Defining Unix Load Average
November 7, 2005
Went looking for a simple definition of load average in Unix, but alas, simplicity is not in the nature of load. I read UNIX® Load Average Part 1: How It Works by Dr. Neil Gunther who breaks load down into much appreciated, yet excruciating detail.
In short it is the average sum of the number of processes waiting in the run-queue plus the number currently executing over 1, 5, and 15 minute time periods.
It’s calculated like this: load(t) = load(t - 1) e^(-5/60m) + n (1 - e^(-5/60m))
(at least in Linux)
From Gunther’s PDF Guide:
Most sys admins tend to refer to and use the m = 1 minute load average For queueing models we want the steady-state average [...] that suggests the m = 15 minute load average is more useful for capacity planning
So load is useful but complex.
Casey experienced some pretty sever load numbers recently…
Tags: definition, dr. neil gunther, gunther, LA triplets, linux top, load, load average, monitoring, performance, performance monitoring, solaris, unix
Comments
5 Responses to “Defining Unix Load Average”
Got something to say?
[...] In his blog entry from late last year, Zach sums it up quite nicely: In short it is the average sum of the number of processes waiting in the run-queue plus the number currently executing over 1, 5, and 15 minute time periods. [...]
[...] Have you ever wondered how load average on Unix is defined? [...]
Friend,
This information is wrong. You forgot the “m” (minute) information:
-5/60m
Attention!!!
Hugs,
Denis
[...] Defining Unix Load Average [...]
[...] 任何一个相对成熟的站点都会利用 Cacti（基于RRDTool） 等工具进行容量规划工作。抓取的 Load 会传 1、5、15 分钟列值过去，这三个度量采用哪个呢? 15 分钟为首选【参见Gunther 的 PPT】。 [...]