Windows Server 2012 R2 brings with it a host of new virtualization features, as well as improvements to existing features and capabilities. Refer to our ‘What’s New in Windows Server 2012 R2‘ article for a more general overview, but read on for a list of some of the top new virtualization features found in the R2 release.
Windows Azure Infrastructure-as-a-Service (IaaS) is built on the same hypervisor as Windows Server. This means that there is complete virtual machine compatibility between the private cloud, partner public clouds, and the Microsoft-owned public cloud. Customers now have to ask themselves: “Where do I want my service to run today?”
A compression engine is built into Live Migration in Windows Server 2012 R2 Hyper-V. The processor in hosts is often underused, so this engine makes use of this spare resource to compress the memory of virtual machines that are being moved before the memory pages are copied across the Live Migration network. Hyper-V will monitor the utilization of CPU resources on the host and throttle compression to prioritize guest services. Enabling Live Migration compression on networks with 10 Gbps or less without Remote Direct Memory Access (RDMA/SMB Direct) support will greatly reduce the time it takes to move virtual machines (not including storage migration).
Live Migration can be configured to leverage SMB Direct (Remote Direct Memory Access, or RDMA) on hosts that that NICs with support for this feature. This feature will provide hardware offloaded accelerated copy of memory pages using SMB 3.0 NICs. This can take advantage of SMB Multichannel to span multiple networks. SMB Direct Live Migration provides the fastest way to live migrate virtual machines (not including storage) from one host to another.
A crazy fact: Memory speed will be the bottleneck on a host with PCI3 support and three RDMA NICs for Live Migration!
This feature allows very interesting new architectures, especially where organizations have decided to deploy SMB 3.0 storage with support for SMB Direct. Investments in RDMA can be leveraged to move virtual machines very rapidly over these physical networks (with QoS applied for SLA). For example, Cluster Aware Updating (CAU) will be performed much more rapidly.
Virtual hard disks of the VHDX format that are attached to the SCSI controllers of virtual machines can be resized without shutting down the virtual machine. VHDX files can be up- and down-sized. Downsizing can only occur if there is unpartitioned space within the VHDX. This feature supports both Windows and Linux guests.
Live Resizing of VHDX files will be of huge value to those running mission critical workloads. It will also offer a new self-service elasticity feature for clouds.
New storage metrics for IOPS have been added to WS2012 R2. With these metrics, you can determine the IOPS requirements of virtual machines and put caps on storage activity. This will limit how much physical disk activity that virtual machines can create, and therefore limit the damage that activity spikes can cause to other virtual machines and their guest services.
One of the concerns with shared storage is the possibility of a race for storage throughput. Enabling Storage QoS will limit the damage that any virtual machine or tenant can do in a cloud.
WS2012 R2 Hyper-V allows you to clone a running virtual machine. This will create an exact copy of the virtual machine that is stored in a saved state. This feature supports GenerationID. That means you can use Live Virtual Machine Cloning to create Active Directory supported clones of a virtual domain controller that is not the PDC Emulator.
This feature will be useful for situations where you need to debug a production system or you want to perform tests, such as guest OS upgrades.
You can export a virtual machine with a checkpoint (formerly known as a snapshot) and you can export a checkpoint of a virtual machine.
Dynamic Memory will be supported in Linux Guest OS’s on Windows Server 2012 R2 Hyper-V. This will give much better memory optimization for Linux virtual machines, and it’ll allow for much greater densities. Linux distributions with this built-in Linux Integration Services for Hyper-V HyH support are already available.
There will be support for online backup of Linux guest OSs. This is not Volume Shadow Copy Service (VSS) for Linux, and it does not give an application consistent backup. Instead, a file system consistent backup is created by freezing the file system. This feature does require an upgrade of any already deploy Linux Integration Services.
You can configure up to 64 virtual machines to share a single VHDX file on some shared storage (such as CSV or SMB 3.0). The VM sees the shared VHDX as a shared SAS disk with SCSI-3 persistent reservations. This is for data volumes to create guest clusters, and not for shared boot volumes. It works with down-level guest OSs, such as W2008 R2 with the WS2012 R2 Hyper-V Integration Components installed. This feature is supported by Service Templates in VMM 2012 R2.
This will drastically simplify guest clustering, where virtual machines are used to create a highly available service at the application layer. This could eliminate the need for guest attachment to physical LUNs and will be accommodating to self-service deployment within a cloud.
The default period for asynchronous replication of the Hyper-V Replica Log is every 5 minutes, but this can be changed to every 30 seconds or every 15 minutes. This allows companies to choose the allowed recovery point objective (RPO) – the maximum allowed amount of data loss in time.
Hyper-V Replica can now be extended to a third site. This is an A-B-C extension, and not an A-B/A-C extension. For example, a company might replicate virtual machines from the primary site to a local secondary site. This might be configured to hdappen every 30 seconds. Replica virtual machines in the secondary site might be replicated to a distant third site (such as a hosting company) maybe every 15 minutes. In the event of an unplanned failover, this would give an RPO of 30 seconds in the secondary site and an RPO of 15 minutes and 30 seconds in the third site.
The performance and scalability of Hyper-V Replica has been improved. Maintaining historical copies of virtual machines in the secondary site is costly (IOPS). This has been reduced, so maintaining historical copies of your replica VMs will not punish your storage in the secondary site.
The crippled virtual machine connection of the past is being replaced by a Remote Desktop experience that is built into the virtualization stack. This has no dependency on the virtual machine’s networking. By default, this feature is disabled in WS2012 R2 Hyper-V and enabled in Windows 8.1 Client Hyper-V.
Things that Remove Desktop VM Connect allow you to do include:
· Copy & paste text/images.
· Copy files to/from the client desktop.
· Do session-based USB redirection. This means you might use a USB stick to copy files. It is not a USB dongle solution.
You can live migrate a virtual machine from WS2012 Hyper-V to WS2012 R2 Hyper-V. This could eliminate downtime when deploying WS2012 R2 Hyper-V. For example, you can deploy a WS2012 R2 Hyper-V host/cluster alongside an existing WS2012 Hyper-V host/cluster. You can then live migrate the virtual machines from the older platform to the new platform with zero downtime to the availability of the services provided by the virtual machines.
Note that you cannot do a Live Migration from WS2012 R2 Hyper-V to WS2012 Hyper-V. It is a one-way upgrade path.
This is the first time that Windows Server (and thus Hyper-V) is being developed with and released at or close to the same time as System Center. There has been closer than ever before cooperation within Windows Server and System Center (WSSC). That means you can deploy System Center immediately and follow it up with WS2012 R2 Hyper-V, without the long delays of the past.