Quickly repurpose HCI hardware for hosting virtual desktops
Virtual desktops are tailored for remote work situations, and hyper-convergence makes a great VDI host that stands up quickly in a pinch. Consider these factors, though.
The COVID-19 pandemic has forced countless organizations to accommodate employees working from home. Versatile IT pros have recognized their existing hyper-converged infrastructure as a great option for hosting virtual desktops.
There are numerous benefits to this transition to virtual desktops. For starters, remote workers get a familiar, fully provisioned corporate desktop. Virtual desktop infrastructure (VDI) also improves security, because users access data from managed (virtual) desktops, rather than directly from personal devices.
In spite of the many benefits to adopting a VDI, repurposing an existing hyper-converged environment for hosting virtual desktops is not trivial. There are several factors to consider beyond simply what to do with the workloads currently running on your hyper-converged infrastructure (HCI), and whether your existing hyper-converged platform can function as a virtual desktop host.
High availability
High availability is one of the most important considerations when designing your virtual desktop deployment. Assuming your users are going to be entirely dependent on the VDI to do their jobs, you must consider the entire platform mission-critical. As with any other such workload, the VDI deployment has to avoid any single points of failure.
A main reason HCI is so often used as a virtual desktop-hosting platform is its built-in redundancy. It also scales easily to accommodate additional virtual desktops. However, redundancy must go beyond hardware. It's a good idea, for example, to have a highly available connection broker and redundant internet connectivity from at least two different providers.
Many of the innate benefits of hyper-convergence -- fast deployment, ability to scale to very large clusters, simplicity of integration and low-latency storage -- make hyper-converged systems a good fit for organizations looking to host virtual desktops for remote users quickly.
Virtual desktop persistence
Another important consideration is whether you want virtual desktops to be persistent, nonpersistent or a mixture of both.
Nonpersistent virtual desktops tend to be easier and more secure to manage than persistent virtual desktops. When a user logs in, the user gets a virtual desktop from a pool. When the user logs off, the virtual desktop is reset to a pristine state and returned to the pool.
Although examining virtual desktop memory requirements and available host memory is a good place to start the capacity planning process, storage ultimately determines how many virtual desktops a node can handle.
A persistent virtual desktop works more like a physical desktop. Rather than being assigned a random virtual desktop, the infrastructure gives users a dedicated virtual desktop. That way they can make changes to the virtual desktop -- such as installing applications or changing the wallpaper -- and those customizations carry from one session to the next.
Most organizations tend to use the same virtual desktop for everyone. But there are situations where it may make sense to assign persistent virtual desktops to certain users but make everyone else's virtual desktop nonpersistent.
VDI capacity planning
When planning to repurpose hyper-converged systems for a virtual desktop deployment, it is important to determine how many virtual desktops each node can host within your hyper-converged infrastructure platform.
Start by looking at the hardware used by your physical desktops. If, for example, most of your physical desktops have 8 GB of RAM and your hyper-converged nodes are equipped with 256 GB of RAM, then you might consider using 32 virtual desktops per node as a starting point.
Keep in mind, though, the actual number of virtual desktops a node can host may be higher or lower, depending on several factors. For example, it's possible your physical desktops are equipped with more memory than they actually need, and you might be able to get away with either allocating less memory to each virtual desktop or enabling dynamic memory. Either may allow you to increase virtual desktop density per hyper-converged node. At the same time, though, remember operating system hosting virtual desktops will consume some of each node's memory.
Although examining virtual desktop memory requirements and available host memory is a good place to start the capacity planning process, storage ultimately determines how many virtual desktops a node can handle. Each virtual desktop will need a specific performance from the storage system. So, the storage hardware must be able to provide enough IOPS to meet the needs of all the virtual desktops running on the node.
As you determine how many virtual desktops will work on each node, don't forget to account for activity spikes. Virtual desktops tend to be more demanding when first powered up and at other times, such as when users open applications. If several users do this at the same time -- for instance, when everyone logs on in the morning -- your virtual desktops can slow to a crawl, unless you've made adequate hardware resources available to each.
