Adding virtual desktops to your environment can put a strain on the data center network. But never fear: You just...
need a VDI network management strategy that keeps bandwidth up and latency down.
For virtual desktop infrastructure (VDI), most remote display protocols are designed to use as little bandwidth as possible and to tune the user experience to fit inside the available bandwidth. End users may experience an improvement in network performance because their virtual desktops likely reside on gigabit-speed networks alongside the servers they connect to on a regular basis.
An example of this is a user who accesses Microsoft Outlook on a remote desktop that communicates on the same network as the organization's Microsoft Exchange server. The WAN's I/O influences the user experience, which can be degraded by the number of concurrent user sessions sharing the link at any given time.
More on VDI network considerations
Ensuring VDI network redundancy and resilience
How to prepare your network for virtual desktop connections
Still, no organization has complete control over this network layer. The source of VDI network bottlenecks could be a user who connects to the virtual desktop via Wi-Fi at an Internet café. At different times of the day, users might connect to their virtual desktops with high-speed Wi-Fi links, but later they may be forced to use the less-reliable 3G connectivity of the devices themselves.
Allowing users to connect to their virtual desktops via 3G can be expensive and increase the risk of VDI network bottlenecks, since the remote displays consume bandwidth and eat into the monthly data allocation. Businesses might want to review the use of such devices as part of their process for selecting data plans to provide services.
Let's get low latency
Remote displays are not throttled by VDI bandwidth considerations alone. The latency of the links from the client to the virtual desktop and the number of dropped packets also affects stability and usability.
As latency grows, the number of hops that network packets must traverse also increases. So while the bandwidth might be rich at both the source and destination networks, organizations have no control over the quality or management of the Internet infrastructures between the client and the server. This can be a problem where variability exists, especially outside the U.S. and Northern Europe.
Users who experience this kind of latency may still be able to connect to the desktop but may suffer keyboard lag, where the screen is slow to update when they type. An average latency of 220 milliseconds (msec) between two locations can present a minor VDI network performance problem. A latency of less than 150 msec is ideal, but anything under 250 msec can deliver acceptable performance with traffic shaping and WAN optimization to mitigate peak congestion.
Latency becomes a real headache when it exceeds 350 msec, with intermittent spikes to 800 msec -- even if those latency spikes last only a few minutes each day and wouldn't be very disruptive for a single application. It might be assumed that users could work on another task while they wait for the latency spike to pass, but virtual desktops often present the entire platform, meaning users can't multitask during blocks of extreme congestion.