Jose R. Ziviani jrziviani

Latency Comparison Numbers

Note: "Forked" from Latency Numbers Every Programmer Should Know

Event	Nanoseconds	Microseconds	Milliseconds	Comparison
L1 cache reference	0.5	-	-	-
Branch mispredict	5.0	-	-	-
L2 cache reference	7.0	-	-	14x L1 cache
Mutex lock/unlock	25.0	-	-	-

Tracing QEMU-KVM Interactions

But default in linux you can figure out how many times and for what reasons there is a VM Exit from a VM into the kvm kernel module. However given the ubiquity of vhost and the ability of kvm to emulate most device models directly in the kernel, most of those VM exits do not result in a transition from host kernel into the QEMU. The transitions from VM -> kvm -> QEMU are typically the most expensive.

Here we try to figure out how many of the VM Exits result in the invocation of QEMU.

Tracking VM-KVM Interactions

This can be done very simply with perf

	#!/bin/bash

	# load resources
	xrdb -merge .Xresources

	#xsetroot -solid '#222'&
	# map caps lock as extra escape
	xmodmap -e 'clear Lock' -e 'keycode 0x42 = Escape'&
	# start clipboard manager
	parcellite&

	Latency Comparison Numbers
	--------------------------
	L1 cache reference 0.5 ns
	Branch mispredict 5 ns
	L2 cache reference 7 ns 14x L1 cache
	Mutex lock/unlock 25 ns
	Main memory reference 100 ns 20x L2 cache, 200x L1 cache
	Compress 1K bytes with Zippy 3,000 ns 3 us
	Send 1K bytes over 1 Gbps network 10,000 ns 10 us
	Read 4K randomly from SSD* 150,000 ns 150 us ~1GB/sec SSD