While the growth of the size of hyper-convergence infrastructure deals indicates HCI is increasingly moving into...
enterprise deployments, there are still plenty of hyper-converged infrastructure use cases for single applications.
When hyper-convergence first came on the scene several years ago, many vendors marketed it as an easy, low-cost, turn-key environment for running virtual machines (VMs). HCI is still widely used as a virtualization platform, particularly in medium-sized businesses and branch offices. To understand why HCI suits specific tasks, one must understand a little bit about the technology's advantages and limitations.
HCI initially gained popularity primarily because of its ability to simplify IT operations. HCI combined virtualization software with performance-matched hardware. This hardware was certified as fully compatible with the hypervisor and any supporting software such as drivers and management tools. The entire HCI bundle was sold as a single product, with one single source for technical support. As such, organizations often adopted HCI in hopes of avoiding common headaches such as hardware compatibility issues or vendor finger pointing.
While the advantages of using HCI are undeniable, there are more appropriate hyper-converged infrastructure use cases, because HCI isn't well-suited to every workload. The biggest problem with prepackaged HCI is that hardware configuration tends to be relatively inflexible. This might not be an issue for organizations that need to host relatively mundane, run-of-the-mill virtual machines, but it can be a big problem for organizations that need to run high-performance workloads. For this reason, HCI is widely regarded as a poor choice for Hadoop clusters, for example.
HCI also tends to be a poor choice for virtualized workloads that require a lot of storage. This is because prebuilt HCI deployments often make it impossible to upgrade a node's storage alone. An organization that needs to add capacity will have to purchase additional nodes to get the storage it needs. This also means that the organization is paying for CPU and networking resources it might not need.
The right hyper-converged infrastructure use cases
The bottom line is that HCI has advantages, but generic HCI deployments aren't the best fit for every workload. But some vendors are starting to custom tailor an HCI deployment to a workload's exact requirements while still retaining all of the advantages that made HCI so popular in the first place.
Consider, for example, the requirements of a backup appliance. A backup appliance requires beyond all else, a large amount of storage -- far more than would typically be found in an HCI deployment. As such, a backup vendor who bases its product on HCI would undoubtedly include as much storage as possible.
This brings up an important point with regard to backup appliances as hyper-converged infrastructure use cases. If storage was the only important thing, a backup vendor could create a NAS-based product at a significantly lower cost than an HCI-based product. HCI becomes a good fit for backup because most modern backup systems support instant recovery for virtual machines. If a VM needs to be restored, the VM can be temporarily hosted directly on the backup server while a traditional restoration happens in the background. Because HCI was initially designed with virtualization in mind, it lends itself really well to hosting instant recovery workloads.
Virtual desktops take advantage of the HCI deployment's virtualization capabilities. But there is another aspect that makes HCI so well-suited to the task of hosting virtual desktops.
In a virtual desktop infrastructure (VDI) environment, each virtual desktop is generally capable of hosting a certain number of virtual desktops. This number varies depending on the underlying hardware and on how the virtual desktops are used. Even so, hyper-convergence makes virtual desktop capacity planning easier. Because each node in the HCI deployment can handle a specific number of virtual desktops, the organization can scale the VDI deployment in a predictable way by simply adding nodes.
Backup and VDI are just two examples of single-purpose hyper-converged infrastructure use cases that have been adapted to take advantage of an HCI architecture. In the future, we are likely to see far more single use cases for HCI. This does not mean, however, that HCI's role as a virtualization platform will suddenly go away. HCI will remain a viable option for the foreseeable future for organizations that need to host general-purpose virtualized workloads.