Computer latency for humans.

If I tell you that each core of a server, based on the Intel Xeon processor E5 v4 family, might perform 2.4 billion clock ticks and might execute many times that number of instructions in the second or so that it takes to read and register this post, would you be able to grasp the sense of this speed ? I am not.

So let’s try to scale this unimaginable speed to something that we, the humans, can perceive and understand.

In the computer chain, cpu is the fastest element, followed by cache, then memory and so on. But how how really fast, or slow, is every item of the chain ?

Let’s assume that one cpu cycle is equal to 1 second of our lives. Then look at the attached table.

System EventActual Latency —————-> Scaled Latency
One CPU cycle 0.4 ns ————————–> 1 s
Level 1 cache access 0.9 ns ——————>2 s
Level 2 cache access 2.8 ns ——————> 7 s
Level 3 cache access 28 ns ——————-> 1 min
Main memory access (DDR DIMM) ~100 ns ——> 4 min
Intel Optane memory access <10 ?s —————-> 7 hrs
NVMe SSD I/O ~25 ?s ————————–> 17 hrs
SSD I/O 50-150 ?s ——————————-> 1.5-4 days
Rotational disk I/O 1-10 ms —————–> 1-9 months
Web page call: San Francisco to New York City 65 ms –> 5 years
Web page call: San Francisco to Hong Kong 141 ms —-> 11 years

Credits: David Jeppesen from Prowess Consulting, who cites the work of performance architect and author Brendan Gregg, “Systems Performance: Enterprise and the Cloud”.