A spares decision is a decision about where the delay sits. If the part is already on site, the repair takes as long as the swap. If it has to be ordered, the repair takes as long as the supply chain.
Three places to hold stock
Field stock sits at or near the site: a locked cabinet in the comms room, a shelf in the colocation cage. It gives the shortest time to repair and the least control. Parts get borrowed, moved and forgotten.
Central stock sits in one warehouse or depot and ships to site when needed. It is easier to count and cheaper to hold, because one pool covers many sites. The cost is transit time, plus customs where the estate crosses borders.
On-demand means holding nothing and sourcing from a distributor or support contract when something fails. It ties up no capital. It also means the repair waits on someone else’s stock position.
Most estates use all three. The work is deciding which component goes where.
Which components justify holding
A part earns a place in stock when it fails often enough, stops enough service when it does, and takes long enough to source. Components that usually qualify:
- Hard drives and SSDs, which fail regularly and are simple to swap.
- Power supplies and fans, which are field-replaceable and fail predictably.
- Optics and patch leads, which are cheap and often the real fault.
- Controllers or line cards for platforms out of warranty or near end of life.
Motherboards and chassis rarely justify field stock. They fail less often, cost more, and usually have a vendor replacement route.
Failure patterns and lead times
Group past tickets by part number and site and the pattern shows itself: one model of power supply that keeps failing, one site with poor cooling that keeps losing fans. Hold stock against what has failed, not a generic list.
Lead time changes the answer. A drive available next day locally may not need holding. A controller for a platform no longer in production may take weeks to find on the secondary market, and one spare on the shelf is worth more than a fast engineer. For hardware past its support dates, multi-vendor break-fix cover works only as well as the spares pool behind it.
The spares nobody reconciles
Stock that is never counted drifts away from the records. An engineer takes a drive from the cabinet, fits it, and puts the faulty unit back in the same slot, unlabelled. A year later the register shows three good spares and the shelf holds one good drive and two dead ones.
The cost appears at the worst moment: an engineer on site and a spare that does not work or is not there. Quieter costs follow. Parts bought for decommissioned platforms sit on the books. Stored spares fall behind on firmware and may need updating before the array or switch stack will accept them.
The fix is routine. Count field stock on a fixed schedule, tag every faulty part at the point of removal, and retire spares when the platform they support leaves the estate. Before signing a support contract that includes spares, ask who owns the stock, who counts it, and what happens to it when the contract ends.
