SLA, SLO, and SLI Metrics

Understanding SLA, SLO, and SLI

Service Level Agreement (SLA)

• Definition: A contract between a service provider and a customer specifying the expected level of service.

• Purpose: Defines obligations and consequences if the agreed*upon reliability is not met.

• Example: "The service will be available 99.9% of the time per month. If this is not met, a refund of 10% will be issued."

Service Level Objective (SLO)

• Definition: A specific, measurable target for the level of service reliability.

• Purpose: Serves as a benchmark to ensure the SLA is met.

• Example: "99.95% of HTTP requests will return a response within 200ms."

• Relation to SLA: Typically stricter than the SLA, giving room to address failures before breaching the SLA.

Service Level Indicator (SLI)

• Definition: A metric that quantifies system performance to track compliance with SLOs.

• Purpose: Acts as the raw data used to measure whether an SLO is met.

• Example: "The percentage of successful HTTP requests over the past 30 days."

Example Relationships Between SLA, SLO, and SLI

• SLI: Measured uptime = 99.93%

• SLO: Target uptime = 99.95%

• SLA: Guaranteed uptime = 99.90%

Calculating Availability and Reliability

Availability is often expressed as a percentage of uptime over a given period:

$$\text{Availability (\%)} = \left( \frac{\text{Uptime}}{\text{Total Time}} \right) \times 100$$

Example:

• Total Time: 30 days (43,200 minutes)

• Downtime: 30 minutes

$$\text{Availability} = \left( \frac{43,200 * 30}{43,200} \right) \times 100 = 99.93\%$$

Reliability measures the likelihood of a system performing without failure over a specific time:

$$\text{Reliability (\%)} = e^{*\left(\frac{\text{Total Failures}}{\text{Total Time}}\right)}$$

Example:

• Failures: 2

• Time: 100 hours

$$\text{Reliability} = e^{*\left(\frac{2}{100}\right)} = e^{*0.02} \approx 98.02\%$$

Examples of Defining Metrics for a Service

Let’s define SLAs, SLOs, and SLIs for a simple web API

1. SLI Examples:

• Latency: 95% of requests complete within 200ms.

• Uptime: Percentage of time the service is reachable.

• Error Rate: Percentage of failed requests.

2. SLO Examples:

• Uptime SLO: "The service uptime will be at least 99.95% per month."

• Latency SLO: "95% of requests will complete within 200ms over the past week."

• Error Rate SLO: "The error rate will not exceed 0.1% of total requests over the past month."

3. SLA Example:

• SLA: "The service will maintain 99.9% uptime per month. For every 0.1% below this threshold, a 5% refund of the monthly fee will be issued."

Monitoring and Implementation

Monitoring SLIs:

• Latency: Use tools like Prometheus to track response time.

• Uptime: Use uptime monitoring tools like Pingdom or a custom Prometheus exporter.

• Error Rate: Count HTTP 4xx and 5xx responses using metrics.

Visualizing in Grafana:

• Create panels for each metric to display SLI performance over time.

• Set alerts when an SLO is violated (e.g., error rate exceeds 0.1%).