Hello, Andy from Sanbolic here… back for another discussion regarding how a cluster file system can help organizations achieve far greater returns on their investments in one of the leading VDI solutions on the market today, Citrix XenDesktop.
Last week, while browsing through some articles on VDI (as I so often do), I came across a blog from a prominent and reputable Citrix XenDesktop VDI architect that caught my attention. (You can find it here: http://virtualfeller.com/2010/09/08/moving-from-a-smb-virtual-desktop-solution-to-enterprise-scale/.) In it, he answered a question that had recently been posed to him regarding when one should add more Desktop Delivery Controllers (DDC) and Provisioning Servers (PVS) to support a growing number of virtual desktops. In his response, the architect emphasized the need to “plan for the future” by designing and building an infrastructure that could support many more desktops than would initially be deployed, thus ensuring that the infrastructure would not have to be rebuilt somewhere down the road, which could significantly increase costs as well as the likelihood for something to go wrong. In other words, as the architect cannily pointed out, “do it now and be done with it.”
This was a great response to a question we hear organizations asking all the time, particularly those who are either considering implementing virtual desktops or who are in the early planning and design phases of deploying virtual desktops.
In conversations I have with IT administrators planning VDI deployments, the topic of discussion that so often takes center stage is scalability. The chief concern of these administrators is not how to initially deploy their virtual desktop infrastructure, but rather once their infrastructure is in place and in use, how to quickly and seamlessly expand it to support more desktops without incurring many of the drawbacks frequently associated with its growth (i.e., diminished desktop experience, additional administrative overhead, additional storage resources that lead to increased costs, etc.), all of which can detract from the initial benefits of VDI and hinder an organization’s ability to realize a greater ROI.
So how are many organizations achieving the initial benefits of VDI, while simultaneously ensuring they have the ability to quickly and seamlessly expand their infrastructures on demand without incurring any of the negative side effects that often accompany VDI expansion?
Answer: They deploy VDI in conjunction with highly scalable and highly available shared storage provided by the combination of a cluster file system and SAN storage.
Unlike file-based storage solutions such as file shares or NAS devices, which rely on file-sharing protocols (i.e., CIFS or NFS) that are prone to locking contention as additional I/O loads are generated by more applications running on more desktops, a cluster file system employs its own distributed locking mechanism that is not susceptible to the locking contention issues often encountered with CIFS or NFS in mid-to-large-size file-sharing environments.
Another issue frequently encountered by organizations deploying VDI with file-based storage solutions is the taxing of the network infrastructure as additional Provisioning Servers are implemented to support a growing number of virtual desktops. More servers means more traffic (vDisk streaming and storage operations) flowing over the same network, introducing the likelihood for network congestion to occur. This can make it extremely difficult, if not impossible to scale the VDI solution to meet future needs while maintaining the high levels of performance necessary to ensure a rich desktop experience.
Using a cluster file system and SAN storage, storage traffic is shuttled across a separate network dedicated to storage operations, allowing organizations to expand their virtual desktop infrastructures while maintaining a rich desktop experience.
Other storage options for VDI, such as replication, also introduce additional complexities that can severely limit or prevent the scale-out of virtual desktop infrastructures. With replication, a separate LUN is provisioned, formatted with NTFS, and assigned to each Provisioning Server in the PVS farm. Copies of the vDisks streamed to the VMs hosting the virtual desktops are stored on each LUN. Every time a change is made to the contents of a vDisk, the updated vDisk needs be copied to each LUN in order to ensure that each Provisioning Server streams the most recent version of the vDisk to the virtual machines. Instead of simplifying vDisk management, this storage option actually introduces additional overhead to the maintenance of vDisks, contradicting one of the primary benefits afforded by the use of Provisioning Services – simplified deployment and upkeep of vDisk image files.
The ability to scale virtual desktop infrastructures may be further impeded as new Provisioning Servers are added to the PVS farm to support an increasing number of virtual desktops. For each additional server, an additional LUN must also be created, forcing IT administrators to copy vDisks to a greater number of LUNs each time a change is made to the contents of a vDisk. Not only does this make vDisk maintenance more difficult and cumbersome, it can also lead to extremely poor storage utilization and increased storage costs as the same data is stored repeatedly on multiple LUNs.
Using a cluster file system, only one LUN is necessary, with all Provisioning Servers in the PVS farm sharing concurrent, block-level read-and-write access to data stored on the LUN. When the need, or desire to expand the virtual desktop infrastructure arises, administrators can add more Provisioning Servers and/or more storage resources “on the fly” (without incurring system downtime or impacting user productivity), improving I/O performance, while maintaining optimal storage utilization to minimize storage costs.
The combination of a cluster file system and SAN storage offers a complete storage solution that isn’t beleaguered by the limitations and caveats associated with storage options such as file shares, NAS devices, or replication. This allows organizations to expand their virtual desktop infrastructures quickly and seamlessly whenever additional demands are imposed upon the infrastructure or when organizations simply want to extend the benefits of VDI to a larger percentage of their users.
And when organizations deploy XenDesktop VDI with an advanced, all-purpose cluster file system such as Sanbolic Melio FS, additional benefits beyond scalability/scale-out can also be achieved! For example:
- Enhanced fault-tolerance via high availability of vDisks to maximize uptime for virtual machines hosting virtual desktops, and high availability of PVS databases to maintain total visibility into and complete manageability of the infrastructure when the server hosting the database fails.
- Simplified management of desktop and VDI component images as only one LUN is necessary to store all vDisks, with changes to the contents of vDisks available to all Provisioning Servers immediately and automatically.
- Data protection in the form of VSS-based (block-level) snapshots of shared volumes, providing reliable backup and recovery of vDisks at various points in time.
The ideal storage solution for Citrix XenDesktop VDI, the combination of an advanced, all-purpose cluster file system and SAN storage allows organizations investing in VDI to meet both current and future demands, without introducing the complexities that often accompany the build out of virtual desktop infrastructures. The end result, organizations realize far greater returns on their investments in VDI. And most important of all… they can proudly proclaim they’re glad they “thought ahead!”
As always, if you have questions or comments regarding any of the information presented in this blog, please let me know.
Notice: The views expressed in this blog belong to Sanbolic, Inc. and have not been authorized by Citrix Systems, nor do they necessarily reflect the views of Citrix Systems.