Virtualize Servers with
Microsoft Hyper-V and NetApp

Microsoft® Hyper-V™ is a new virtualization technology that is a key component of Windows® Server 2008. Hyper-V provides the ability to virtualize server resources and manage them within the Windows management framework and is a good choice for customers whose core infrastructure is built primarily on Windows Server and other Microsoft application technologies.

Like all server virtualization solutions, Hyper-V increases the demand on back-end storage in terms of reliability, performance, and data management capabilities. NetApp® storage solutions are well suited for Hyper-V environments, offering storage virtualization capabilities that complement and extend the core capabilities of Hyper-V. Key NetApp technologies help provide data protection, disaster recovery, deduplication, thin-provisioning, and efficient cloning for faster virtual server provisioning. Going forward, NetApp will continue its close partnership with Microsoft to provide the same deep level of integration for Hyper-V as it does for other Microsoft solutions such as Microsoft SQL Server™, SharePoint®, and Exchange.

	Before Virtualizing Servers*	After Virtualizing Servers
Number of applications per server	1	10+
Number of physical servers	10+	1
Number of apps down on storage failure	1	10+
Data lost on dual-disk failure	1x	10x
Backup data volume	1x	10x
Meeting backup window	Feasible	Maybe not
Provisioning	Slow/complex	Storage ≠servers

*Typical configuration: DAS, RAID 5, tape backup

Table 1) Impact of virtualization on storage infrastructure.

This article will help you understand Hyper-V with some guidelines for getting started using the technology in NetApp environments. It includes:

• Introduction to Hyper-V
• Storage considerations for Hyper-V
• Deduplication and thin provisioning
• Backup and disaster recovery

Introduction to Hyper-V

Microsoft Hyper-V (previously known as Microsoft Server Virtualization) is a hypervisor-based server virtualization technology that is an integral part of all Windows Server 2008 editions (as of its release in late June 2008). Hyper-V significantly extends the virtualization capabilities that Microsoft provided through its still shipping Microsoft Virtual Server product.

Hyper-V is designed to allow multiple virtual machines (VMs) to run unmodified on a single physical server while providing strong isolation between partitions. Its inherently secure architecture has a minimal attack surface with no third-party device drivers.

Among the noteworthy features of Hyper-V are:

• Support for 32-bit (x86) and 64-bit (x64) VMs
• Support for both Windows and Linux® guest operating systems (full list)
• Up to 64GB of memory per VM
• Up to 4 cores per VM (SMP)
• VLAN support
• Up to 12 NICs per VM (eight synthetic and four legacy)
• Virtual machine Snapshot™ copies
• Integrated cluster support for HA and VM migration (quick migration)

Windows Server 2008 can be installed using either a full or server core installation. The server core is a new minimal installation option providing essential server functionality, while eliminating nonessential code. This improves availability and security and reduces management and servicing overhead.

Hyper-V management is provided through System Center Virtual Machine Manager (SCVMM), part of the Microsoft System Center Suite of management products. SCVMM allows management of both virtual (Hyper-V, Microsoft Virtual Server, and VMware® ESX) and physical infrastructure from the same interface and assists with VM management, resource optimization, and both physical-to-virtual (P2V) and virtual-to-virtual (V2V) conversions.

See the Microsoft Hyper-V Web page for more general information on Hyper-V.

Figure 1) Microsoft System Center Virtual Machine Manager (SCVMM) architecture.

Hyper-V Storage Provisioning

Hyper-V supports three possible storage infrastructure options: direct-attached storage (DAS), Fibre Channel SAN, and iSCSI. However, many of the advanced features of Hyper-V, such as quick migration, require the use of shared storage technologies, making iSCSI or FC SAN a better choice than DAS for installations that need to scale beyond more than a few physical servers.

Hyper-V provides two options for presenting storage to virtual machines: virtual hard disks and pass-through disks. A third option is to install an iSCSI software initiator in the child OS and access iSCSI LUNs directly, bypassing the Hyper-V mechanisms. VMs may also access NAS (CIFS and NFS) file systems directly.

Virtual hard disks, or VHDs, allow you to assign storage to a VM in which the actual storage is kept in a VHD file located on a disk attached to the Hyper-V parent partition. The benefits of VHDs are the improved manageability and portability that come from having VM storage encapsulated in a single file. There are three different types of VHDs:

• Fixed-size VHDs allocate all of the storage for the VHD up front. Though the data storage requirements upfront can be large, NetApp technologies such as thin provisioning and deduplication allow you to recapture much of this storage. (See the following section.) NetApp recommends utilizing fixed-size VHDs.
• Dynamically expanding VHDs grow the VHD file each time data is added. While this method is extremely space efficient, it does present potential problems in terms of performance, fragmentation of the child file system, and alignment. NetApp recommends limiting use of this type of VHD for these reasons.
• Differencing VHDs are created not at the time the VM is created, but, as an example, when a Hyper-V Snapshot is created of an existing VM. A differencing VHD points to a parent VHD file and grows in a manner similar to a that of dynamically expanding VHD, and therefore it suffers from the same performance concerns.

Pass-through disks are disks that are attached directly to the Hyper-V parent, but assigned directly to a virtual machine and formatted with the child OS file system. One limitation with pass-through disks is that Hyper-V Snapshot copies are not supported. Because of this fact, NetApp recommends limiting use of pass-through disks in your Hyper-V environment, except where considered necessary.
For more information on the performance characteristics of VHDs and pass-through disks, refer to the “Storage I/O Performance” section (pages 65–67) of the Microsoft Performance Tuning Guidelines for Windows Server 2008.

With VHDs and pass-through disks, you have the further option of presenting them as either an “IDE” or “SCSI” type device to the child. The limitations of each of these options versus direct access via iSCSI are summarized in the following table.

	DAS or SAN on Host, VHD or Pass-Through Disk on Host, Exposed to Guest as IDE	DAS or SAN on Host, VHD or Pass-Through Disk on Host, Exposed to Guest as SCSI	Not Exposed to Host, Exposed to Child as iSCSI LUN
Child boot from disk	Yes	No	No
Additional SW on Child	Integration components (optional)	Integration components	iSCSI SW initiator
Child sees disk as	Virtual HS ATA device	Microsoft virtual disk SCSI disk device	Microsoft virtual disk SCSI disk device
Child max disks	2 x2 = 4 disks	4 x 64 = 256 disks	Not limited by Hyper-V
Child hot-add disk	No	No	Yes

Table 2) Hyper-V storage comparison. (Integration components install drivers to optimize the performance of a VM. These drivers provide support for synthetic I/O devices, which significantly reduces CPU overhead compared to emulated I/O devices.)

A Hyper-V server can access LUNs on NetApp FAS storage systems using either Fibre Channel or iSCSI. LUNs must be masked so that the appropriate Hyper-V parent and child partitions can connect to them. With a NetApp FAS system, LUN masking is handled by the creation of initiator groups (igroups). NetApp recommends creating an igroup for each Hyper-V server, cluster, or child VM (when using direct LUN access by child OS with the iSCSI software initiator). NetApp also recommends embedding the name of the igroup and the protocol type in the name of the Hyper-V server, cluster, or child VM. If a Hyper-V server or cluster will use both Fibre Channel and iSCSI protocols, separate igroups must be created for Fibre Channel and iSCSI.

You can find complete recommendations and instructions for configuring NetApp aggregates, FlexVol® volumes, and LUNs for use with Hyper-V in the NetApp and Microsoft Virtualization Storage Best Practices Guide.

Using NetApp Deduplication and Thin Provisioning with Hyper-V

NetApp offers thin provisioning and deduplication capabilities that enhance the virtualization of storage used by Hyper-V, providing considerable storage savings. Both technologies are a native part of the NetApp Data ONTAP® operating environment and don’t require any special Hyper-V configuration options.

Virtual server environments typically have high levels of data duplication with numerous copies of nearly identical operating system and application code for various VMs. NetApp deduplication can eliminate this deduplication, yielding typical storage savings of greater than 50%. By reducing the amount of storage consumed by virtual environments, NetApp deduplication also significantly reduces the bandwidth and cost needed for replication, making disaster recovery configurations more economical.

To recognize the storage savings of deduplication with LUNs, you must enable NetApp LUN thin provisioning. The value of thin provisioning is that storage is treated as a shared resource pool, and additional storage is consumed only as each individual VM requires it, increasing the total utilization rate of storage.

NetApp thin provisioning allows LUNs and VHDs to be provisioned to their total capacity (fixed size VHD), yet consume only as much storage as is required to store the actual VHD files. Pass-through disks can also be thin provisioned. See the Hyper-V Storage Best Practices Guide for more details.

Figure 2) Advantages of thin provisioning.

Data Protection and Disaster Recovery

Virtualizing a number of servers on a single physical server raises the stakes for both data protection and disaster recovery. A problem that might have affected a single application in the past now has the potential to impact dozens of applications. Because the risks are higher, the measures you take to protect data must increase accordingly.

These problems may be further compounded by higher server utilization and limitations on I/O bandwidth. A single multiprocessor server might have more than enough I/O bandwidth to accommodate the needs of multiple applications during normal operation, but might not have I/O bandwidth equivalent to the multiple physical servers it has replaced. This can become apparent during backup.

The solution is to offload the I/O that results from backup and disaster recovery processes from the server to the storage system as much as possible, freeing server CPUs and I/O channels for VMs and associated applications.

NetApp facilitates offloading these workloads from Hyper-V servers with its Snapshot, SnapVault®, SnapManager®, and SnapMirror® technologies.

• NetApp Snapshot copies provide an instantaneous, point-in-time view of volumes storing VM data and are highly space efficient because they only consume additional storage as changes are made.
  
  

  Figure 3) Using traditional backup in a virtual environment versus NetApp Snapshot.

• NetApp SnapVault makes it possible to back up saved Snapshot copies to secondary storage for longer-term retention.
• The NetApp SnapManager suite of products simplifies the creation of fast, consistent application backups for popular applications (Microsoft SQL Server, Exchange, Oracle, SharePoint Server, and SAP®) running within VMs.
• NetApp FlexClone® can be used to clone one or more VMs for testing or other purposes and can also provide the basis for extremely space-efficient virtual server provisioning.
• NetApp SnapMirror simplifies the replication of important VM and application data to a remote site for disaster recovery.

Conclusion

Microsoft’s new Hyper-V technology provides a full-featured server virtualization environment that might offer particular advantages for sites that rely primarily on Windows infrastructure for their core operations. NetApp has made a strong commitment to support Hyper-V with full integration. NetApp can be deployed today to provide back-end storage for Hyper-V (either iSCSI or FC SAN). You can already take advantage of NetApp technologies such as Snapshot, SnapMirror, FlexClone, and so on to simplify data management in your Hyper-V environments. Going forward, we will increase the level of integration of NetApp products with Hyper-V and its associated management tools to further simplify and enhance operations in Hyper-V environments.

Brian Gracely Manager, Reference Architecture NetApp Brian manages the Virtualization Reference Architecture team within NetApp’s Virtualization and Grid Infrastructure business unit (VGIBU). His team of Microsoft Hyper-V experts is based in Seattle, Research Triangle Park (RTP), and Bangalore and is focused on design, implementation, and training for Hyper-V deployments using NetApp storage.