Service Level Objective
An SLA must contain one or more thresholds that defines the “service level” that is implied by the name. This is called the Service Level Objective (SLO). Likely the most common SLO in the world of IT measures Availability and is often expressed as a percentage (99.99%), or in a number of “nines”, e.g. “four nines” (99.99%) or “five nines” (99.999%).
Whatever measurement you are providing as the SLO, it should be black and white: your service either achieves it, or it doesn’t. There should not be room for subjectivity or interpretation.
How many SLAs do you need?
Figure 2: The 3 types of SLA in VM as a Service
Many service providers will only commit to the most obvious one: the Availability SLA. This may seem the simplest SLA to track and manage because it’s binary: the service is either up, or it’s down. In reality, Availability can be deceptively complicated depending on where and how you need to measure it, but this will be discussed more in section later in this chapter about Availability SLAs.
But just because something is “up” doesn’t mean it’s “fast”, and a service that’s slow to the point of being unusable may as well be down.