SLA

An SLA turns vague promises into numbers someone can be held to. It's the contractual layer that says, in writing, what minimum performance the customer will get — typically availability (uptime, often quoted as a number of nines), throughput, latency, and sometimes things like fault-response and restoration times. It also has to define how performance is measured and what happens when it's missed: service credits or penalties are the usual teeth.

This is where KPIs stop being an internal dashboard and become money. The NOC tracks the live numbers against the committed targets, and MTTR (how fast faults get fixed) often sits right in the contract because restoration speed is part of what the customer is paying for. SLAs come up most sharply in enterprise and private-network deals, where a customer running a factory or a hospital on the network cares far more about a guaranteed availability figure than about a peak speed. The detail that bites is the measurement clause: "99.9% availability" is almost meaningless until the SLA pins down how, where and over what window it's measured, and what counts as an excused outage.