Every cabinet is metered at the busbar, billed to a customer, and on a plan that has to match the fire certificate. J14 draws between 1.9 and 2.2 kilowatts, belongs to nobody, and the aisle it is in goes J13 to J15.
I am a night shift data centre technician. Hands and eyes: I swap drives, reseat cards, walk the floor, and answer the phone to customers who cannot get to their own kit at three in the morning.
Power is the whole business so here is how it is counted.
The room is fed from the main switchgear to power distribution units, and from the PDUs to overhead busbars running the length of each aisle. A cabinet hangs off the busbar through a tap box, and the tap box is metered. Two of them, actually, because everything is dual fed from A and B.
Each meter is read by the building management system every minute. That reading is the customer's bill. It is also how we know whether an aisle is approaching its design limit, which matters, because an aisle that trips takes out forty cabinets.
So every cabinet has: a position on the floor plan, a meter, a customer, a contract, and a line on the fire certificate, because the fire strategy depends on the heat load.
Those five have to agree. They are audited annually and I have sat in on the audit twice.
The building management system lists a meter called J14.
It draws between one point nine and two point two kilowatts. It has done since before I started, which was 2020, and the historian goes back to the commissioning in 2017 and it is there from the first week.
It is a completely ordinary load. About what a half populated cabinet of older servers draws. It varies a little through the day, more in the afternoon, less at four in the morning, which is what a real load does when real people are using real machines.
Aisle J goes J01 to J24 and there is no J14.
Not an empty floor tile. The cabinets are bolted to the floor and they touch. J13 is beside J15. I have put my hand on both at once.
The floor plan has twenty three cabinets in aisle J, numbered J01 to J24 with 14 missing, and it has been like that since the drawings were issued.
I asked about it in my first month, the way a new technician asks about everything, and the shift lead at the time told me there had been a numbering error at commissioning and 14 was skipped. Which is an extremely ordinary thing to happen and which I believed for two years.
What stopped me believing it.
The billing.
Every metered load is attributed to a customer account. The finance system pulls the meters monthly and raises invoices.
J14's consumption is attributed. It has a customer account number.
The account is not in the customer database. I know that because in 2022 I was asked to do a reconciliation of cabinets to accounts ahead of a migration, which is the sort of job they give the night shift because nights are quiet, and I went through all nine hundred and forty cabinets in the building.
Nine hundred and thirty nine matched. J14's account number is the right format and has no record behind it.
And the invoices go out. About two thousand one hundred pounds a month of power on that account, raised, and I do not have access to the ledger so I do not know whether it is paid, and the one time I asked the day shift manager whether a particular account was in arrears he quite properly told me that was not my business.
The fire certificate.
This is the one that means somebody with a professional qualification has signed something.
The heat load calculation for the room is the sum of the cabinet loads, and the certificate is based on it. I have seen the calculation, because it is on the wall in the plant room in a frame.
The total on that sheet is about two kilowatts higher than the sum of the cabinets on the floor plan.
I have added the floor plan up three times on three different nights. The difference is J14.
So either the fire engineer added a cabinet that is not on the plan, or the plan is missing one, and the certificate is the document that would be in front of a coroner.
What I have done about it, which is very little, and the reason is honest and unflattering.
I raised a ticket in 2022. Category: BMS data quality. It was closed after six weeks with the resolution that the meter point was a legacy configuration item and had been flagged for removal in the next BMS upgrade.
The BMS was upgraded in 2024. New head end, new database, points re-mapped from the field devices.
J14 came across. It is drawing two point one kilowatts as I type this, on a screen eleven feet from where I am sitting.
If I raise it again I am the night tech who will not let go of a meter point, and I have eleven months left on a contract I want renewed.
The last thing, and it is the thing that made me write this down instead of forgetting it again.
A tap box is a physical object on a busbar. If a meter reads two kilowatts, something is connected to a tap box, and a tap box is bolted into the busbar overhead where you can see it.
In the August I took a ladder into aisle J at about half two in the morning, which is allowed, and I walked the busbar from J01 to J24 with a torch, reading the labels on the tap boxes.
There are twenty four tap boxes.
They are evenly spaced, one above every cabinet position, and the gap between J13 and J15 does not exist up there, because up there the spacing is the spacing of the busbar and it does not care what is underneath it.
The fourteenth one is labelled J14 and it is live and its breaker is closed, and there is a pair of power cables coming out of the bottom of it, going down, and they stop about nine inches above the top of the gap between two cabinets that are touching.
Not cut. The ends are terminated in a plug. The plug is the right plug for a cabinet inlet and it is hanging in the air, and the cable is warm.