How To Assess the Cost of Implementing VDI
Unfortunately, a proof-of-concept rarely reflects the true cost of taking a new solution into production. Given the many moving parts that a VDI implementation can involve, there is strong potential for unexpectedly high infrastructure costs and complexity.
In this post, I will outline how to scope and assess VDI projects, so that initial costs are included, as well as what will be needed to help users scale, provide high availability, and manage the system.
In selecting a VDI solution, make sure to have a detailed, informed conversation with your vendor to fully understand the real costs of the implementation both up-front and over time. The vendor should be able to provide concrete figures based on a TCO or ROI calculator; while these will always be estimates rather than guarantees, they will provide valuable information to help evaluate options and make the best choice.
The cost assessment process should include the following areas:
Virtual Desktop Host Servers
How many servers are necessary to run the virtual desktops for users? This should be a straightforward calculation based on the number of virtual machines than can be run on each host, and the total number of users in any given environment. Ask the VDI vendor for a sizing guide with their recommendations on density based on the types of virtual desktops you plan to run (e.g. for task workers using basic Office applications vs. power users who may need lots of graphics capabilities).
If the solution being evaluated requires shared storage, i.e. a SAN, it will be necessary to estimate the cost of both the SAN and the related high-speed interconnects typically required for their production usage. This can be quite expensive, and also has implications for scalability--as users scale, they will not only need to add more SAN capacity but also increase the interconnect speeds to account for factors such as boot storms. On the other hand, a solution that leverages off-the-shelf, direct-attached storage (DAS) will usually prove much less expensive both initially and over time. If the solution uses DAS, ensure that it will still provide high-availability so end users will have business continuity in the event of a server failure.
Traditional VDI architectures require organizations to install separate connection brokers, provisioning servers, load balancers and sometimes clustered SQL servers as well. Work with the vendor to determine how many servers are required to run these components, and add their cost to your estimate. You also need to consider the cost of any Microsoft and other software licenses needed to run on company servers.
One way to avoid these costs is to base the VDI implementation on a shared-nothing architecture, which eliminates the need for management servers (as well as the need for a SAN). For example, the Citrix VDI-in-a-Box software appliance integrates connection brokering, desktop provisioning, load balancing and user profile management and lets users connect servers so they comprise a highly available grid which balances loads automatically. All that's required are the virtual desktop host servers.
Though often omitted from proofs-of-concept, high availability is a critical element of any VDI implementation designed to increase availability and make sure users don't lose access to their virtual desktops. As discussed in our last post, a VDI solution which does not provide high availability out of the box will require you to re-architect your implementation as you scale to provide redundant connection brokers, provisioning servers, load balancers and high-speed interconnects. The shared-nothing architecture mentioned above makes it possible to maintain high availability as implementations scale without having to add these components, reducing both cost and complexity.
Scaling the typical VDI infrastructure means more than simply adding virtual desktop hosts. It also means proportionately scaling the various other moving parts that make up the deployment to ensure optimal performance. If the VDI solution relies on shared storage, scaling may create bottlenecks such as boot storms that can only be overcome by over-provisioning with high-speed interconnects and higher-end SANs.
As an alternative, consider a solution that can scale in cost-effective steps. A solution with an all-in-one architecture that eliminates the need for separate connection servers, management servers and shared storage pools will also eliminate the kind of centralized bottlenecks that slow down traditional VDI architectures. As a result, users can scale more affordably in phased deployments with no need to re-architect the solution to ensure optimal performance.
Providing virtual Windows desktops to users requires the Microsoft VDA license. This is free for Microsoft customers with Software Assurance; otherwise, users need to include this cost in their estimates. It is not necessary to separate Windows licenses, as the Microsoft VDA license provides access to Windows. With traditional VDI, users may also need additional Microsoft Windows Server licenses for the servers which run management components. A solution integrated on a single box eliminates the need for these.
The cost of access devices can start from nothing if you re-use existing PCs or institute a "Bring Your Own" program. The costs go up from there when providing individual thin clients, tablets or other devices to each user. One popular strategy for SMBs is re-using existing desktops, then refreshing them one-by-one as they break or become obsolete. Look for software that will turn an existing PC into a locked-down thin-client; this can be a cost-effective way to get started with virtual desktops. Whichever approach is chosen, this is an important element of your estimate and should be calculated carefully.
Beyond the components above, keep in mind the time and resources it will take to roll out virtual desktops in production environments. Remember that as the number of components in the solution increases, so will the complexity of the architecture, the number of potential failure points, and the time it will take to diagnose and debug any issues that arise. To install and manage especially complex architectures, if may also be necessary to get a new staff with specialized skills.
A VDI implementation can deliver tremendous benefits, but it can also incur large up-front and ongoing costs. To make sure there is a solid ROI, be diligent about cost assessment, ask hard questions of each vendor in the mix, and choose the most cost-efficient solution.
We hope you've found this guide useful. In our next installment, we'll talk about how and when to use persistent versus pooled desktops.
Posted by Krishna Subramanian on 02/22/2012 at 12:47 PM