Hyper-converged infrastructure software aids IT flexibility
Implementing a software-based hyper-converged platform can make it easier to scale components separately, but it has drawbacks as well, like multiple support sources.
Software-based hyper-convergence is designed to bring down the cost of scaling a hyper-converged platform. Rather than tying the hyper-converged infrastructure to super rigid hardware requirements, vendors that offer hyper-converged infrastructure software base it on a thin software layer that runs on top of an organization's existing hardware.
Traditional hyper-converged infrastructure takes an appliancelike approach to IT infrastructure. Vendors build server nodes to exacting specifications, preloaded with a hypervisor and management software, and sell them as integrated products. There are compelling benefits to using a hyper-converged platform, but there are disadvantages as well. Software HCI products can overcome some of the limitations inherent in traditional hyper-converged platforms.
The main disadvantage to investing in a traditional HCI platform is that its appliancelike architecture tends to make scalability expensive. Each node in a normal hyper-converged architecture includes its own compute, memory, storage and perhaps network resources. These nodes typically plug into a chassis that provides power, cooling and connectivity.
Suppose for a moment that an organization purchases four hyper-converged nodes but then realizes that it needs additional storage for the workloads it runs on the nodes. Depending on which product it uses, upgrading the storage may not be an option. Remember, hyper-converged systems are sold as self-contained appliances, and making changes to a node's hardware may invalidate the vendor's support agreement. That means the organization would likely have to enlarge the size of its hyper-converged cluster by adding additional nodes, rather than simply upgrading the existing nodes. Not every vendor requires this approach, but the practice is fairly common.
Purchasing additional nodes just for the sake of gaining more storage essentially means paying for compute and networking resources that are not necessarily needed and, sometimes, additional software as well. Furthermore, the chassis used by hyper-converged systems can only accommodate a finite number of nodes. If the organization in this example had originally purchased a four-node chassis, then it would not only have to purchase additional nodes, but an additional chassis as well.
Software-based HCI equals added flexibility
The advantage of a hyper-converged infrastructure software approach is that it makes using hardware more flexible and cost-effective. An organization might build a software-based HCI that fits the hardware it already has. When it needs hardware upgrades, it can upgrade individual nodes -- or even individual components -- without jeopardizing its hyper-converged environment in the process.
On the surface, hyper-converged infrastructure software would seem to be an ideal path.
VMware, Hewlett Packard Enterprise (HPE), Red Hat and Maxta sell hyper-converged infrastructure software products, and the way that these solutions work vary from one vendor to the next. Generally speaking, however, components such as compute and storage are treated as pools of hardware resources that can be dynamically allocated to workloads on an as-needed basis. This is often know as DIY (do it yourself) hyper-convergence.
Nutanix also separates its software from the underlying hardware but relies on server and channel partners to package it on appliances rather than selling software directly to customers without hardware.
Hyper-converged infrastructure software offers many of the same advantages of traditional hyper-converged systems, while greatly loosening the restrictions on hardware. Even so, there is one major issue that must be considered.
Traditional HCI appliances include software and hardware that have been extensively tested and certified to work together. The hardware and software are sold by a single vendor, using a single SKU (stock-keeping unit), which greatly simplifies getting technical support.
A potential problem with software-only hyper-convergence is that the hardware and software may or may not be certified to work together. In some cases, this is a nonissue. Vendors such as HPE have created an entire architecture around their own hardware and software (HPE calls this a composable infrastructure).
Other hyper-converged infrastructure software vendors may offer a hardware compatibility list or a hardware reference architecture. This approach ensures that the hardware and software are fully compatible with one another but lacks the advantage of having a single point of contact for technical support.
Still other vendors may offer general-purpose, highly flexible, hyper-converged infrastructure software products that are completely hardware-agnostic. While such products offer the greatest flexibility, there are no guarantees that a random piece of hardware will work as intended with the HCI software.
