Cost-saving technologies remain a priority for IT in 2011 and virtual desktop infrastructure (VDI), with its ability to streamline operations, is one of the technologies at the top of the list.
With VDI, IT administrators can manage desktops and applications from a centralized location, eliminating the need to physically touch and update every single desktop. This, in turn, enables faster provisioning and deployment - a framework that is especially attractive for rapidly expanding computing environments. End users also benefit, gaining the ability to seamlessly access critical applications from any location with a myriad of devices.
So what's the catch? Why do VDI pilots fail? As the computing landscape has changed, so have user expectations. With mobile and ubiquitous computing fast becoming the norm for most corporations, end users don't tolerate availability or performance problems. In fact, end user satisfaction has been identified as the No.1 factor in determining success of any VDI pilot/proof of concept (POC). If the plan includes thousands of desktops, ensuring the first hundred users' happiness is critical to satisfying the next hundred, and so on.
The network is key to VDI satisfaction, being the conduit by which the virtual desktop continuously feeds the VDI client desktop activity. This video feed "paints" the monitor's screen via a desktop presentation protocol, such as PCoIP, ICA or RDP. When the visual display depends on network performance, understanding the difference between LAN, WAN and VPN activity is critical to project success.
VDI pilots often stall when employees start accessing their desktops via WAN, VPN and other lower speed links. On the LAN, contention is usually a non-issue, so pilots that only involve LAN links can create a false sense of accomplishment. If the VDI team has not taken a baseline on WAN and VPN links to see what headroom is available for VDI traffic, there may not be enough resources for even a small pilot.
When looking at bandwidth, be sure to account for peak utilization and not just average use. If there are legitimate spikes of activity expected across the links, there must be room for those peaks once VDI has been added, or non-VDI users will complain.
Armed with this information, you can work with the VDI vendors to adjust protocol parameters to ensure performance within the available headroom. These parameters include things like screen resolution, audio quality, USB redirection, and other user experience settings.
Once configured, it is important to monitor the infrastructure continuously for sudden bursts in network load that drown out VDI users across the WAN. Over the course of the VDI pilot deployment, these optimizations will provide a stable baseline from which to extrapolate full deployment feasibility.
Real-time visibility is required
The network, shared storage, connection brokers, desktop hosts, application virtualization servers, Active Directory servers, DHCP servers, security gateways, etc. must all work seamlessly for the successful delivery of a VDI desktop.
The performance, availability and constraints of each infrastructure component impacts the quality of the end user experience. For example, when desktop logins are slow, it might be a connection broker problem due to a "login storm," but it might also be a lengthy anti-virus update that needs to be scheduled to run after login. A sufficiently granular performance management solution is the only way to peer into these critical seconds.
The many moving parts in a VDI ecosystem demand an accurate, timely and comprehensive picture, or reactive management will be the sad reality. Performance problems can't be re-created from historical logs and/or disparate reports from a small subset of the virtual desktop infrastructure components. A real-time system allows you to navigate through the infrastructure and drill down to where the issue is happening even while it is occurring. Administrators should look for a single dashboard that covers all the components - whether physical or virtual - so that they are not blind-sided by the fluctuations common in complex IT systems.
Response time, also known as latency, has a direct and immediate impact on end-user experience. Virtual desktop performance is highly sensitive to sudden shifts in storage latency as well as network latency across the WAN. If latency shifts aren't measured in real-time, VDI administrators will have their phones ringing off the hook with user complaints while the lights in their virtualization management systems are still green.
This puzzling situation occurs because most management solutions poll for data every 5 to 15 minutes, often averaging the data over large intervals as well. A 30-second latency hit is invisible on these intervals, but the user still complains or chalks it up to "poor IT support."
That is why it is essential to have a real-time system continuously analyzing the infrastructure and comparing current performance against the historical response time thresholds. If anything changes, administrators will be immediately notified. If the storage latency for a key host serving up VDI desktops has jumped to 80 milliseconds, several minutes is far too long to wait. Traditional threshold crossing events should be bolstered by recordings or other contextual data so that VI administrators, network engineers, and storage administrators can collaborate on a common interface.
Give power to the users
When evaluating VDI management tools, special consideration should be given to solutions that directly capture user experience. If users can trigger a DVR recording of their activity when problems occur, administrators can capture the real-time load on the infrastructure, for example CPU, memory, storage and the activity in and out of the desktop. This rich information eliminates the impossible task of trying to "re-create" the problem. Load balancers, dynamic cluster rebalancing algorithms, and on-demand resource schedulers ensure that this morning's infrastructure configuration will be entirely different in the afternoon when they log back in.
A recording like this gives users a simple and proactive way to immediately communicate with IT as soon as they encounter performance problems — instead of learning about them a day later during a status meeting. By giving users a "visual trouble ticket," the service desk also benefits because it has immediate, actionable user information.
With some analysts calling 2011 the year of VDI, it's well worth taking the time to understand the new management criteria essential to succeeding with VDI. By ensuring that your infrastructure performance management system is up to the task for a VDI initiative, you can meet and exceed user expectations, increase overall business productivity and improve operational efficiencies.
Xangati, the emerging leader in infrastructure performance management solutions, accelerates virtualization by providing critical VM visibility previously unavailable, enabling superior design, optimization and troubleshooting for the virtual infrastructure. Xangati, Inc. is a VMware Technology Alliance Partner and is tightly integrated with VMware vCenter Management Suite.
Read more about data center in Network World's Data Center section.
Join the CIO Australia group on LinkedIn. The group is open to CIOs, IT Directors, COOs, CTOs and senior IT managers.