How to choose HCI? Three evaluation axes

Hyper-converged infrastructure (HCI) virtualizes what used to be separately purchased servers (compute), disk arrays (storage), and networking, integrating them in software on a common set of standard x86 servers, managed from a single interface.

The question customers ask most is "which is better, VergeIO or Arcfra?" It has no real answer. Their feature sets map onto each other almost one to one, and the spec sheets do not settle it. What actually makes the difference is three things: what your servers look like, how much you are willing to cut over at once, and whether your organization has an origin review requirement.

Axis one: how licensing is calculated

HCI licensing follows two mainstream models. The difference is not cheap versus expensive, but what unit pricing is based on. Apply a different model to the same environment and the bill's structure changes completely.

• One license per server (per node): one license per physical server, regardless of how many CPUs, cores, memory, or disks it holds. VergeOS uses this model.
• Priced by core (per core): priced on total CPU core count, with no separate charge for storage capacity or cluster node count. Arcfra Enterprise Cloud Platform (AECP) uses this model.

Which server configurations favor which model

• Few servers, but each one heavily specced (dual CPU, high core count, load concentrated on a few large machines): per node is relatively favorable, since core count is not in the pricing formula.
• Many servers, but each with a modest core count (spread across more mid-sized servers, or small clusters across multiple sites): per core is relatively favorable, since node count is not in the pricing formula.
• Storage capacity growing fast: neither model charges by capacity. The gap is most visible for environments coming from software-defined storage licensing priced per TiB.

Two things easy to overlook

The pricing unit is only half the bill. The other half is how many separate products you used to buy: virtualization hosts, software-defined storage, network virtualization, a centralized management console, and third-party backup software - previously five quotes and five renewals. Add up all five together, not just the hypervisor line.

Both vendors currently use subscription licensing. In practice, the procurement approach is to lock in pricing with a multi-year contract up front, with the cap and conditions on renewal increases spelled out in the contract.

Axis three: supply chain origin

Raise this at the first selection meeting, because it can override every other evaluation result.

Public-sector agencies, the defense supply chain, some financial institutions, and Taiwan subsidiaries of foreign companies have explicit origin review requirements for ICT products. Reviews typically ask three questions: where the vendor is legally incorporated, the background of its technical and R&D team, and the tier of the supply channel in Taiwan. The third question is the one most often overlooked. Even when the vendor's home base is straightforward, product and support may still pass through a regional channel in a third country.

Taking the two options on this page as examples:

• VergeIO: the vendor, VergeIO, Inc., is a US company headquartered in Ann Arbor, Michigan.
• Arcfra: the vendor is Singapore-incorporated ARCFRA PTE. LTD. (established May 2024), with a technical and founding team spun out from the international business of HCI vendor SmartX. Public information shows investors including Temasek's Vertex Holdings and Samsung Ventures.

The criteria for origin review fall under your organization's compliance, audit, and procurement functions, and vary by industry and tender. Our role is to help obtain a country of origin declaration and written compliance statements from the vendor, so your organization can judge based on the actual documents.

Once the documentation is in hand, evaluate it against your own requirements. Confirm the origin review first - the other two evaluation axes only matter after that.

1. Should backup and ransomware protection be replaced together

Switching platforms may mean redoing your backup jobs, restore procedures, and audit report formats from scratch. Both vendors include built-in backup, immutable snapshots (unable to be deleted even by an administrator before expiry), and remote replication, which can eliminate the separate line item for backup software. But if your audit process is already tied to a specific backup product, confirm the new platform is on that product's support list first.

2. Is the underlying data center switch fast enough

HCI moves storage traffic onto Ethernet. Traffic that used to run over a dedicated storage network - replica synchronization between nodes, rebuilds, live migration - now all lands on the switch. A legacy 1G environment will almost certainly need to upgrade to a 25G/100G leaf-spine architecture. Buy the platform without upgrading the underlying network, and performance problems will surface at the worst possible time. See Enterprise Networking for related planning.

3. Will operations staff need retraining

The management console changes, the terminology changes, and even the first troubleshooting step changes. IT staff familiar with the old platform have to relearn it, usually right as go-live overlaps. Training and knowledge transfer are unavoidable - they are part of the migration plan. Operator training, management console walkthroughs, and documentation handover should all be scheduled and written into acceptance criteria. See Services for details.