SLI / SLO / SLA

• SLI: quantitative metric (uptime %, p99 latency, error rate)
• SLO: target value for the SLI — internal goal
• SLA: customer-facing contract, usually with penalty on breach
• Error budget: 100% - SLO (e.g. 99.9% SLO = 0.1% budget = 43 min/month)
• Multi-dimensional: separate SLOs for availability, latency, error-rate

SLI: % requests with status 2xx\/3xx
SLO: 99.9% of requests succeed (monthly)
SLA: refund 10% if <99.9% in a month

• p99 latency
• Rate Limiting

A Service Level Indicator (SLI) is a quantitative measure used to gauge the performance of a service. SLIs are essential for understanding how well a service is functioning in relation to user expectations. Common SLIs include metrics such as response time, availability, error rate, and throughput.

For instance, if you are measuring response time, you might define the SLI as the p99 response time, which indicates the response time at the 99th percentile. This means that 99% of requests are served within this time frame, providing a clear indicator of service performance under normal operating conditions.

Another example could be the error rate, defined as the percentage of requests that result in an error. If your application processes 1000 requests and 5 result in errors, your error rate SLI would be 0.5%.

SLIs should be carefully chosen to align with user expectations and business goals. They provide the foundation for setting Service Level Objectives (SLOs) and Service Level Agreements (SLAs). By monitoring SLIs consistently, organizations can proactively identify performance issues and ensure services meet user demands.

A Service Level Objective (SLO) is a specific target that an organization aims to achieve for a given SLI. SLOs provide a clear framework for measuring service reliability and performance. When crafting SLOs, it is crucial to consider the business impact and user expectations.

For example, if your SLI indicates that your service has a p99 response time of 300ms, you might set an SLO that states: p99 response time must be less than 200ms. This target establishes a clear goal for the engineering team to work towards.

Another common example is setting an availability SLO. If you define your availability SLI as uptime percentage, you might set an SLO of 99.9% uptime monthly. This translates to an allowable downtime of approximately 43 minutes per month. Monitoring these SLOs regularly helps organizations assess if they are on track to meet their performance goals.

Setting realistic SLOs requires a deep understanding of both the service's capabilities and the end-users' needs. Regular reviews and adjustments of SLOs can ensure that they remain relevant and achievable, thus driving continuous improvement in service delivery.

A Service Level Agreement (SLA) is a formal, legally binding contract between a service provider and its customers that outlines the expected level of service. SLAs typically include performance metrics such as uptime, response times, and support availability, along with the consequences for failing to meet these standards.

For instance, an SLA might stipulate that the service provider guarantees 99.9% uptime per month. If the service provider fails to meet this commitment, the SLA may specify that customers are entitled to a refund or credit. This establishes a clear expectation for service reliability and accountability.

SLAs also often include provisions for maintenance windows, response times for support requests, and escalation procedures for unresolved issues. For example, an SLA might state that critical incidents will receive a response within 1 hour and resolution within 4 hours.

It is essential for organizations to regularly review and revise SLAs to ensure they align with business objectives and customer needs. By doing so, companies can foster trust and transparency with their customers, ultimately leading to improved service satisfaction and loyalty.