All the datacenter jargon you'll meet in the game, in plain words. 🤓
| Term | Meaning |
|---|---|
| Autoscaling | Extra instances added automatically to a customer's VMs or Kubernetes cluster when they are busy, and removed when the load drops. |
| Backbone | The cable tray linking the aisles of a server room. |
| Cable manager | A 1U accessory that routes the cables of nearby devices neatly along the rack. |
| Cold / hot aisle | The aisle in front of the racks (cold, where servers take in air) and behind them (hot, where they blow it out — and where the cables are). |
| COP | Coefficient of performance of a cooling unit: heat removed per unit of electricity drawn. Higher is better. |
| DAC | Direct-attach copper: a short, cheap, fast cable between two SFP or QSFP cages. Reaches only a few metres. |
| Degraded | A unit that runs, but badly (saturated server, slow or lossy network). Still billed, but counts as unavailable. |
| Disconnected | A unit that can't be reached (server off or failed, no network path). Not billed, counts as unavailable. |
| Execution unit | One instance of a service you sold (a VM, a database…) running on one of your servers. |
| Failover | Switching traffic to a backup path when the main one fails. |
| Fibre | An optical cable with its two optics, for long distances. Expensive. |
| L2 / L3 | Simple switching over a single path / routing over several paths at once with instant failover. |
| Load balancing | Spreading traffic over several links at the same time. |
| Management port | The socket that opens a device's management page instead of carrying data. |
| N+1 | Having one more unit (cooling, power…) than strictly needed, so that a single failure changes nothing. |
| Off-peak / peak hours | Night hours with cheap electricity / day hours with expensive electricity. |
| Operator uplink | The box in the operator rack that connects your building to the Internet. |
| Overcommitment | Selling more CPU or network than your servers physically have, betting that not everyone needs it at once. |
| Pending | A unit that found no server with enough room. Not billed, counts as unavailable. |
| Prospect | A potential customer, waiting for your answer on the Prospects page. |
| QSFP28 | A 100 Gb/s network cage, for network cores, routers and fast Internet uplinks. |
| Rack | The cabinet in which servers and network devices are mounted. |
| Reputation | Your standing on the market (0 to 100): it drives how many prospects show up, tier unlocks and renewals. |
| RJ45 | The classic copper network socket. Capped at 1 Gb/s in the game. |
| Saturation | Real demand above the capacity of a server or a link: everything on it suffers. |
| Shedding | Switching off whole racks when your power demand exceeds what your grid can deliver. |
| SFP+ / SFP28 | Network cages for 10 / 25 Gb/s links, taking DAC or fibre cables. |
| Single point of failure | A device or cable whose failure alone cuts services off, for lack of a backup path. |
| SLA | Service Level Agreement: the availability you promise a customer, and its measured value. |
| Spine | A very fast (100G) switch linking several network cores and routers together. |
| Tier | A customer's category — Starter, Pro, Business, Enterprise — setting its needs, its pay and its demands. |
| Transformer | Equipment in the electrical room that brings power to your racks. Their total sets (part of) your available power. |
| U | Rack unit: the height slot of a rack (4.45 cm). A 1U server takes one, a 2U server two. |
| Uplink | A switch's fast port used to connect it to the rest of the network. |
| vCPU | A virtual processor. A server offers two per physical core. |