Fixed and dynamic Virtual Hard Disks (VHDs) are popular storage mediums for Microsoft Hyper-V environments, but...
By submitting your email address, you agree to receive emails regarding relevant topic offers from TechTarget and its partners. You can withdraw your consent at any time. Contact TechTarget at 275 Grove Street, Newton, MA.
there are scenarios in which these VHDs do not achieve the disk I/O performance that administrators require.
Luckily, Hyper-V also supports pass-through disks, which are persistently attached to virtual machine (VM) hosts. This storage configuration, as with all VM storage disks, has its strengths and shortcomings.
So why use pass-through disks for VM storage? How do these storage disks fit into your overall virtual environment and strategy? This tip provides insights into these questions and how to best implement pass-through disks in your virtual environment.
Why should you use pass-through disks?
IT administrators use this storage method for performance reasons. Because VMs access RAW disks without an abstraction layer, disk I/O capabilities can be greatly improved with pass-through disks.
Pass-through disks can also save you more disk space than a fixed-type VHD because you are utilizing only the space that houses the files. This configuration is similar to the space utilization of dynamic VHDs but without the performance penalty.
Note that the enhanced speed of pass-through disks is dependent on the disk subsystem type. Do not use cheap, slow storage disks and expect high levels of performance.
How to set up pass-through disks
Pass-through disks can use any type of storage that is persistently mounted to the host. Storage that connects through USB, Firewire or other external ports is not available as a pass-through disk. But iSCSI and Fiber Channel disks, as well as direct-attached storage, can be set up as pass-through disks.
For more information on how to configure pass-through disks, check out the Microsoft Core Team blog.
Host hardware considerations
Before using pass-through disks, consider your host hardware. When ordering hardware or architecting storage , consider the best possible disk subsystem for your budget. You can strictly utilize local storage disks, but many VMs connected to pass-through disks can saturate the disk I/O capabilities of the local disk controller and spindles.
Under ideal circumstances for this scenario, I would recommend iSCSI or Fiber Channel storage, because both methods are more flexible. These configurations, however, require that you have some sort of storage area network (SAN) infrastructure -- and more importantly, as many fast storage disks as you can afford. Ultimately, more disk spindles will collectively produce better disk I/O performance.
Backup infrastructure considerations
Using host-based Hyper-V Virtual Shadow Copy Service Writer backups, such as Diskshadow, is not possible with pass-through disks. But there are reliable ways to perform these types of backups. Since the number of VMs on a host is usually reduced, the use of more traditional backup strategies, like installing a local backup agent within the VM, can be used. With fewer VMs on the hosts, host resource saturation is less likely during VM backups.
In SAN infrastructures, SAN-based snapshots -- which have been used on UNIX servers for years -- are another commonly overlooked backup solution. Another option is to script a SAN-based snapshot of a pass-through disk at the SAN level, then mount it to another server that can be backed up to tape.
Depending on the workloads, you may have to perform some pre-snapshot work for a valid backup (e.g., a SQL backup job to ensure that the consistency of the database remains intact).
Live Migration and pass-through disks
Seamless live migration is not possible with pass-through disks. But these storage disks can be used in conjunction with Hyper-V clustering technologies and paired with other hard drive options (i.e., fixed or dynamic storage disks) that reside on a Cluster Shared Volume. The resulting configuration can be faster than Quick Migration but slower and slightly more disruptive than Live Migration.
Because pass-through disks are a part of a VM configuration within the Hyper-V cluster that needs to be dismounted and remounted on another node during the migration, there is a period of time, usually three or four seconds, where the VM is in the transitional state and not accessible.
Increasing disk I/O performance in virtual hosts
In most cases, administrators try maximizing disk I/O resources with pass-through disks on workloads that are more demanding than typical VM workloads. This setup, however, can affect your virtualization strategy, which may lead to fewer, more powerful VMs on your hosts.
Conventional wisdom focuses on maximizing the VM-to-host ratios. But a growing number of virtualization architects are running workloads that require more host resources. This alternative strategy emphasizes the management benefits over raising consolidation ratios.
Placing four virtual SQL servers on pass-through disks for adequate disk I/O, for instance, is a better and more cost-effective approach than running four physical servers. This tactic leads to better host resource utilization, easier management and quicker VM restorations.
It may be challenging to convince management to adopt this strategy, because it decreases hard cost savings that are traditionally seen with the higher VM-to-host ratios. But maximizing a host's resources and including multiple VM workloads that do not ordinarily run on the same virtual platform are still compelling reasons. In these scenarios, it's often necessary for pass-through disks to capture these types of workloads.
Pass-through disks are not popular in virtual server infrastructures, but they are useful for taxing workloads that need dedicated disk I/O resources for reliable performance. But provisioning fewer, more powerful VMs per host can be a mindset barrier that is difficult to overcome.
How have you creatively deployed pass-through disks in your virtual environment? Send me a note at VirtuallyAware.com.
About the expert:
Rob McShinsky is a senior systems engineer at Dartmouth Hitchcock Medical Center in Lebanon, N.H., and has more than 12 years of experience in the industry -- including a focus on server virtualization since 2004. He has been closely involved with Microsoft as an early adopter of Hyper-V and System Center Virtual Machine Manager 2008, as well as a customer reference. In addition, he blogs at VirtuallyAware.com, writing tips and documenting experiences with various virtualization products.