Error Rate and Success Metrics
You'll learn how error rates, timeouts, and retries function as vital performance indicators, not just correctness checks.
Beyond Pass or Fail
When we think about errors, we often frame them as binary: the system works or it doesn't. But in performance engineering, how often your system fails and how it handles failures are critical performance dimensions. A system with a 0.1% error rate has very different performance characteristics than one with 5% errors, even if both are "mostly working."
Error Rate as a Performance Metric
Error rate measures the percentage of requests that fail. If you're serving 1,000 requests per second with 10 failures, that's a 1% error rate. This tells you:
- Whether your system is degrading under load
- If recent deployments introduced instability
- How close you are to violating your error budget (from your SLOs)
Track errors alongside latency and throughput—they're interconnected. High error rates often signal resource exhaustion, cascading failures, or configuration issues that impact overall performance.
Timeout and Retry Behavior
Timeout rates reveal how often operations exceed acceptable wait times. A 2% timeout rate means users are hitting dead ends frequently—a performance problem even if the underlying service eventually succeeds.
Retry patterns affect both success rates and load. If 10% of requests retry twice, you're actually processing 1.2× the apparent traffic. Aggressive retries can amplify problems during incidents, creating retry storms that worsen performance.
Success Rate: The Flip Side
Track success rate (100% - error rate) as a positive indicator. Pair it with request volume to understand total impact: serving 10,000 successful requests at 99% success is better than 1,000 at 99.9%.
Key Takeaway: Error rates, timeouts, and retry behavior are first-class performance metrics that reveal system health under real load conditions—track them continuously alongside latency and throughput to understand true system behavior.