Case study 06 · Business case

Same node, new price

Red Hat moved to one subscription per node, so I re-ran the comparison at public list prices.

My role
Solution architect
Context
An air-gapped digital engineering platform on OpenShift and IBM Fusion HCI
Stack
Bare metal OpenShift, IBM Fusion HCI, Red Hat subscriptions

Problem

The platform was designed for bare metal OpenShift, where one subscription covers the whole node regardless of virtual CPUs. The licence saving against core-based subscriptions, as published in the IBM Fusion product guide in 2023, was part of our case.

Red Hat has since changed the bare metal subscription, and from January 2026 there is one subscription per physical node, regardless of sockets or cores. The unit is simpler, but the price of a node subscription has risen relative to Red Hat’s core-pair subscription. Did the case still hold?

Constraints

  • Public prices only. The comparison uses Red Hat’s public list prices, with no organisational or contract pricing.
  • Prices move. Subscription terms and prices change over time, and customer pricing varies.
  • A scale-out design. We favoured more, moderately sized workers over fewer high-capacity servers, because that spreads workloads across more failure domains and reduces the impact of maintenance and hardware failures. A subscription per node now places a higher relative cost on exactly that design.
  • Pricing built for virtualisation. Unlimited cores per node suit organisations consolidating virtual machines onto fewer, high-core-count servers. Many microservices scale the other way, spreading replicas across nodes for resilience.

Decision

Re-run the guide’s comparison with its own method, at current public list prices, comparing core-pair subscriptions with a bare metal node subscription for an IBM Fusion HCI worker node.

Publish it as a community proposal. I wrote a proposed revision of the product guide’s cost comparison and published it on the IBM TechXchange Community, labelled as a community proposal, not an IBM publication, and revisited our own business case on Medium.

Tell the owners. I contacted the IBM Redbooks team to flag the section and recommend an update in the next revision.

Count resilience, not just subscriptions. Consolidating onto fewer, denser nodes may reduce subscription costs, but a failed dense node takes a larger share of the cluster with it, each drain for maintenance becomes a larger disruption, and keeping equivalent resilience can need extra hardware that offsets the licence saving. Our scale-out design proved itself when a node’s hardware failed: the platform kept running while replacement parts were sourced and fitted.

Outcome

Still lower
OpenShift licensing on bare metal, for every IBM Fusion HCI worker node
Smaller
saving than the product guide states
  • Bare metal is still the lower-cost option for every IBM Fusion HCI worker node, but the advantage is no longer as large as the guide states.
  • Our decision to scale out was driven by workload requirements, not licensing economics. Those requirements haven’t changed, but the financial incentive now favours a different architecture. We kept the design.
  • The recommendation: model OpenShift subscriptions against Red Hat’s current subscription guide, not the figures in the 2023 product guide.
  • The full comparison, with its figures and assumptions, is in the two articles below.

For your team

Re-run the business case when the vendor changes its price list, not only when you buy. The cheapest architecture to license is not necessarily the cheapest, or most appropriate, architecture to operate.

My role

I re-ran the comparison, published it on Medium and the IBM TechXchange Community, and contacted the IBM Redbooks team with a proposed revision.

Re-running a platform business case? Let’s compare notes.

Connect on LinkedIn