Match your hyper-converged hardware pick with your workload
Overall workload demands and growth projections, not just user applications, should be top of mind when you choose among the hardware options in hyper-converged appliances.
Hyper-converged infrastructure vendors offer many hardware choices. You must look past end-user applications and consider your organization's overall workload needs before selecting the right hyper-converged hardware.
The first thing that you need to think about when selecting HCI hardware is why you are purchasing an HCI platform in the first place. Defining your objective could greatly reduce the amount of time it takes to make a selection. If your goal, for example, is to use HCI as a virtual desktop infrastructure host, then there are vendors that offer HCI systems preconfigured for VDI. Similarly, if you want to use HCI for backup and disaster recovery (DR) purposes, there are vendors that specialize in those functions.
Unfortunately, purpose-built HCI platforms do not exist for every conceivable workload, but if one does exist for your intended use, then it should be on your shortlist. After all, such systems may save you a substantial amount of money and will have software optimized specifically for your needs.
Match the hardware with the workload
Regardless of whether you can get a purpose-built HCI platform or not, you must consider whether the hyper-converged hardware will address your workload requirements, from back-end operations to user-facing applications. Suppose, for example, that you want to use HCI for backup. Although purpose-built HCI systems exist for backup and DR, they will only be viable if they contain sufficient storage to accommodate the data that needs protecting.
Hyper-converged hardware capabilities are also an important consideration for more general-purpose workloads. Suppose you are going to use HCI for several virtualized database servers that make extensive use of in-memory tables. In that case, the HCI nodes will need lots of RAM and will probably also benefit from high-performance storage. An entry-level HCI system with a mediocre amount of RAM and low-performance/high-capacity disks would not be a good fit for this particular workload.
Future-proofing
Workload demands often grow over time, so it is important to think about future capacity needs before buying an HCI system. Ideally, the platform that you choose should be able to handle the workload requirements you project for five years from now. Capacity planning estimates can end up being way off, however, so it is also important for an HCI system to be expandable in case you need additional capacity. There are a few different questions that you need to consider with regard to expandability of your hyper-converged hardware.
First, can you upgrade any of the components individually? Some hyper-converged platforms will allow you to upgrade things like memory and hard disks, just as you would on any other system. Most do not allow for component-level upgrades, and expanding capacity requires adding nodes, expanding all resources equally.
In spite of its reputation for simplicity, you must select a hyper-converged system as carefully as any other computing system.
Another consideration is the system's requirements for when you will add more nodes. Although adding a node or two seems relatively straightforward, some hyper-converged systems require you to add four nodes at a time and also require the purchase of a chassis when adding nodes. As such, you may need to think about how adding multiple nodes all at once would affect your IT budget.
How will you manage it?
Presumably, you already have management tools that you use to manage your organization's servers. The HCI system that you choose should ideally work with your existing management tools. Otherwise, you risk creating inefficiencies stemming from management silos.
In spite of its reputation for simplicity, you must select a hyper-converged system as carefully as any other computing system. Choosing an HCI platform that is not matched to your workload needs can negatively affect the workload's performance and availability. Furthermore, choosing the wrong hyper-converged hardware can lead to other issues, such as management silos, vendor lock-in or lack of capacity to handle future growth.
