Course contentsShow
← Cloud and Infrastructure
Lesson 2114 of 2,98945. Observability - Dashboards and SLOsFree lesson

Common SLI Categories: Availability, Latency, and Throughput

Explore the most common types of SLIs used to measure system reliability and performance.

Common SLI Categories: Availability, Latency, and Throughput

What you'll learn: The three foundational SLI types that form the backbone of reliability measurement in production systems.

Understanding the Big Three

While you can measure countless aspects of a system, three categories dominate reliability engineering: availability, latency, and throughput. These aren't arbitrary choices—they directly map to what users care about most: "Is it working?", "Is it fast enough?", and "Can it handle my load?"

Availability SLIs

Availability measures the proportion of valid requests that succeed. Think of it as uptime from the user's perspective.

Common availability SLIs:

  • Percentage of HTTP requests returning non-5xx responses
  • Percentage of successful health checks
  • Ratio of successful database queries to total attempts

Analogy: A store's "open" sign—users care whether they can get in, not whether the lights are on in the back office.

Latency SLIs

Latency tracks how quickly your system responds. This is typically measured as percentiles (p50, p95, p99) rather than averages, because outliers matter enormously to user experience.

Common latency SLIs:

  • 95th percentile API response time under 200ms
  • 99th percentile page load time under 1 second
  • Median database query latency under 50ms

Analogy: Restaurant service speed—most diners tolerate some wait, but nobody wants to be the unlucky table waiting 45 minutes.

Throughput SLIs

Throughput measures how much work your system successfully processes over time—requests per second, transactions per minute, or messages processed per hour.

Common throughput SLIs:

  • Successfully processed requests per second
  • Completed batch jobs per hour
  • Messages consumed from queue per minute

Key Takeaway: Availability tells you if your system works, latency tells you if it's fast enough, and throughput tells you if it can handle the volume—together, these three SLI categories capture what users experience as "reliability."