Hyper-converged infrastructure offers an easy way to stand up a VDI deployment, but planning the size of an HCI...
implementation is key to making sure VDI runs smoothly.
When trying to right-size your HCI storage system, your primary concern should be the user experience. If you set up a system that does not deliver the necessary performance, you could end up with a failed VDI implementation. Only by carefully calculating the required CPU, memory and storage resources can you implement an HCI system that delivers the services necessary to keep users happy.
To appropriately size an HCI system, you must determine how many appliances to implement and consider the resource capabilities of each one. Before you can arrive at this number, however, you must calculate your virtual desktop requirements and the infrastructure needs to support them.
Calculate the virtual desktop requirements
Virtual desktop workloads can be varied and unpredictable. The better you understand your users and their work habits, the better you can estimate their usage patterns and correctly size the HCI system.
To start, determine the number of users you must support and when they work at their desktops. This shows how many people are on their desktops at any given time and helps you prepare for usage spikes from boot storms and virus scans.
Timing: As vital to VDI as it is to comedy
If 5,000 users all log in and work at the same time, you must allocate resources to support running those desktops concurrently. But if half your users work in another part of the world and work hours between and the two locations seldom overlap, you can allocate resources based on a lower concurrency rate.
In addition to calculating the concurrency rate, estimate resource usage during steady states, or the periods when users do regular work.
To calculate steady state resource usage, identify how each user works based on daily activities, then break users into categories that determine what resources their desktops require. Typically, these categories are task users, knowledge users and power users, but establish as many categories as your user base requires.
When assessing your users, use a system that accurately reflects how they work. The goal is to get as complete of a picture as possible of the resource requirements for all your desktops.
Once you understand how users work, you can determine who needs persistent desktops. Nonpersistent desktops require fewer resources, but they're suited to situations where user settings do not need to be maintained.
You should also determine what software needs to run within each virtual desktop. Software can have a significant impact on system resources, so you need a thorough inventory of everything in use and the type of resources they require. If you are planning to virtualize applications, you should estimate their resource usage separately.
Once you've collected all the information you need, you can determine the specific processing, memory and storage requirements for each desktop. Be sure to include the estimated number of IOPS, as well as the disk space.
Calculate the system-wide requirements
After you calculate your virtual desktop requirements, it's time to step back and look at the big picture: the software infrastructure that will support the virtual desktops.
To begin with, you must account for the host operating system running on each node, as well as the hypervisor. They require their own CPU, memory and storage resources, which can differ between products and editions. You should also take into account any components that interact with external systems and services because they too can affect resource usage.
Then there is the software that manages and maintains all those virtual desktops. For example, a VMware Horizon View implementation can include View Manager, View Composer, Virtual Infrastructure, a DHCP server and a database server. On the other hand, a Citrix XenDesktop implementation can include Desktop Delivery Controller, Studio, FlexCast and Virtual Desktop Provisioning.
VDI products can vary substantially in terms of what components are available and what you choose to implement. They might include databases, store services, application virtualization technologies, and the components necessary to balance virtual machine allocation. You must account for everything to ensure that you have the necessary compute and storage resources to support the server-wide systems without negatively affecting virtual desktop performance.
When calculating HCI storage requirements, include both the physical disk space and anticipated IOPS. You must account for the resources required for the software to carry out its operations, as well as the files that support those operations.
You should also consider the resources necessary to implement clusters, failover systems and disaster recovery structures. If your system is based on XenDesktop, you might want to stand up two Desktop Delivery Controllers and two StoreFront servers to provide data redundancy. If your VDI implementation uses a database, back it up.
Some HCI appliances, such as SimpliVity's products, include built-in data protection that provides backup, replication and restore capabilities. Even in such cases, you must account for the resources they require to house the software, run operations and store the data, just like you do for all the components.
Right-size your HCI storage platform
After you've calculated the resource requirements for your virtual desktops and software infrastructure, you should have a good sense of the amount of processing, memory and storage necessary to deliver the desktop services. You can then compare this information to the available capacity of each appliance to determine how many desktops it will support and how many appliances you'll need.
One of the challenges with right-sizing your HCI storage system is that appliances often come with fixed resource ratios. If your VDI operations fit neatly into the workloads supported by those ratios, you need only add an appliance to scale out your implementation. If your workloads require additional processing power, but nothing else, you might still have to purchase the entire appliance because you cannot scale up HCI storage incrementally. In such cases, you have no choice but to overprovision some resources to meet the needs of others.
Hyper-converged infrastructure can be an assembly required play
Hyper-convergence can boost the VDI storage game
Listing the benefits of HCI for VDI implementations