Latency Numbers Every Engineer Should Know

Interactive chart from L1 cache to cross-continent network calls.

Latency numbers every engineer should know. Hover over a bar for details.

L1 cache reference

1 ns

Branch mispredict

3 ns

L2 cache reference

4 ns

Mutex lock/unlock

17 ns

Main memory reference

100 ns

Compress 1KB (Snappy)

2.0 μs

Read 1 MB sequentially (memory)

3.0 μs

SSD random read

16 μs

Send 2KB over network

44 μs

Read 1 MB sequentially (SSD)

49 μs

Round trip same datacenter

500 μs

Read 1 MB sequentially (disk)

825 μs

Disk seek

2.0 ms

Send packet CA → NL → CA

150 ms

Memory is ~200x faster than SSD for random reads

SSD is ~100x faster than spinning disk for random reads

Same-datacenter round trip is ~300x faster than cross-continent

Compressing before sending saves time if payload > ~1 KB