Defining reliability expectations via SLOs
Service consumers (users) need a service to be reliable, and the reliability of the service can be captured by multiple characteristics such as availability, latency, freshness, throughput, coverage, and so on. From a user's perspective, a service is reliable if it meets their expectations. A critical goal of SRE is to measure everything in a quantitative manner. So, to measure, there is a need to represent user expectations quantitatively.
SRE recommends a specific technical practice called a SLO to specify a target level (numerical) to represent these expectations. Each service consumer can have a different expectation. These expectations should be measurable, and for that they should be quantifiable over a period. SLOs help to define a consistent level of user expectations where the measured user expectation should be either within the target level or should be within a range of values. In addition, SLOs are referred to as internal...
