Latency Numbers Intuition¶

Memorizing order-of-magnitude latency differences (L1: 1ns, RAM: 100ns, SSD: 100us, network: 100ms) enables rapid identification of performance bottlenecks and explains why certain operations feel "slow."

Tension Addressed¶

Speed vs. Capacity

Recognizable As¶

"Why is this slow?" investigations, capacity planning, architecture decisions about what to cache, when to batch, where to colocate.

Canonical Sources¶

Dean & Barroso, "The Tail at Scale" (2013); Latency Numbers Every Programmer Should Know

Transfers To¶

Distributed systems design, database tuning, API design, user experience optimization

Key ratios to internalize: L1 cache reference (1ns) -> Main memory (100ns) = 100x. Main memory -> SSD random read (100us) = 1000x. SSD -> Network round trip (100ms) = 1000x. These ratios guide architectural decisions.