Addressing VDI’s Fork in The Road

Organizations embark on Virtual Desktop Infrastructure (VDI) projects to drive down end-user operational costs and to enhance infrastructure security. These projects often evaluate well. During “test mode” there are a limited number of users, making it easy to meet their demands with a cost efficient, hard disk-based storage array. User counts increase as these systems move from test to production, and IT professionals face a fork in the road. To the left is continuing to use a hard disk-based array to keep costs per desktop down and executive management happy. To the right is upgrading the storage infrastructure to give the users the performance and data access they need, which encourages storage sprawl and increases the price per desktop.

Why VDI?

While the return on investment is harder to determine, a successful VDI project does indeed deliver a sizable return on investment. Organizations embark on VDI projects to drive down operational costs instead of hard costs. While administrators can reduce hard per-desktop costs, they cannot eliminate operational costs. Unlike server virtualization, which eliminates physical servers, in desktop virtualization the users still need an endpoint device of some sort. There should be some cost reduction since the user device can be a less powerful and less frequently upgraded computer; it can be a tablet or even a Chromebook.

The real gains are operational. IT administrators can manage more virtual desktops per-headcount than physical desktops. VDI environments significantly reduce the time to deploy, upgrade and maintain desktops. Additionally, the organization gains greater control and security over its end-point devices, which means better protection from deletion, loss or theft.

Fork Left: The Hard Disk Only Problem

The problem with a hard disk only storage system is that it can’t deliver the consistent performance that users expect. VDI is an unusual workload requiring modest performance most of the time with occasional extreme high-performance spikes. Also, techniques like thin provisioning, cloning, and deduplication, which VDI architectures use to keep capacity consumption costs down, are also storage I/O and storage compute intensive. That further hinders performance. If the performance of the virtualized desktop is not at least as good as their physical desktop, users resist adoption. If it is demonstrably worse, they will rebel against it. Most organizations come to the conclusion that user adoption is a higher priority than driving down the price per-desktop and look to upgrade their storage infrastructure.

Fork Right: The All-Flash Problem

Most IT professionals view all-flash arrays (AFA) as the quick way to solve VDI problems. The high performance and low latency of an AFA are not only able to handle VDI performance spikes, but also the write I/O demands of techniques used to lower capacity consumption. Of course, because of the higher cost per GB of the AFA, these techniques become a requirement. Even with a high return rate on data efficiency, the AFA still has a cost per GB problem that increases the cost-per-desktop. To overcome the cost-per-GB issue, IT professionals must increase the number of desktops per-physical host, which may involve buying additional or more powerful servers. Increased desktop density also increases the size of the failure domain; one server failure could leave thousands of users without a desktop.

The other problem with an AFA is that most are block-based storage systems, which is fine for hosting virtual desktop images, but users may prefer to access their data from a network attached storage (NAS) system or file server. While it is possible to map an existing NAS into the VDI design, IT professionals tend to purchase a NAS specifically for the VDI architecture citing performance concerns. The new system needs either to replace those NAS services with services of its own or compliment them.

A Silo of Performance Sprawl

In its attempt to address user’s performance concerns, IT creates a storage silo. The AFA dedicated to the VDI project is often not available to other environments like virtual servers or databases. In many cases, the VDI project is the data center’s first experience with an AFA. When they experience the performance improvement that the AFA brings, they naturally want other environments to also benefit. The problem is the AFA bought for the VDI environment is often too limited and an administrator will frequently purchase a separate AFA for other environments, creating several storage silos. Finally, even within the VDI environment, there is sprawl because of the addition of a dedicated NAS for VDI.

Merging the Forks: The Performance Tier Solution

In an attempt to address the performance demands of the VDI environment administrators replace storage systems before they reach the end of their lifecycle. Often these systems are not reaching their capacity limits. In fact, many have excess capacity, meaning the problem is a lack of performance. The AFA is an attempt to solve the performance problem by replacing the current storage system with capacity that is more expensive. Instead, IT professionals should consider adding a performance tier that compliments their current capacity instead of replacing it.

Architecting a VDI Performance Tier

Read heavy boot storms and virus scans occur in the morning, but during the day the VDI storage profile is write heavy. For example, thin provisioning and clones extend data volumes on the fly, as data and the metadata management tasks are significant. Hard drive-based systems will struggle to keep up with both of these I/O profiles, which is why so many vendors recommend replacing storage with an all-flash array. The reality is the above I/O profile is well suited for a performance tier that automatically caches active data and is the initial storage area for inbound writes.

A VDI storage performance tier should be flash-based and store all the dynamic data that the virtual desktops are accessing. When clones are used, storing the core virtual desktop image in the performance tier is straightforward since so many virtual desktops will count on the same image for most of their boot information. The performance tier should also store write data so that during the day activities also respond quickly. Primarily, the VDI storage performance tier acts as a shock absorber to more cost-effective disk capacity.

Designing a Complimentary Performance Tier

A popular alternative to all-flash arrays is hybrid arrays, but these systems still require a wholesale change out of storage systems. IT professionals who have a significant investment in their current storage systems may want to look for a performance tier that can complement the system instead of buying a replacement.

There are typically three options when designing a complementary performance tier; server-side cache, storage system cache or network cache. A server-side cache requires placing flash memory in every server in the environment. While it does provide local flash performance, it is not considered safe for write caching, as required in VDI. Buying flash devices for every server can get expensive and is extremely difficult to manage. The second option is to add flash to the existing array and leverage its caching or tiering capabilities. The problem with this choice is that while the original storage system may support flash, it may not be able to fully exploit it. It also may be limited in scale and buying extra flash capacity for each storage system can get expensive.

Network Caching

The final option is network caching, which are purpose built appliances that will take full advantage of the performance of flash. They are also highly redundant so caching writes are safe. A single network cache applies its performance acceleration capabilities to potentially every server and storage system on the network.

The entire VDI deployment and even to other environments like server virtualization and desktop virtualization can take advantage of the network cache investment. It also avoids creating storage silos and sprawls by providing flash-level performance across multiple storage arrays, accelerating your VDI environment any other application.


As VDI projects move into production, they reach a fork in the road. Does the IT team optimize performance or optimize dollars? Typically, they choose performance since user adoption is so critical to success. The result is a reduced ROI and, more than likely, a more modest rollout, which means a more limited gain in operational efficiency. If the organization selects a performance tier that can leverage the current storage investment while delivering the same performance as an all-flash system, it can have the best of both worlds.

Sponsored by Cloudistics

Twelve years ago George Crump founded Storage Switzerland with one simple goal; to educate IT professionals about all aspects of data center storage. He is the primary contributor to Storage Switzerland and is a heavily sought after public speaker. With over 25 years of experience designing storage solutions for data centers across the US, he has seen the birth of such technologies as RAID, NAS and SAN, Virtualization, Cloud and Enterprise Flash. Prior to founding Storage Switzerland he was CTO at one of the nation's largest storage integrators where he was in charge of technology testing, integration and product selection.

Tagged with: , , , , , , , , ,
Posted in Article
One comment on “Addressing VDI’s Fork in The Road
  1. […] To read the complete article, CLICK HERE […]

Comments are closed.

Enter your email address to follow this blog and receive notifications of new posts by email.

Join 22,223 other followers

Blog Stats
%d bloggers like this: