastructure implementation with one of the big 4 accounting firms and the numbers were potentially huge. The savings were big too, but it required a big investment, and a huge leap of faith.Complexity is a real concern – enterprise virtualisation doesn’t live in it’s own domain – it is intertwined with physical infrastructure, SANs, networking – and there are management considerations too. In the project I mentioned the biggest issue was not technical, but one of organisation – where does the network team end and the server team start. It used to be simple when all the network kit had a Cisco badge on the front, but what happens when some of the switches are virtual?Then there are the management tools - customers want to use the same tools for physical and virtual infrastructures – and those tools need to look inside the VM and check that the workload is healthy, not just the VM container.Virtual sprawl is a real issue as simplified deployment means that people really do deploy lots of VMs... how do you managethat? If you’re not careful, then you could end up with even more servers than you started with as the latent demand for new servers is unleashed.Then there is the single point of failure that I mentioned previously–safeguards need to be built in.Until recently, software licensing was an issue. There have been a lot of changes made as physical licensing didn't make sense in the virtual world. Microsoft’s licensing is per CPU socket, not per core but many ISVs have stuck their heads in the sand re: licensing and support – that’s changing as the argument that “we don’t support virtualisation” no longer holds up as virtualisation is shipping in the hardware of every new server now - and some OEMs are shipping hypervisors with new servers too.Hardware was an issue until recently, now we have Intel-VT and AMD-V but to quote Jeff Woolsey, a Senior Program Manager in the Windows virtualisation team at Microsoft "we've had to go through a number of software gymnastics just to make it work" [Jeff Woolsey, Program Manager].Techniques such as ringcompression and binary translation to jack up a guest operating system in a virtual machine so that it thought it was operating in ring 0 but was actually in ring 1. We don't have to do that with hardware assisted virtualisation.Performance has been another area of concern – virtualisation is fine for some workloads but its big Achilles’ heel is IO - going to disk or getting a packet off the wire – Hyper-V (and Xen) address that to offer near-native performance.I mentioned politics earlier – with SQL admins, Exchange admins, operating system admins, network admins, storage admins, etc. - who owns what when it’s all consolidated?And security goes hand in hand with the politics. We need delegated access.
• #14: Microsoft’s view is that virtualisation is more than just a hypervisor and a suite of management tools – there are a number of different solutions:(click) What we’ve discussed so far (and will concentrate on in the rest of the presentation is server virtualisation, currently provided with Virtual Server 2005 R2, soon to be replaced by Windows Server 2008 Hyper-V. Microsoft see four core scenarios for this technology:Server consolidationBusiness continuityDynamic data centreTest and development(click) Desktop virtualisation (as in virtual desktop infrastructure – for example the Vista Enterprise Centralised Desktop) is a fairly niche market but I think there is a growing opportunity for virtualisation on the desktop, where Microsoft’s product offering is Virtual PC but a number of acquisitions have also been made recently which could lead to some exciting developments.(click) Application virtualisation (formerly known as SoftGrid) is very exciting - instead of virtualising hardware, it offers a virtualapplication interface to the operating system, COM, DCOM, fonts, registry – a real world solution for deploying legacy applications (and Microsoft’s most popular volume license product ever).(click) Presentation virtualisation overlaps with application virtualisation in that it allows the rapid deployment of desktops. It will take you 90% of the way there (especially with some of the enhancements in Windows Server 2008 that Austin Osuide will be talking about today) but there is always one important legacy application that doesn't work in Terminal Services – using application virtualisation takes the solution the rest of the way.(click) Finally, System Center provides a single suite of tools to manage it all.
• #15: So, with the scene set, finally we can start to talk about Hyper-V.
• #16: What is it?It’s a hypervisor-based virtualisation platform but hypervisor is really just a buzzword right now – think of it as virtualisation on bare metal.Hyper-V is just a role on Windows Server 2008 (x64 editions only), available in standard edition, but it makes sense to use enterprise or datacenter edition for licensing purposes. And for those who don’t want it (why?), there are SKUs for Windows Server 2008 without Hyper-V, which I think will sell about as well as Windows “N” Edition (“N” for nobody wants it).Hyper-V is standards based – the management APIs are interoperable (they are desktop management task force standards).Finally, looking at the hardware, it does require an x64 processor with hardware assisted virtualisation (AMD-V or Intel-VT). Hardware enabled data execution prevention is also a requirement but to be honest, NX/XD and 64-bit have been in mainstream computing platforms for a while now and it’s difficult to buy a computer without virtualisation support too.
• #17: There are two main types of virtualisation today – hosted and hypervisor (which may or may not be paravirtualised). In a hosted environment, the virtual machine manager runs partly on top of and partly within the operating system. Examples of this model are VMware Workstation, Server (formerly GSX Server) and Fusion, Parallels Desktop, the Linux KVM, Microsoft Virtual PC and Virtual Server.Hypervisor-based virtualisation products have their roots in 1960s IBM mainframes but the most popular product today is VMware ESX (previously known as ESX Server) and some OEMs have begun to offer ESXi (which was previously known as ESX Server 3i) “embedded” in servers (actually, it’s not embedded at all – it runs on a flash drive on an internal USB port!).Xen is paravirtualised because it requires kernel modifications to support virtualisation; however Windows Server 2008 uses the same kernel with or without the Hyper-V role installed.
• #18: So, let’s look at the requirements in order to support virtualisation.Firstly, with multiple virtual machines all attempting to accessing resources, we need a method of scheduling resource access.Memory Management is required to assign memory to virtual machines and ensure that it is secure and isolated (Microsoft does not allow the over-commitment of resources or transparent page mapping in its virtualisation products)A VM state machine is required in order to record the state of running virtual machines (network, storage, configuration, how it is running at any point in time).Storage and network stacks are pretty self evident but they also need to multiplex IO from physical to virtual machines so that multiple virtual machines can share the same physical resource.We need a virtualised device model so that the virtual machine can talk to physical machines - abstracted through emulation in Virtual Server but Hyper-V takes a different approach (more on that in a moment).Binary translation is only required for legacy hardware without hardware support for virtualisation. Such old servers will have limited storage, bus bandwidth etc. so are probably not well-suited to virtualisation and are not supported by Hyper-V (you could still use Virtual Server) Drivers are required to access physical resources.Finally, a management API is required in order to create and monitor virtual machines.
• #19: Looking at the architecture for Hyper-VWe have the physical server hardware, (click) on top of which the hypervisor is installed. (click) This all runs in ring -1 (provided by the Intel-VT or AMD-V processor support).(click) The parent partition is the first installed operating instance in a Hyper-V installation. The vital components that we discussed in the previous slide are all there - scheduler, memory management, VM state machine, storage and network stacks, etc.This is one area where Hyper-V is very different to ESX – Microsoft’s hypervisor is just over 100k in size (the entire Hyper-V role is around 100mb in size) and includes no non-essential code (http://blogs.msdn.com/virtual_pc_guy/archive/2008/02/25/hyper-v-terminology.aspx) whereas VMware ESX includes drivers for supported devices in the hypervisor. Which would you rather have, a smallhypervisor with a tiny surface attack area or new drivers added to the hypervisor every time there is a hardware change? This approach avoids the requirement for a new driver model required too, which would have been a showstopper – think of the fuss that people have made around deriver support for Windows Vista (which it has to be said is the fault of the IHVs and not Microsoft) – drivers for Windows Server 2008 are the same regardless of the presence or otherwise of Hyper-Vand are run in the parent partition.The Hyper-V and Xen architectures are very similar, but there is an important difference – Windows uses the same kernelwhether it is running on hardware or software - ISVs don't want to test their software in in two environments.Protection rings are used to separate kernel mode (ring 0) processes from user mode (ring 3) and provide a security boundaries (http://en.wikipedia.org/wiki/Ring_(computer_security)).The VM worker processes (VM state machines) run in user mode, ensuring that the virtual machines are sandboxed far away from the hypervisor and run as a network service so have almost no privileges. The WMI provider is the management API – again, safely tucked away in user mode.(click) Second and subsequent operating system installations on Hyper-V become child partitions and, although this is not a hosted environment, they are reliant on the parent partition. If the parent partition fails, then the whole machine has failed. This is why Windows Server 2008 server core is such a great choice for Hyper-V – the reduced codebase improves security and reliability, as well as requiring less servicing (i.e. fewer patches).(click) Looking first at the way virtualised devices have been implemented in previous products, emulation has been used (and is still available as a legacy solution). Emulation is highly inefficient however – in order to retrieve a packet from the network, the application makes a call down to kernel mode and, in order to access the network, an emulated DEC NIC lives in the parent partition’s user mode (sandboxed for security). This provides great compatibility but very poor performance due to the many, many context switches before the packet gets off the wire.(click)Hyper-V uses a new, VSP/VSC/VMBus model, so that (for example), a synthetic device is provided – the Microsoft virtual network adapter. This is known as a VSC (virtual service client) and such synthetic devices have a fast path from the application to hardware, using the VMBus (for interpartition communication) to the network VSP (virtual service provider), which communicates with the NIC driver and providesaccess to hardware without bouncing back and forth between user and kernel mode. The lack context switching results in a huge improvement in performance and efficiency.Microsoft is providing the VSPs and VSCs in the form of integration components (like the VM Additions in virtual server) - OEMs provide device drivers as for non-virtualised Windows.(click) So what about Linux? Is it relegated to a second class citizen with emulated device support? No! Microsoft has partnerships with Citrix (XenSource) and Novell partnerships so that Linux support for the VMBus is available for Hyper-V (there is no Red Hat support, at least not officially - I'm told it works but I've not managed to get anything other than emulation to work in my testing).
• #21: Microsoft cannot afford for security issues to be uncovered in Hyper-VTo that end, it has been designed with some basic security assumptions in place:All guests are untrusted – would you like a web server in a VM to be compromised and bring down the entire system?The parent partition must be trusted by the hypervisor and by the children - effectively anadministrator on the server can manage the entire system.Code in guests can run across all parts of the machine. No more ring compression or binary translation – hardware assisted virtualisation is used instead.The hypercall interface (the interface from the kernel to the hypervisor) is an open specification, available royalty free, and thereforeattempts will be made to compromised its security (not security throughobscufation).It is possible to detect that a system is running on a hypervisor – the version number is even supplied (in any case, this is possible to do today with Virtual Server – the motherboard manufacturer is Microsoft)!As for the hypercall interface, the internal design of the hypervisor will be well understood.
• #22: Because of the basic security assumptions, isolation of virtual machines from one another is critical. To that end, Microsoft has ensured that there is no sharing of virtualised devices - each VM has own devices; a separate VMBus is provided from each VM to the parent partition; and there is no sharing of memory (each VM has its own address space).Indeed the only permitted communication between VMs is through traditional networking and guests cannot perform DMA attacks as they are never assigned to physical devices.Guests cannot write to the hypervisor, and even the parent cannot write to the hypervisor to inject code.
• #25: Installation of Hyper-V is simple – it’s just one more role on Windows Server 2008.When Hyper-V is installed, the hypervisor slides in between the operating system and the hardware - until that is done, Windows is running on bare metal as usual.In order to save time here, I’ve pre-recorded a couple of videos showing the process for a full installation and for a server core installation.
• #26: (click the green button)1:53
• #27: (click the green button)1:32
• #28: Of course, installation is just the start – management is the most important aspect of any virtualised infrastructure.In a few moments, we’ll look at managing Hyper-V with a variety of methods, but first I want to take some time to look at the broader management issues for a virtualised environment.I recently published an opinion article (I have copies here if anyone is interested – hold up a copy) which looks at the reality of managing a virtual environment in which I put forward the opinion that virtualisation is neither a quick fix for complex environments nor a guaranteed source of bottom-line benefits.It all depends on how you manage the technology – and how you adapt the processes and culture of your organisation to new ways of working.
• #29: If you start off with a physical mess and virtualise it, you will end up with a virtual mess – that much I promise you.It’s not just about zipping up the IT so that it takes less room (and less thought) – if anything it needs more attention to detail – and it’s not just about technology: it’s not just a cliché - this really is about people and process too.In order to tame the virtual world:Make an inventory of what servers do and how they relate to one another – a structured model which also needs to consider SLAs so that you know which servers have the greatest business value.Build a provisioning model, with supporting workflow to allow costs to be attributed to a particular individual or department. The cost can be real or indicative, but also allows for different levels of service to be offered (e.g. paying less for a VM that is rarely used).A lifecycle mechanism is required to ensure that disks are not filled with duplicate or unused virtual machines – are they all still needed?Carry out capacity planning exercises in order to find workloads that complement one another, for example applications with peak loads at different times.Record metadata for each VM, including details of its owner, dependencies, and SLA.
• #30: Having talked about the practicalities of managing pretty much any vendor’s virtualisation technology, let’s look at the specifics of managing Hyper-V – starting out with the graphical tools that ship with the product.
• #31: Let’s take a look at the out-of-the-box administration tools for Hyper-V. (click) Here you can see Server Manager being used to examine the status of the Hyper-V role on a server. I can see all the events relating to the service (ignore the errors, these are peculiar to my configuration) and, perhaps more importantly, that all the services are running. (click) If I scroll down, I can see links to useful support and information resources.(click) Hyper-V Manager is installed on Windows Server 2008 when the Hyper-V role is added; however, there is also a version available for Windows Vista SP1.(click) If I connect to my server, I can see that there are two virtual machines running. The first is my 32-bit Windows XP installation (UK090922VM2). There’s no heartbeat shown because this machine does not have the integration components installed (XP SP3 is required for those).(click) If we look at the other virtual machine (UK090922VM3), we can see a number of snapshots. This is the VM that I will be using for my server core session this afternoon, and the snapshots are useful for if I need to return to a particular point in the configuration. We can also see that there is a heartbeat present as the Windows Server 2008 Hyper-V integration components are installed in this VM. All VMs show thumbnails and context-sensitive actions but if I double-click on a VM (click) then a Virtual Machine Connection is opened to work directly at the VM’s console. Note that the VMC no longer uses the VMRC protocol (which Microsoft said wasn’t VNC, but used the same ports and acted a lot like it!) but a version of the remote desktop protocol (RDP) instead.(click) Returning to Hyper-V Manager, I can click New, Virtual Machine, and (surprise, surprise) start the (click) New Virtual Machine Wizard. This takes me through a number of simple steps to (click) specify the machine name and storage location, (click) select the amount of RAM to use (up to 64GB, or the maximum installed in the physical host, minus some overheads – in this case around 4GB), (click) select a network to connect to (more on that in a moment), (click) connect or create a virtual hard disk, (click) and install an operating system from a number of sources, (click) before providing a summary and the option to start the VM immediately.(click) We won’t start the VM right away but can click on Settings in Hyper-V Manager to look at some of the VM configuration options. (click) Starting at the top, we can add hardware – SCSI controllers, network adapters (up to 12) or legacy virtual NICs (more on that in a moment). There is no longer a noticable performance difference between virtual SCSI and virtual IDE controllers but the SCSI controllers can support up to 256 devices each (and there can be multiple controllers) – that’s a lot of devices.(click) BIOS options are now exposed via the management interface, with the ability to set the bootup device priority and also the Num Lock key status at startup.(click) Next, we can change the amount of RAM assigned to a particular VM.(click) This machine has one virtual CPU but could have up to 4 (or the maximum number of cores in the physical machine – in this case 2).(click) We then see the IDE controllers – two of them (click), each with the ability to control 2 devices.(click) In the network adapter settings we can define the MAC address and also any VLAN settings.(click) Serial ports are available (click) but there is no parallel port support in Hyper-V.(click) Virtual floppy drives are also supported (there is no physical floppy drive access).So that’s the hardware – is anything missing that anyone would like to see?For me, the main thing that’s missing is USB support. Microsoft say that there is insufficient customer demand for USB support on a server and see it as a virtualisation client feature – if you, like me, think that’s nonsense, I suggest you provide some feed back to them. There is a third-party USB over Ethernet solution, but it’s not free, and I haven’t tried it.(click) From a management perspective, we have the ability to edit the virtual machine name and provide some notes, (click) we can control the integration services offered by the VSC/VSP model (when supported integration components are installed), (click) the snapshot location, (click) and automatic start (click) and stop actions.(click) I mentioned that there are two types of network adapter. In this XP VM with no integration components we can see that we have the same (legacy) emulated driver that Virtual Server used – shown as an Intel 21140-based PCI Fast Ethernet Adapter (Generic). (click) If I switch to my Windows Server 2008 server core VM with the guest integration components installed, there is a VMBus NIC – show as a Microsoft VMBus Network Adapter.(click) Hyper-V Manager includes a virtual network manager for managing its network switching capabilities. Unlike Virtual Server, where the switch was really a hub – this really is network switching and we have the ability to create a number of types of switch – external (bound to a physical NIC); internal (for child virtual machines and the parent, but nothing on the physical network) and private (child virtual machines only).(click) Looking at how this appears in the parent partition, we can see that I have a number of adapters and my physical NIC is (click) only bound to the Microsoft Virtual Network Switch Protocol but the logical network adapter created by Hyper-V (click) is has all the usual bindings. If I look at the network configuration (click), I can see that the device is described as a Microsoft Virtual Network Switch Adapter.Finally, if you want to bind Hyper-V to a wireless NIC, you can’t, at least not directly. There is a workaround, in which you enable Internet connection sharing on the Wireless NIC and connect it to a virtual switch configured for internal networking. The result is a NATted connection via using the 192.168.0.0/24 network with the ICS on the wireless NIC acting as a local DHCP server. I’ve found that this approach can be used on other NICs too – for example on a 3G mobile data card. If you’d like to know more about how this works, you can read more about this on my blog (http://www.markwilson.co.uk/blog/2008/03/hyper-v-and-networking.htm)
• #32: A while back, I had a look at managing Virtual Server from PowerShell using the Virtual Server COM API and it was a pain in the neck because I had to load a special assembly to create a wrapper for PowerShell to have sufficient privileges to communicate with Virtual Server’s COM interfaces.It’s a whole lot easier for Hyper-V, as the beta of the Hyper-V WMI APIs was recently released (http://blogs.technet.com/virtualworld/archive/2008/01/31/hyper-v-wmi-api-s-publicly-available.aspx). Hyper-V uses WMI APIs (similar to the Virtual Server COM API) to create, manage, monitor, configure virtual resources. Microsoft expects the Hyper-V WMI APIs to be used widely: ·By third party management vendors who want to write tools to manage Hyper-V.·By enterprises who want to integrate with an existing management solution.·By developers who want to automate virtualisation in test and development environments through scripts. Details of the APIs are available on the MSDN website (http://msdn2.microsoft.com/en-us/library/cc136822(VS.85).aspx)I say why limit it to test and development? – managing Hyper-V from PowerShell is a great idea for all administrators, although with VMware developingPowerShellcmdlets for Virtual Infrastructure 3 (currently in beta - http://blogs.vmware.com/vipowershell/2008/03/we-are-beta.html) and SCVMM being built on PowerShell, there is a lot of work going on in this area – so much that I think we could do with some standardisation.Getting back to the point, let’s have a look at PowerShell interacting with Hyper-V via WMI. Readers of James O’Neill’s blog may recognise this demo (http://blogs.technet.com/jamesone/archive/2008/02/01/getting-to-grips-with-hyper-v-s-api.aspx) and James was kind enough to let me use it (when I though to ask him – sorry James), even helping me out when I was having some difficulties with the code (although they turned out to be a permissions issue on my machine – running an elevated PowerShell fixed that).
• #33: PowerShell rocks. But this isn’t a PowerShell demo! I’m just using PowerShell to show how straightforward it is to control Hyper-V using WMI.(click) We start off with a PowerShell (run elevated) and type:get-wmiobject -class "MSVM_ComputerSystem" -namespace "root\\virtualization" This will return details of all the objects from the MSVM_ComputerSystem class (on the local machine) – including Microsoft Hosting Computer System (Hyper-V parent partition) and (click) Microsoft Virtual Machine (this has changed between beta and release candidate – it used to be Microsoft Virtual Computer Systems (http://blogs.msdn.com/virtual_pc_guy/archive/2008/03/27/wmi-changes-in-hyper-v-rc.aspx).(click) Next, instead of returning all objects in the class, we’ll perform a WMI query to search for objects containing virtual in the caption (i.e. just the VMs) (click):get-wmiobject -namespace "root\\virtualization" -query "SELECT * FROM MSVM_ComputerSystem where Caption like 'Virtual%' "(click) That works, so let’s put it into a variable instead:$vms = (get-wmiobject -namespace "root\\virtualization" -query "SELECT * FROM MSVM_ComputerSystem where Caption like 'Virtual%' " )And if I (click) just type the variable name:$vmsI get the same output (click)(click) It’s a bit unweildy, though, so let’s format it as a table and give it some meaningful headings:$vms | Format-Table -autosize @{Label="VM Name"; expression={$_.ElementName}} , @{Label="VM GUID"; expression={$_.Name}} , Enabledstate(click) We can even change the running state from a number to a known state (it that looks a bit long, remember that this is a scripting language, not necessarily something to enter at the command line!):$vms | Format-Table -autosize @{Label="VM Name"; expression={$_.ElementName}} ,@{Label="VM GUID"; expression={$_.Name}} , @{Label="State"; expression={switch ($_.EnabledState) { 2 {"Running"} 3 {"Stopped"}   32768  {"Paused"} 32769 {"Suspended"} 32770 {"Starting"}   32771 {"Snapshotting"} 32773 {"Saving"} 32774 {"Stopping"} } }}(click) I’m just going to paste in a function here that James wrote to retrieve the list of VMs and give a choice for which one to act on:Function Choose-VM{$global:Counter=-1  $vms = get-wmiobject -namespace "root\\virtualization" -query "SELECT * FROM Msvm_ComputerSystem WHERE Caption Like '%virtual%' " format-table -InputObject $vms -Property  @{ Label = "ID"; Expression={($global:counter++) }} ,                                          @{Label="VM Name"; expression={$_.ElementName}} ,                                           @{Label="VM GUID"; expression={$_.Name}} ,                                        @{Label="State"; expression={switch ($_.EnabledState) { 2{"Running" }      3 {"Stopped"}  32768{"Paused"}                                                                                             32769{"Suspended"} 32770 {"Starting"} 32771 {"Snapshotting"}                                                                                            32773{"Saving"}    32774 {"Stopping" } } } } | out-host $vms[ [int[]](Read-Host "Which ones? ").Split(",")]}(click) As you can see, the VM I created earlier (UK090922VM1) is turned off (click) but if I run this command:(choose-vm) | foreach-object {$_.requestStateChange(2)}(click) And select the VM, (click)(click) We can see that the VM is now starting up, (click) the console reconnects, (click) and we get a boot error because that VM does not have an operating system installed.(click) If I do the same again, but this time use a state of 3 (stopped):(choose-vm) | foreach-object {$_.requestStateChange(3)}(click) Guess what – (click) The VM is stopped (click).That’s not a good way to turn off a VM (it’s the virtualisation equivalent of pulling out the power cord), so let’s do something more controlled on a VM with the Hyper-V guest integration components installed. (click) We can see that the machine is running (click) but if I type:(choose-vm) | foreach-object { (get-wmiobject -namespace "root\\virtualization" -query "SELECT * FROM Msvm_ShutdownComponent WHERE SystemName='$($_.name)' ").InitiateShutdown($true,"Because I said so") } (click) And select the correct VM (click) it starts to shut down and (click) eventually switches itself off, in a controlled manner.
• #35: Looking at the SCVMM interface, we can see how it allows the grouping of VMs (production, test and dev, by geographic regions, etc.); (click) how it allows for viewing VMs by status, (click) or by the users who have delegated control. (click) The libary provides a common location for virtual machine images, whilst (click) a live thumbnail provides a view of the current VM and (click) context-sensitive actions are presented.
• #36: Looking at the roadmap, Hyper-V RC0 is available now, along with a version of the Hyper-V Manager MMC for Windows Vista SP1. Hyper-V will RTM within 180 days of Windows Server 2008 (which was 4 February 2008) and my own feeling is that the product is shaping up well and will hit that milestone.An SCVMM vNext beta with Hyper-V support is also available now and the final version will be available later this year, although Microsoft will not commit to me whether this will co-incide with Hyper-V release (I believe there were some statements to that effect earlier on in the product lifecycle but they don’t seem to be the current position). The best information I have now is that RTM will be early in the second half of 2008SCVMM vNext will include support for managing VMware ESX, with Xen integration coming in the followingversion and is being positioned as a heterogeneous VM management solution (competitive products tend to focus on a single platform).There’s more about SCVMM vNext on Rakeshm’s VM Management Blog (http://blogs.technet.com/rakeshm/archive/2008/02/27/scvmm-vnext-is-getting-closer.aspx) but two features that are being banded around include expanded support for delegated administration and improved resource calibration and optimisation.
• #37: This slide shows the architecture for System Center Virtual Machine Manager vNext, although I don’t have any supporting information at this time.
• #38: So, we’ve looked at what virtualisation is about – in terms of the technologies and also the opportunity, we’ve also examined the architecture that Microsoft has used for Hyper-V and seen installation of the product on full and server core installations of Windows Server 2008. Finally, we looked at management, with Hyper-V Manager, and using the WMI API as well as a quick discussion of how System Center completes the picture.Just to recap, Hyper-V is Microsoft’s enterprise virtualisation platform and I am really excited about the product. It may not do everything that Virtual Infrastructure does, but it’s a lot less expensive (despite what VMware say) and most of the important items are there.It’s also worth considering some of the issues about managing a virtual infrastructure – the virtual reality.
• #39: This slide lists a number of resources to find out more about Windows Server 2008 Hyper-V.
• #40: As I mentioned earlier, the Windows Server Team UK website is where we are trying to build a community around Windows Server 2008 and my blog includes some information on Hyper-V as well as many other subjects.There’s also an article I was involved in writing about some of the management considerations for virtualisation on the Fujitsu Services website.I’ve also linked here to some good virtualisation blogs at Microsoft and elsewhere – although Dugie recently crossed to the other side and now works for VMware (http://blog.windowsvirtualization.com/virtualization/i-have-a-new-role-with-vmware-2)! Alessandro Perilli’svirtualization.info is useful as it is independent and cross-platform.As I mentioned earlier, the WMI demo that I gave is based on work by James O’Neill from the Microsoft UK IT Pro evangelist team, and James covers a number of subjects. Meanwhile, security and management fall under Steve Lamb’s remit.Finally, this slidedeck (including full speaker notes and additional information that I didn’t present today) is available on my SkyDrive (http://cid-1453622c71a8a08e.skydrive.live.com/browse.aspx/Public) – that’s not a very friendly URI so I will post a link on my blog.
• #41: And now a word from our sponsors ;-) Well, not strictly sponsorship, but they gave me the time to prepare this presentation and be here today! As I mentioned earlier, I work in the Microsoft Practice at Fujitsu Services – if you want to find out more about our partnership with Microsoft check out the link above or get in touch with me – either after this session, at this evening’s event, or through my blog.We now have a few minutes for a question and answer session. If anyone has any questions around the content of this presentation, I’ll do my best to answer them!
• #42: I hope you’ve enjoyed this presentation. If you have any comments, good or bad, please feel free to provide feedback to me via my blog at http://www.markwilson.co.uk/blog/ .Thanks for listening.
• #43: Copyright © 2008, Mark Wilson ([email protected]). Some rights reserved.This work is licensed under a Creative Commons Licence.You are freeto copy, distribute, display, and perform the work and to make derivative works under the following conditions:Attribution. You must give the original author credit.Non-Commercial. You may not use this work for commercial purposes.For any reuse or distribution, you must make clear to others the licence terms of this work.Any of these conditions can be waived if you get permission from the copyright holder.Nothing in this license impairs or restricts the author's moral rights. Your fair dealing and other rights are in no way affected by the above.This is a human-readable summary of the Legal Code (the full licence - http://creativecommons.org/licenses/by-nc/2.0/uk/legalcode).