IT infrastructure migration to a new London data centre

The client is a major global digital service provider. Their UK infrastructure had to move to a new colocation data centre in south London: seven cabinets of servers, switches, routers and patch panels, around 250 servers in all.

The equipment was in active use. Every hour it spent in transit or waiting to be reconnected was an hour the service was affected, which is what makes a migration different from a build. There is no option to take your time.

The client had no local resource. Doing this with their own people would have meant flying staff over from the United States and keeping them here for the duration.

The window was five days of downtime. That was the maximum the business could absorb, and the whole project had to fit inside it.

Done in sequence, a migration of this size takes a long time. Decommission, pack, transport, unload, build the cabinets, install, cable, configure, test — each step waiting on the one before it and the service down for all of it. Five days is not generous for that. It is tight.

The way out is not to work faster on the day. It is to move as much of the work as possible to before the window opens.

The second problem is that nobody would be there to see it. The client’s engineers were on another continent, so whoever ran the project had to be trusted to plan it, sequence it and execute it without supervision.

The client set the requirement and the window, and supplied the rack layout they wanted. Everything else was ours: the surveys, the sequence, the pre-build, the transport, the move itself, the reconnection and the documentation.

From doing this work repeatedly we know which parts of a migration can be finished before the window opens, so we go through every task and pull as much as possible forward. On this project that led to a specific proposal: build the new network fully first, before a single server moved.

Surveyed both data centres well in advance. The old site and the new one, before any planning was finalised. The purpose was to find the risks that cause delays and, just as importantly, to identify work that could be completed before the equipment moved at all.

Proposed building the new network before the move, not during it. The plan to stand up the new network in advance came from us. It is the single decision that turned five days of downtime into three, and it was put forward at planning stage rather than discovered halfway through.

Prepared the cabinets to the client’s layout. The client supplied the layout they wanted and we worked to it. What needed planning was everything that hangs off it: where the rails and the cable management bars go, how the power cords are routed. Getting those right in advance is what makes an installation tidy and easy to maintain later.

Set up the new cabinets and brought the network up while the old site was still running. Rather than waiting for the old cabinets to be decommissioned, an advance team set up the new ones: rails and blanking plates fitted, switches and routers racked, and the fibre cross connections between cabinets laid in. With the hardware in place, the client’s network team configured and tested everything remotely, including the ISP connection, so the new network was live and proved before a single server moved. Those steps take a significant amount of time. Doing them in advance meant that on the day the change came down to rerouting the traffic.

Booked a secure, insured carrier for the transport. We arranged the transport ourselves rather than leaving the client to find someone, using a carrier with the right containers and insurance for equipment of this value.

Labelled and packed every server by destination. Each server was labelled with the cabinet slot it was going to and packed deliberately rather than loaded loose, with every box marked with the target cabinet name. Unloading then followed the build sequence, so the team installing at the new site was never hunting for the next machine.

Moved and reconnected. On the day, servers were decommissioned, securely packaged and transported. At the new site they were installed, power supplies connected, cabling run, labelled and patched in. Because the network was already up and tested, the work came down to disconnecting the servers at one end and connecting them at the other, with the client redirecting traffic to the new router.

Returned the old switches for reuse. The switches that came out of the old data centre were packed and returned under RMA so the client could put them back into service elsewhere.

Documented the installation in full. The client received a record of when and where every piece of equipment was installed, with every cabling interconnection labelled to our standard. The labels carry the information in plain form rather than in codes only we can read, so anyone who opens those cabinets later can work from them, including another provider.

Three days of downtime against the five the client expected, and the project delivered a day early and within budget.

“We were very impressed with the overall organisation of the project. Their communication with us was excellent, the attention to detail was remarkable.” — VP Global Network Infrastructure

The saving did not come from working faster. It came from the sequence: surveys done early, the layout settled before anything moved, and the new cabinets set up, the network configured and tested and the ISP connection live while the old site was still running. By the time the window opened, most of the project was already finished.

No member of the client’s team travelled from overseas.

The client was already ours. We had been working on their infrastructure before this, but not on anything like this: a migration on this scale was a new type of project for the relationship.

It went well, and more work followed. An existing supplier being given a bigger and unfamiliar job, and then further projects after it, is a different kind of decision from being hired once.

Migration and data centre relocation, with on-site surveys, project management, structured cabling, labelling and full installation documentation, run by DACPROS. Smart hands and ongoing maintenance run alongside it on the same infrastructure.

If you are moving infrastructure between data centres and your team is not local to either site, we will survey both and plan the sequence before a window is agreed.