Deploying VDI is complex, and it brings both cost and risk to IT departments. One way to reduce the risk is to...
use converged infrastructure or hyper-converged infrastructure.
Both converged infrastructure (CI) and hyper-converged infrastructure (HCI) aim to simplify the back-end resource layer for VDI and allow simple, size-based platform selection. But CI and HCI simplify things differently and are useful in different situations.
Hyper-converged vs. converged infrastructure
Converged infrastructure packages existing technologies: a set of servers, some network switches, a storage array and a storage network. These are the same parts IT can buy individually from a catalog, but with CI, the parts all come together. Converged infrastructure vendors specify every part and configuration option, and validate the collection of components, firmware, software and settings. The CI vendor even assembles the hardware into a rack and delivers a working virtualization platform to a company's site.
CI removes the uncertainty of building a virtualization platform and turns it into a single line item to purchase. Companies typically choose converged infrastructure systems based on capacity. For VDI, that is the number of light workload users the platform can host. Usual CI sizes are medium, large and extra-large. A medium converged infrastructure system might support 500 desktops whereas an extra-large might support 10,000. CI vendors include VCE, NetApp and Hewlett Packard Enterprise.
The hyper-converged vs. converged infrastructure debate is really over differences in architecture. Rather than simply using existing components as CI does, hyper-converged infrastructure redesigns the basic infrastructure for running VMs. One of the key characteristics of HCI is a node-based, scale-out architecture. Virtualization platforms usually have scale out for CPU and RAM with multiple hypervisor nodes, but the storage is centralized in an array that the hypervisor nodes share. HCI eliminates the storage network and array and replaces it with clustered storage. Software on each node forms the storage cluster and makes it available to the hypervisor nodes to run VMs. Usually a 10 GB Ethernet network serves as the storage network, but IT can also use it for management and VM networking.
An HCI cluster is a group of these nodes, usually with a minimum of two or three nodes. A typical node can handle 100 or 200 lightweight desktops. IT can add extra nodes until the cluster has the required capacity to support the company's VDI deployment. Another key feature of HCI is that adding nodes typically includes additional resources of all four types -- CPU, RAM, network and storage.
Hyper-converged infrastructure's scale-out architecture removes the risk that one of the resources will run out as the environment expands. With a conventional storage array, the storage performance is fixed. It can be exhausted by adding compute nodes for additional desktops. With an HCI platform, adding nodes also adds storage capacity and performance. Some vendors do offer flexibility in this area. For example, SimpliVity's OmniStack and Cisco HyperFlex platforms both allow customers to add compute-only nodes. To date, the most successful HCI vendor is Nutanix with its Acropolis platform, and SimpliVity is in second place, but the market should continue to expand.
How converged and hyper-converged infrastructure are alike
CI and HCI have some characteristics in common, such as simplified data center infrastructure management. Both architectures aim to simplify the deployment and operation of virtualization infrastructure. CI delivers a working infrastructure to a company's data center, and it requires very little configuration. HCI automates the setup process through wizards and scripts to ensure a quick deployment as soon as the product arrives. CI and HCI both use policy-based management.
Trials and tribulations of CI and HCI
One of the challenges of using a converged infrastructure system is that it is built for a specific VDI deployment size. If an organization's deployment is bigger or smaller than the available CI sizes, then IT may have to compromise. In this scenario, there are two options: pay for more resources than the company needs to support its VMs or try to strong-arm VMs onto a too-small CI stack, which hurts performance. Neither of these is very palatable in VDI.
CI is far easier to implement in large deployments where IT needs multiple CI pods to meet the workload. In large VDI environments, IT can usually manage the ongoing maintenance of the dedicated CI storage networks and arrays. Smaller deployments may lack the processes and controls required to keep consistent configurations and patching on these complex systems.
Still, HCI is not without its faults. Cost can become an issue at large scale. There is a resource cost to having software on each node act as the storage cluster. This is sometimes called the HCI resource tax and can be as much as 10% of each node's resources. These are resources that IT could otherwise use to run VMs if they were not consumed by the storage cluster.
HCI can also lead to challenges within an organization's IT personnel -- most commonly opposition from the established storage team. HCI effectively makes the storage team redundant and so is a threat to those jobs. Companies should understand before moving to HCI that it might end up affecting the structure of the IT department. Finally, HCI tends to come from startup vendors that lack the history or financial wherewithal of a more established enterprise supplier.
Both HCI and CI are good options for many VDI deployments, but IT ultimately has to decide between hyper-converged vs. converged infrastructure. Both simplify the processes of deploying and operating the virtualization platform that underpins a VDI deployment.
CI is likely to suit larger VDI deployments where there are larger support teams and where the larger block size won't limit a deployment. HCI suits smaller deployments where the simplicity of management is valuable along with the lower entry cost. HCI also suits a slower growing VDI deployment, whereas the large purchase cost of each CI unit would cause financial issues.
Three benefits of using hyper-converged infrastructure
Showdown: Nutanix vs. SimpliVity HCI for VDI
What HCI can and can't do for VDI deployment