Resources

Four numbers carry the night. Three of them live inside a single shift; the fourth is the only thing that survives it.

Budget

You start a shift with a fixed allocation — $500 on night one, rising by $50 per campaign night to $950 on night ten. Hard difficulty cuts it to 75%.

Only some interventions cost money, and the ones that do are permanent fixes rather than stopgaps. Money left over at sunrise converts into score at half a point per dollar, so an unspent budget is not wasted — but neither is it worth protecting past the point where the fleet needs it.

Everything else is free. That asymmetry is deliberate: the cheap actions are reversible, the expensive ones are structural.

Bandwidth

Incoming traffic in requests per second, measured against the fleet's total capacity. Only web (1500 req/s each) and lb (3000 req/s each) contribute capacity. Go over and latency climbs; stay well over and machines start saturating.

Because capacity is a property of the fleet, it also grows when you do — which is why adding machines is a real, if expensive, answer to a bandwidth problem.

Uptime

A composite measure of service quality, expressed 0–100 and averaged across the whole shift. It falls when machines are down, when latency is above target, and separately when an incident sits unresolved past its grace window.

Uptime is the lose condition below 50% and the largest single term in the score. Practically everything you do is an uptime decision wearing a costume.

Reputation

The only meta-resource — it lives between nights rather than inside one, and only in the campaign.

A strong run compounds: better grades buy more budget, which buys more room to absorb the next night. A bad night bites twice.