Calculating SLA creditsGet a second set of eyes26

Calculating SLA credits [/ˈkælkjəˌleɪtɪŋ ˈsla ˈkrɛdɪts/] n - A service level agreement promises a customer some measure of uptime or performance, and sets a price upon missing it. When we miss, the customer is owed an SLA credit. To reckon that figure aright, neither too generous nor too stingy and ever defensible, is among the more exacting labors of IOC operations, for it stands where the monitoring record, the contract, and finance meet.

First, Measure Against the Proper Clock

The reckoning begins from measured downtime, and the measurement must answer to the contract’s own definition, which is seldom “ping failed.”

  • What counts as down? [/ˈwhæt ˈkæʊnts ˈæs ˈdɒwn/] n Full outage, or degradation past a performance floor? Per-rack, per-service, or the whole site?
  • When did it start and stop? [/ˈwhɛn ˈdɪd ˈɪt ˈstært ˈænd ˈstɒp/] n I anchor this to the monitoring timeline and the incident ticket, never to memory; see Data integrity and log retention for why the source data must be trustworthy.
  • What’s excluded? [/ˈwhæts ˈɛxklʌdɛd/] n Scheduled maintenance windows and customer-caused troubles are usually excepted, and that exception must be applied honestly.

Availability for a period is but measured uptime over scheduled uptime:

Second, Map Availability to Its Credit

The contract sets the tiers: below 99.99% earns back one percentage of the monthly fee, below 99.9% a greater, and so on. The labor here is looking up the right tier and applying it to the right billing base, which means reading the specific agreement rather than presuming a house standard. Two customers upon the same site oft keep different tables.

Third, Validate Before It Goes Abroad

A credit is a number that leaves the building and strikes a bill, so it endures the same scrutiny as an On Running the Root-Cause Inquiry:

  • Reconcile [/ˈrɛkɒnkɪlɛ/] n the downtime window against the incident record and the alert history; the start and stop times must agree.
  • Apply exclusions [/ˈæppli ˈɛxklʌsɪɒns/] n for maintenance and out-of-scope causes.
  • Sanity-check [/ˈsænɪtikhɛkk/] n the tier and the base against the contract language.
  • Get a second set of eyes [/ˈgɛt ˈæ ˈsɛkɒnd ˈsɛt ˈɒf ˈɛiɛs/] n upon anything large before it is processed.

Fourth, Process and Retain

Once validated, the credit is filed through the agreed finance and ticketing path, and the calculation is kept with its evidence [/ˈkept ˈwith ˈits ˈevidense/] n: the contract clause, the downtime math, and the source telemetry, so that if it is ever questioned, the answer is a folder and not a reconstruction. SLA-credit volume and value also feed Operational metrics and reporting, for a rising credit trend foretells that something upstream, most often detection or a recurring failure, wants attention.