Service Level

• Define the metrics that matter
• Define the expected range for those metrics (what a healthy service looks like)
• Define the action to take when we can't provide the expected service

Service Level Agreement (SLA)

• Days to accomplish something

Service Level Indicators (SLI)

• Quantitative measure (metrics!) of some aspect
• Data is collected over a measurement window and turned into a rate, average, percentile
• E.g.
• request latency
• error rate
• system throughput
• availability/yield: well formed requested that succeeded. (99.95%: 3 nines and a half)

Service Level Objectives (SLO)

• Numeric thresholds to define acceptable levels for the SLIs
• E.g.,
• p99 latency < 200ms for 5 min