Large single-site heavy engineering plant, South Yorkshire (confidential)
A plant network designed around production, not legacy cabling
Industrial network redesign: resilient core, rationalised uplinks and zone-based access control
The problem (a legacy estate that had outgrown its own cabling)
Every production area depended on a handful of communications cabinets that had accumulated links to each other over years of extensions. A single failure could take out more than the area it served, and nobody could say in advance how far the impact would reach.
The site covers about 5.2 hectares across five production zones, served by 24 access cabinets. It had grown in stages. New buildings stood alongside old ones, clean production and finishing areas sat next to dusty ones, and successive cabling projects had been layered on top of each other. Each addition made sense at the time. Together they created inter-cabinet dependencies that cascaded, so failure domains were far larger than anyone had designed for.
That shaped day-to-day operations as much as outages. Changes were hard to plan because the knock-on effects were difficult to predict, and fault-finding started with tracing cables rather than reading documentation.
The estate was Cisco in principle. In practice, some of the cabinet additions had been built with cheaper switches, so capability, support status and configuration varied from one part of the site to the next. That variation made the dependencies harder to reason about, and any design had to work with equipment of mixed capability.
The route in was a conversation rather than a tender. The plant had appointed a new IT manager who had worked with Oxspring at a previous employer, and he called us for advice before committing to a plan.
The brief called for a topology with redundant cores at opposite ends of the site, cabinet uplinks rationalised into something the site team could understand and support, and access control that reflected how the plant actually operates. Replacing serviceable equipment was not the aim.
What “good” looks like (outcomes)
- Failure domains sized to the area they serve, with no cascading dependencies between cabinets.
- Redundant cores at opposite ends of the site, so losing a core or a single link does not stop production.
- A documented uplink hierarchy that can be changed without a site-wide outage.
- Access control aligned to production zones rather than to historical cabling convenience.
- One switching standard across the site, so the estate behaves predictably and can be supported as a single network.
- A topology the customer’s own team can support and extend without rediscovering it.
Risk & Challenges
- Live production throughout, with change confined to agreed windows.
- Mixed building stock, so distances, containment and environmental conditions vary across the site.
- A mixed switching estate, with capability and support status varying by cabinet.
- End-of-support access switches in around 40% of positions, which had to be replaced rather than carried forward.
- Existing fibre that could not be assumed to support 10 Gbps, so the survey had to prove it before design.
- Legacy patching that was poorly documented, which made discovery a prerequisite for design.
- Dirty and clean areas requiring different equipment choices and access arrangements.
- Cascading dependencies that made staged migration necessary, with a rollback position for each stage.
