Virtual Desktop Storage Architectures

Tweet

 

To Persist, or Not To Persist – That is the Question
Back in May of this year I wrote about how all virtual desktops are not created equal.  I would like to expand on this now to talk about the different storage architectures, their uses, and how they apply to virtual desktop deployments. 

There are two primary storage-based architectures for virtual desktops:

• Floating Pool (non-persistent) storage
• Dedicated (persistent) storage

Floating Pool (non-persistent) storage

First, I will talk about the most common storage architecture associated with non-persistent virtual desktops – this storage design is commonly referred to as floating pool architecture.  Basically, the floating pool solution generates a fresh desktop for each user when they log in; the user’s  standard folders, user environment and other information is mapped to the desktop and everything they need is there for them to interact with.  If the host connection is lost, the user logs in again and receives a fresh new desktop. 

Because the only storage this solution manages is the master image of the floating pool, it relies heavily on Microsoft Active Directory for folder redirection and profile management (see the floating pool architecture diagram below).  It is no surprise that the two most important performance metrics for floating pool are the retrieval speed of the local storage medium for base images, and the responsiveness of Active Directory which depends on load and architecture.

As the state and personal information of each desktop is not maintained with floating pool, this storage architecture is optimal in situations where it is desirable for users to receive a fresh desktop each time they log in.  Such scenarios include shift worker environments like call centers and other environments where kiosk type environments are used.

Dedicated (persistent) storage

Dedicated virtual desktops are the most similar to physical desktops because they keep all the files for an individual’s desktop catalogued and indexed.  This allows each person’s desktop to be personal and uniquely theirs, while providing all the benefits of pool grouping and management.  There are additional benefits that are unique to this type of architecture including offline use and the ability to move desktop information between environments.

For the discussion on dedicated virtual desktops below, I would like to simplify by grouping them into the two major types of dedicated storage pool solutions: dedicated desktops with shared storage, and dedicated desktops with local storage, and then introduce how hybrid storage can deliver the most scalable and measureable solution for desktop replacement grade virtual desktops.

Dedicated virtual desktops with shared storage

Shared storage architectures closely mirror server virtualization technologies, and gain all the benefits of server virtualization including fault tolerance, VMware vMotion, and high availability.  This model is commonly used for "converged infrastructure" models because your server virtualization and desktop virtualization technologies can share a common fabric and therefore in theory be managed by the same team as generic computing resources. Because the desktop images are stored entirely on network-based storage (see diagram below), the performance of the virtual desktops is throttled by the performance of the (SAN) with metrics like IOPS, Throughput, and Read/Writes, and the performance of the network. This storage architecture is optimal in scenarios where you want to offer additional virtual desktops in the cloud, due to its ability to leverage overlapping resources with server-based cloud resources.

 
Dedicated virtual desktops with local storage

Dedicated virtual desktops with local storage architectures are becoming increasingly common in smaller implementations. They use local storage to provide a “VDI in a box” solution, which greatly simplifies the number of components that must be managed to make VDI work.  In this model, performance is measured similarly to physical desktops, except that it is shared among the number of desktops on the solution.  Appliance-based solutions like V3’s use this technology to provide guaranteed performance as well as guaranteed utilization (or number of desktops supported on each appliance).

Generally speaking, because all of the user’s personal information is kept locally on each appliance (see the diagram below), the functionality utilized for the shared storage scenarios does not work and then scaling becomes an issue.  This storage architecture is optimal in situations where smaller numbers of desktops are being used and capex cost is the main interest in the architecture, as opposed to measurability, scalability, and opex cost.

Dedicated virtual desktops with hybrid storage

The V3 Appliance uses a flexible hybrid storage architecture (see diagram below) that combines local and shared storage to deliver guaranteed performance and utilization without sacrificing the shared storage features required to scale.  Being able to provide computing with local storage in different building block sizes simplifies scaling and architectures, while delivering new management tools for uptime and reliability.  

Just ask the many excited V3 customers about this hybrid storage architecture.  Many have chosen to replace their primary physical desktops with primary virtual desktops to enable new productivity, reliability, functionality and in general, new freedom to use whatever device works best for the work they want to do at any given time.

All virtual desktops are not created equal and shouldn’t be, just like their physical counter parts.  One should use the desktop which works best for their type of workload.  I hope that understanding these different architectures will help you select the storage architecture that works best for your needs.  Let me know your thoughts.