DataTerminology.comNEW
Also known as: SLI
Quantitative measures that indicate the real-world performance of a service against user-centric aspects like availability, latency, and error rate.
DESCRIPTION
Service Level Indicators (SLIs) quantify service performance from the user’s perspective. They underpin Service Level Objectives (SLOs) and Service Level Agreements (SLAs). Common SLIs include request latency, availability, error rate, and freshness. Well-defined SLIs are precisely measured, documented, and aligned to user journeys. They guide error budgets and reliability trade-offs. Strong observability and high-quality telemetry are prerequisites to trustworthy SLIs.
ALTERNATIVE DEFINITIONS
Google SRE Book
An SLI is a carefully defined quantitative measure of some aspect of the level of service that is provided.
Emphasizes careful definition and quantitative measurement tied to service level.
Atlassian
Service level indicators (SLIs) are metrics that measure the performance of a service.
Highlights SLIs as metrics for service performance distinct from goals and contracts.
Google Cloud Architecture
SLIs are the metrics over time that you set to measure the performance of your service as experienced by users.
Centers SLIs on user-experienced performance over time.
RELATED TERMS
Error Budget Availability Delays Reliability Observability Monitoring Uptime Site Reliability Engineering Throughput Reliability Engineering
COMPETING TERMS
KNOWLEDGE AREAS
Observability & Telemetry Software Engineering Platform & Tooling MLOps Data Services Data Engineering
TARGET AUDIENCE
DevOps Engineer MLOps Engineer Software Developer Cloud Architect Data Engineer Chief Technology Officer IT Manager
EXTERNAL SOURCES
- Wikipedia - Service-level agreement
- Google SRE Workbook - Practical SLI/SLO implementation guidance
- Microsoft Azure WAF - Reliability pillar with SLI/SLO guidance
- Google SRE Book - Monitoring to support SLIs
TOP QUESTIONS
How do Service Level Indicators relate to SLOs?
SLIs are measurements; SLOs are targets for those measurements. SLIs feed error budgets and reliability decisions.
Category: Relationship
What are the most common SLIs?
Availability, latency, error rate, throughput, and freshness are standard. Choose those that reflect user experience.
Category: Overview
What mistakes undermine SLIs?
Using vanity metrics, ambiguous definitions, or unrepresentative percentiles makes SLIs misleading and unactionable.
Category: Mistakes
How should SLIs be measured?
Collect high-quality telemetry at user-facing boundaries, define precise windows and percentiles, and validate via synthetic and real-user data.
Category: Measurement
What is a good starting point for SLIs?
Pick one critical user journey and define availability and latency thresholds; iterate with stakeholder feedback.
Category: Getting Started
How often should SLIs be reviewed?
Regularly, typically quarterly, or when product changes alter user journeys, data quality, or traffic patterns.
Category: Management
What risks arise from poorly chosen SLIs?
They may incent the wrong trade-offs, hide real user pain, and erode trust between teams.
Category: Risk
How do SLIs apply to ML systems?
Include data freshness, feature availability, inference latency, and prediction error rates to reflect end-user experience.
Category: Example
Can SLIs be standardized across services?
Yes, adopt a common SLI taxonomy to ensure comparability while allowing domain-specific metrics when necessary.
Category: Best practices