Service Level Agreement (SLA)
A service level agreement (SLA) is the contractual definition of measurable service quality: availability, service hours, response time and restoration time, each with a measurement method, measurement window and consequence for a miss. The figures are arithmetic, not rhetoric: 99.9 percent monthly availability permits roughly 43 minutes of downtime in a 30-day month, 99.99 percent only about 4.3 minutes. Distinctions: an SLA governs the customer relationship, an operational level agreement governs internal teams, and an underpinning contract governs upstream suppliers.
Service Level Agreement (SLA) in practice
The most common misreading concerns the time metrics. Response time states when work begins, not when the fault is fixed; that is restoration time. Both usually run only inside the agreed service hours, so a report on Friday evening against Monday-to-Friday service hours starts its clock on Monday morning. Priority is set by an impact-and-urgency matrix, and only that classification determines which commitment applies at all.
Availability commitments also depend on architecture rather than contract wording. Microsoft guarantees 99.99 percent for Azure VMs only where two or more instances are spread across two availability zones, 99.95 percent for two instances in one availability set, and 99.9 percent for a single VM using Premium SSD or Ultra Disk; Standard SSD drops that to 99.5 percent and Standard HDD to 95 percent.
Where a provider falls short, service credits are the market norm: for VMs, Azure grants 10 percent of the monthly fee below the committed level, 25 percent below 99 percent and 100 percent below 95 percent. These are the platform operator's figures and do not substitute for the commitment of the provider building on top of them. A sound SLA therefore also names its exclusions: planned maintenance windows, customer-initiated changes, upstream provider outages and force majeure.