by Gary Orenstein

Organizations looking to scale desktop deployments and reduce costs with virtual desktop infrastructure (VDI) face critical choices to pick the right solution. Storage array vendors claim that their products can scale performance by adding controllers and disk shelves. But with mechanized storage, performance scaling is far from linear, making it impossible to guarantee that a number of users can be supported at a set cost.

With the performance and low latency of server-attached, NAND flash memory, VDI is guaranteed to scale with increasing numbers of users and at a far lower cost than traditional storage arrays. Performance and cost scale linearly with each VDI server added, whether the system is supporting 100 or 10,000 users. This article will focus on the performance guarantees, costs and energy savings of server–deployed NAND for VDI to help IT professionals protect their companies as they move to these environments.

VDI Basics
Centralizing desktop infrastructure simplifies management and offers an opportunity to consolidate costs if architected properly. But placing virtual desktops on the other side of a network from operating systems, applications and data has a dramatic impact on network traffic that cannot be easily solved with slower mechanical drives in storage arrays. The demand for low latency IOPS is too high.

Additionally, when user counts increase, determining the number of SAN arrays and disk drives needed to support the load becomes a guessing game.

Figure 1: Placing the entire VDI load on a SAN leads to performance bottlenecks.

Placing Virtual Desktop Images on NAND Flash in the Host
At the root of each user’s desktop is a clone of a golden operating system image. By moving golden images and clones associated with those images to NAND Flash within the server, the VDI solution now has the horsepower to drive desktop image creation and operation with ease.

Figure 2: Moving golden images and clones to flash in the server boosts IOPS for VDI performance.

Placing golden images and clones on flash memory located within the server provides a number of benefits:

• Greater number of virtual desktops per virtual server;
  
• Guaranteed performance, even during peak loads;
  
• Faster generation of clones from golden images;
  
• Low-latency for faster end user desktop experiences;
  
• VDI that is cost-effective enough to localize servers and eliminate network latency;
  
• Reduced reliance on expensive, complex external storage; and
  
• Lower overall cost per desktop.

Predictable Scalability Benefits
The success of any IT installation resides in the ability to scale. Using flash memory to augment a VDI installation means that administrators can easily calculate growth. Since a larger number of desktops are supported per server, scaling becomes much simpler. And with each server accompanied by an equivalent amount of IOPS performance, the overall desktop growth can be supported predictably.

Figure 3: Server-based flash enables more desktops per server.
 
Cost Benefits
VDI allows IT administrators to centralize and simplify the provisioning and administration of user desktops. And consolidating desktops to fewer servers provides a direct positive impact to the bottom line. More desktops per server cuts server spending, minimizes server software licenses and shrinks the data center footprint.

VDI, when architected properly, allows companies to streamline desktop deployment, scale performance and reduce costs. Customers considering VDI should investigate flash-enabled solutions to meet these demanding criteria.

Gary Orenstein is the vice president of products at Fusion-io (Salt Lake City, UT). www.fusionio.com