Things on this page are fragmentary and immature notes/thoughts of the author. Please read with your own judgement!
| Latency |
Operations |
| <=1ns |
- accessing CPU registers
- CPU clock cycle
- L1 cache
|
| 1-10ns |
- L2 cache
- branch mispredict
|
| 10-100ns |
- L3 cache
- mutex lock/unlock
- main memory referencing
|
| 100-1000ns |
- trapping of system call
- md5 hash of an u64
|
| 1-10us |
- process context switching
- compress 1k with zippy
- memory copy of 64k
|
| 10-100us |
- network proxy / http request
- read 1M from memory
- SSD latency (read a 8k page)
|
| 100-1000us |
- SSD write latency / write a page
- intra-zone networking round trip
- memcache/redis get operation
(measured by client including network round trip)
|
| 1-10ms |
- inter-zone networking latency
- hard drive latency
|
| 10-100ms |
- network round trip between US coasts
- network round trip between US east coast and Europe
- read 1G from memory
|
| 100-1000ms |
- bcypt a password
- TLS handshae (250-500ms)
- network round trip from US west coast to Singapore
- read 1G from SSD
|
| >=1s |
- transfer 1G over a network in the same region
|
References