Matthew Dodd — home Search the stacks

Consolidation as an operating-model change

Shifting infrastructure planning from like-for-like refreshes to shared enterprise platforms, with each consolidation gated on capacity and licensing, and its cost made visible to the business that uses it.

  • Medavie Inc.
  • 2026–now
  • Manager, IT Services
  • All work

Where it started

When a dedicated environment reached the end of its life, the default was an evergreen, like-for-like replacement: new hardware and licences for the same footprint. At the same time, Medavie Health Services and Medicentres were being consolidated into Medavie's data centres, alongside the VMware Cloud Foundation and VKS work.

The Medavie Health Services reporting environment was the test case. Replace it like for like, or fold it into a Medavie corporate cluster?

The decisions

Real consolidation, not a like-for-like refresh

My preferred target was real consolidation into a Medavie cluster, using existing capacity instead of buying dedicated hardware and licences.

Turned down: a like-for-like evergreen replacement, which renews the dedicated footprint and its costs for another cycle.

Gate it on capacity and licensing

Shared infrastructure isn't automatically cheaper. I asked for capacity to be validated first, because SQL Server and per-core licensing rules can change the economics. The dedicated investment would come down only if capacity and licensing supported it.

Turned down: assuming that consolidation always saves money.

Make consumption visible

Consolidation was tied to a consumption and chargeback model, so that Medavie Health Services would eventually see what infrastructure it uses and what it costs, rather than relying on less visible operating-cost estimates.

Where it stands

In progress. Infrastructure planning is moving from dedicated, like-for-like refreshes toward shared enterprise platforms, and Medavie Health Services' standalone infrastructure is being folded into shared Medavie Inc. infrastructure.

The budget ask for dedicated refreshes may come down significantly as a result. That saving hasn't been realized yet, so I don't claim a figure for it.