Tail Latency

Definitions

1. The response time experienced by the slowest requests — typically measured at p99 (99th percentile) or p99.9. In distributed systems, tail latency compounds: if a request fans out to 100 services, the overall latency is determined by the slowest one. Jeff Dean's famous insight is that tail latency matters more than average latency.

   In plain English: How slow the slowest 1% of requests are, which matters a lot more than the average speed.

   Example: "Our p50 is 20ms but p99 is 2 seconds — those tail latency outliers are caused by garbage collection pauses."

Related Terms

• toil
• tail latency
• Thundering Herd
• Allocator
• Big-O
• Cache
• Comptime
• Concurrency
• Golden Signals