This content is part of the Essential Guide: Hyper-converged infrastructure options, vendors and installation tips

Hyper-converged platform configuration key to success in SMBs

Big companies that must support hundreds of VMs and scale quickly are prime HCI customers, but small companies can buy a smaller configuration and reap hyper-convergence benefits.

Hyper-converged infrastructure generally suits larger companies better than small ones, but small businesses can purchase a hyper-converged platform in a configuration that works for them.

Businesses deploy hyper-converged infrastructure (HCI) so they can think less about their data centers. For some large organizations, HCI is the only way to scale to hundreds -- or even thousands -- of hypervisor hosts.

Bigger businesses care a lot about scale. They want to know their infrastructure can accommodate a surge in virtual machine (VM) population if they acquire a new company or launch a new product. Big businesses also care about whether they can run ten thousand VMs and manage them all from a single console, which HCI allows them to do.

Policy-based management -- which is baked into HCI products -- is the only way to handle the huge number of VMs at these businesses. Larger companies might also like the ability to add to their hyper-converged platform over time as their workloads grow. In such cases, buying extra HCI nodes at regular intervals makes financial sense.

For small businesses without dedicated IT specialists for storage, virtualization and networking, HCI can be attractive. But finding the right configuration at the right price is a bigger challenge for smaller companies. Using HCI can actually make it difficult to control the cost of running only one or two dozen VMs.

Small businesses can still use hyper-converged platforms

Most HCI platforms require a minimum of three physical server nodes and an N+1 configuration for redundancy. HCI also requires some server resources to run the storage cluster. The result is that small businesses may only get to use half of the compute and storage resources they buy. Small businesses do not usually care that they could expand their HCI deployment. It is far more important that the product they choose to deploy is available in a small configuration.

Small businesses may only want a single CPU socket per node, and 64 GB of RAM may well be enough for each host. These nodes may only need a few terabytes of storage, and, provided some of it is flash, there will be ample performance, too. With a reduced VM count and lower performance targets, these nodes may not need 10 gigabit Ethernet. Scaling all of the resources down to match the intended VM workload reduces hardware costs.

That said, it is hard for some vendors to scale their overhead down for these clients. The design decisions that suit a 30-node cluster with half a petabyte of storage may not be appropriate for a three-node cluster with 5 TB. Some HCI platforms come in configurations for remote and branch offices, scaling below the usual three-node minimum.

One architecture is to have two physical nodes at each site and the third node in a VM at a remote site. This way, there are still three nodes, and each site has resilience against node failure. But each site only needs two nodes, which takes a third off the cost of the hyper-converged platform.

These configurations suit branch office deployments where the virtual third node is run at the head office. Multiple branches can each have two physical nodes with multiple virtual third nodes all running at the head office. This suits branches with a dozen or more VMs and approximately 100 staff members.

For some shops, two hypervisor nodes is still a lot of hardware. The only way to reduce the per-branch cost of HCI in this case is to forego resilience inside the branch. A couple of HCI vendors offer configurations for a single physical node in a branch. If the one host fails, then the branch has no local servers. To protect data inside the branch, all of the VMs are replicated, either to another branch, the head office or to both locations. If the branch node fails, the VMs can be recovered at the head office and accessed over the network. Some companies might accept the reduced application performance after a node failure in favor of cutting the cost of hardware per branch in half.

Most HCI platforms were designed with medium or large organizations in mind, but many HCI vendors have started to target smaller offices and businesses. The challenge is that scaling a large HCI configuration down might involve significant architectural changes to meet the required price, which is often the most important thing to a small business. 

Next Steps

Can you customize your HCI platform?

Possible problems with HCI configurations

Managing HCI right is critical to success

Dig Deeper on Hyper-Converged Infrastructure Management