300 servers decommissioned and returned from a constrained London site

The client owned around 300 1U servers across 10 cabinets in a London colocation facility and had decided to retire the infrastructure. The hardware was not being scrapped — it was going back to their own warehouse for reuse. That meant everything removed had to arrive somewhere else in a known condition and be reconcilable on receipt.

Their technical team is fully remote and based in another country. The client had no staff at the facility and nobody available to travel to it.

Every part of a decommissioning is physical. Servers have to be unracked, boxed against a manifest, moved to a loading bay and handed to a haulier on a booked slot. Cross connects have to be cancelled. Cables and anything else not going back have to be cleared. Cabinets have to be emptied and formally handed back so charges stop.

Like many inner-London sites, the facility occupies a converted building. Circulation space is tight, there is no spare floor area to work in, and the loading bay holds five half-Euro pallets for a maximum of two days. The bay is shared with every other customer in the building, so it cannot be occupied for the length of a project.

Three hundred servers need twenty shipping boxes. Flat-packed they take almost no room. Assembled they take more floor space than was available anywhere on site.

The default approach would have been to move a few servers whenever bay space came free. That runs for weeks, most of it spent waiting rather than working, while colocation charges continue on a footprint that is no longer doing anything useful.

The client defined the outcome they needed and retained control of the data wiping decisions. Everything else was ours. We planned the site operation, coordinated the colocation provider and the haulier, managed the removal day by day and produced the evidence their warehouse needed to reconcile the shipment.

The plan came from us. The client did not issue instructions on how to run the removal — their job was to approve what we proposed and confirm which cabinets were ready.

Solved the space problem first. Our team leader asked the colocation operator whether a secure but unused room could be made available for the empty boxes. They agreed. That one conversation turned an unworkable job into a schedulable one, because twenty assembled boxes could be built in advance and drawn on as needed. Boxes containing servers were never left unattended, there or anywhere else.

Built the plan around the bay limit rather than around the hardware. With five pallets as the ceiling, the removal was split into four batches. We wrote a document setting out which servers went into which box, which boxes made up each batch, and the order in which cabinets were stripped. Collections were booked with the haulier in advance against that document rather than called off on the day.

Booked the bay around other customers, not against them. The same team leader worked with the operator to reserve loading bay slots that did not collide with other customers’ projects. A clash on the day would have cost a full collection and a full day.

Ran every stage in parallel rather than one after another. The haulier collected in the morning and the next batch was filled and staged during the afternoon. Nothing ran in single file. While one pair of engineers stripped a cabinet, others were wiping failed machines, cancelling cross connects and preparing boxes for the batch after next. Engineer numbers were worked out in advance rather than decided on the day. Each stage of the cycle had a headcount set in the plan, and the team scaled up and down to match it. The figures held for the whole project.

Stayed in real-time contact throughout. Our engineers worked in a shared Slack channel with the client’s team rather than trading emails. Questions were answered while the engineer was still standing at the cabinet, and anything needing a decision was raised and resolved the same day. With the client’s team several time zones away, that was the difference between a query costing ten minutes and a query costing a working day.

Handled everything around the servers. We audited and cancelled the existing cross connects, arranged recycling collections for cables and any other equipment that was not going back to the warehouse, and worked directly with the colocation engineers on the cabinet handover.

Each batch ran against a written checklist, so the engineer on shift never had to reconstruct the plan from memory. Every server was recorded by serial number against the box it went into, with the date it was collected. The client received that record in full, and their warehouse was able to receive the shipment and reconcile it without a single query coming back to us.

The client’s remote team owned the data side. They wiped the servers, confirmed which machines completed and which did not, and issued the instructions for handling the failures. Every cabinet held two or three that had not completed, so across ten cabinets that was twenty to thirty machines needing a physical wipe. Our engineers carried those out on site from a bootable Linux USB, following the client’s instructions, and reported each result back. No machine was boxed until its status was confirmed.

That division worked because it played to both sides. The client kept control of a decision that was theirs to make. We supplied the hands and the record — which is the part they could not do from another country.

The site was cleared in eight working days. The pace was set by the loading bay rather than by the work. Four collections was the minimum the building allowed, and the project finished at that floor. Every collection went out on its booked slot. No lorry was rebooked, none left part-loaded, and none clashed with another customer.

No box contents differed from the manifest, so there was no reconciliation work at the receiving end. The twenty to thirty failed wipes were absorbed by running that work alongside the removal rather than after it, so unplanned effort did not become unplanned delay. Nothing waited overnight for an answer.

The cabinets were emptied and handed back to the colocation provider, which ended the client’s charges on that footprint. Against the approach the building would otherwise have forced, the client paid for working days instead of waiting days.

Decommissioning, with project management, logistics coordination and on-site data handling run by DACPROS.

If you own hardware in a UK colocation facility and need it removed properly, we’ll review the site and give you a plan before you commit to anything.

We’ll review the site and give you a removal plan before you commit to anything.