Hyper-converged infrastructure systems can fundamentally change how an organization approaches VDI, including reducing IT's management workload and making it easier to resolve data center issues.
Virtual desktops can benefit IT and workers, but only if a deployment is set up and managed properly, and has all the necessary back-end resources. One issue with VDI is that it's an easy scapegoat. Boot storms, CPU loading, density issues, profile entanglement, network congestion and virtual machine (VM) resource contention are all pain points that often result in fingers pointing at the VDI deployment.
Separate VDI from server infrastructure
Why does VDI cause problems? It boils down to one basic issue: Infrastructure for VDI is -- or at least should be -- fundamentally different than infrastructure for server workloads. The I/O profile, the memory footprint and the user-facing surface area of VDI are very different than a server environment.
One way to solve this problem is to separate VDI and server infrastructure. IT can simplify its job by using hyper-converged infrastructure (HCI) specifically for VDI. Hyper-converged systems include tightly integrated storage, network, hypervisor and compute resources delivered through a single pod-style platform. Hyper-converged infrastructure systems have a management interface that allows IT to balance resources and shape performance.
Organizations can also purchase infrastructure without that software layer -- the management interface -- which is called converged infrastructure (CI). CI provides many of the same benefits, but requires a little more IT management. Vendors such as Nutanix, SimpliVity, Cisco and Dell offer CI and HCI platforms, or IT can build one of these pods itself.
Hyper-converged systems that are tuned for VDI performance usually contain VDI-specific resource cards, such as offload or GPU cards. The VDI deployment lives entirely inside this one HCI box and doesn't need to share compute or storage with other infrastructure. One HCI or CI pod supports whatever volume of VDI users the IT staff sizes it for. If an organization needs to scale its deployment out, upgrading an HCI system is as simple as adding another node.
SimpliVity offers the best user density, with published Login VSI performance-testing results showing that one SimpliVity OmniCube node can support up to 250 task workers at a time. Most of the other HCI vendors can support roughly half that number of virtual desktops per node, although that's just one consideration when choosing a vendor. Nutanix, for example, offers 17 reference architectures compared to SimpliVity's seven, and it has a VDI Assurance program where Nutanix helps size the back-end infrastructure for a VDI deployment and provides the extra hardware if it gets it wrong.
Reduce the effect of data center issues
VDI is a user-facing data center technology. Exchange and SQL servers are technically user-facing too, but if an admin accidentally unplugs the Ethernet cable from the organization's Exchange server, all that will happen is that workers' email accounts will freeze for a few moments until the admin plugs the Ethernet cable back in. But if IT accidentally downs a VDI deployment, every user is suddenly unable to work. Workers are kicked out of their desktops, they lose their trains of thought and they are unproductive during that time. Worst of all, they are immediately incensed.
HCI-deployed VDI eliminates the cross-pollination that causes most problems. A sudden hit to the storage area network because of a SQL reindex won't affect VDI users whose desktops live on an HCI platform. Likewise, a memory leak in the Web server won't slow down performance at all. The reverse is also true: When a large number of users log in at 9 a.m., they won't cause a boot storm that affects the application server. HCI reduces the number and severity of potential issues in the data center.
Some experts argue that IT shops can reap these same benefits by segregating VDI on conventional infrastructure, which is true and false. Conventional enterprise data center infrastructure requires a lot of folks working behind it. There are storage guys, network guys, virtualization experts and more. HCI simplifies the management of these disparate platforms by providing insight into VM-level performance issues, and also correlating performance degradation to resource constraints.
For example, if a specific subset of users complains about slow performance in conventional infrastructure, the whole IT team has to come together and the storage guys say, "The logs don't show any problem." Then the compute guys say, "The servers are storage-constrained." Or a lone IT person in a smaller company will have to log into each system separately to correlate events and identify a root cause. Those processes are time consuming and rarely as efficient and effective as possible.
Those issues disappear in a robust HCI environment. Logging in to the management interface shows IT which VMs aren't performing well and then gives a reason for the issues. For example, it might show that both storage and compute are adequate, but that network fabric between the host and the Internet is congested. It might then suggest adding an uplink to the virtual switch to resolve the issue.
Finally, don't forget the boss in the corner office. Most IT managers like predictability, which the scale out architecture of HCI-deployed VDI provides. It only takes some simple math to figure out the cost per desktop as well as the total cost of ownership. And for broad VDI deployments, HCI or CI platforms can save administrators time, reduce errors, improve employee adoption and provide predictable metrics for management.
What HCI can and can't do for VDI
Compare Nutanix vs. SimpliVity HCI
How HCI improves VDI scalability