Border router upgrade and relocation — thirty fibre connections moved inside a four hour window

A border router had to be replaced. At the same time it had to move to a different position in the rack, which meant its 30 fibre connections had to come off the old unit and go back onto the new one in a new place.

This is a critical piece of infrastructure. With the border router out, the system loses a significant amount of bandwidth, so the only question that mattered was how short the window could be made.

Replacing a router in place is one job. Moving it to a different rack position at the same time is another, and it is the second one that creates the real work.

Cabling that reached the old position does not necessarily reach the new one. Every connection has to be checked for length before the window is booked, not discovered on the day. Anything unlabelled has to be traced to find out where it actually goes. Anything too short has to be relaid in advance and recorded, and the runs it replaces have to come out afterwards.

Thirty fibre connections have to come out and go back, each to a specific port. A single one of them going back wrong has to be found and traced while the window is open and the clock is running. So does a single cross connect from another cabinet that turns out to be a few centimetres short. One cable of the thirty is enough to put a four hour window out of reach.

That is where this kind of work usually overruns. Something is discovered on the day that nobody had counted on, and the window either stretches or the change gets rolled back and rescheduled.

DACPROS planned the window. The client agreed the plan and approved the change. Everything else was ours, including the window itself.

The four hours was our estimate, put forward with enough margin to absorb a surprise. None of the sequence is invented job by job. It comes from years of doing this work and is written down as our playbook.

Audited every connection against its new position. Moving the router meant a full cable audit rather than a tick-list. All 30 connections were accounted for and each one checked for length against where it was going, not where it had been. Anything without a label was traced to establish where it ran and then labelled.

Relaid in advance anything that would not reach. Connections that were too short for the new position were run again ahead of the window and recorded as they were done, including a cross connect coming in from a different cabinet. That is exactly the item that ends up breaking a window when it is found on the day.

Agreed the plan with the client and cleared the questions in advance. The method was discussed and anything ambiguous was resolved before the window was booked. Questions asked in advance cost minutes. The same questions asked with the router out cost the window.

Did everything else that did not need the suspension window. The work was separated into what genuinely required the router to be down and what did not, and the second group was completed beforehand. By the time the window opened, only work that could not have been done earlier was left.

Had the new router configured before the window, not during it. The replacement arrived brand new and unconfigured, shipped straight to the data centre. Normally a unit like that goes to the client’s technical team first, is configured there and then sent on to site, which is two deliveries and the waiting that goes with them. Instead we powered it temporarily outside the cabinet during the preparation phase and gave the client’s network team console access through our engineer’s laptop, so they could configure it remotely and confirm it was ready. The window was never going to be spent finding out whether the new unit was good.

Executed the window as a checklist. On the day the job was a list to be worked through rather than a problem to be solved. Nothing came up that had not been accounted for, and no troubleshooting was needed once the connections were in.

Cleared up and returned the old hardware afterwards. The runs that had been replaced were removed, and the router that came out was packed properly and returned to the client under RMA, so the job closed with nothing left in the cage and the old unit accounted for.

No missing connections, no rollback, no rescheduling.

The four hours were our own figure, proposed as a safety net rather than a target, and the work finished ahead of it. The preparation is what bought that. Every cable had already been proved long enough, every unlabelled run had already been traced, and the connections that would not reach had already been replaced. What was left on the day was thirty fibres going back to thirty known ports.

Shipping the router straight to site also took two delivery legs out of the project.

Smart hands and on-site hardware replacement in Slough, covering the connection audit, labelling, pre-window cabling, the change itself and RMA management for the unit that came out, run by DACPROS.

If you have a change to make on infrastructure that cannot be down for long, we will audit it and plan the sequence before the window is booked.