Primary Reference

• Overview

SLI (Service Level Indicator)

• Indicator of Availability of your application/service
• Sample Indicators:
  • Latency of the home page over last 5 minutes will be less than 300ms for 99.9% of requests

SLO (Service Level Objective)

• SLI binds over a period of time
• Sample:
  • Latency of homepage over a year will be less than 300 ms for 99.9% of the requests

SLA (Service Level Agreements)

• Sample:
  • Customer will be offered free credits if 99.5% of the requests over a year fail to achieved the latency of less than 300 ms

Problem

• We build systems and they fail at some point.
• What’s the SRE approach towards failures

Risks

• You can make aggressive deployments as long as you are with in Error Budget.
• If Error Budget is exceeded no more deployments

Error Budget

• Allowed time in minutes or hours of failure.
• Sample:

  • SLO : Latency will be less than 300 ms for 99.9% of request over the year

  • ERROR budget is what is left of total time after removing SLO (100-99.9) * 365 * 24 * 60/ 100 = 525.6 minutes/year

  • SLO : Latency will be less than 300 ms for 99.99% of request over the year

  • ERROR Budget (100-99.99) * 365 * 24 * 60 /100 = 52.5 minutes/year

Error Budget Burndown

• Error Budget used
• Fast Burndowns
• Slow Burndowns

Toil

• Repetitive manual work that can be automated
• Focus on Toils which are more frequent than infrequent ones