[ILUG] Linux uptime statistics

[Niall O Broin]

On Tuesday 4 May 2004, rick at linuxmafia.com (Rick Moen) wrote:

>Quoting James McCarthy (jamesmccarthy at dublin.ie):
>
>> What are the current linux uptime statistics?
>> Something like 99.99999% or something?
>
>Sometimes, I consider spending ninety seconds per quarter rebooting to
>get sundry new-kernel benefits.  But then I'd suffer a humiliating (1 -
>90 / (60 x 60 x 24 x 90)) * 100  = 99.988402% uptime figure.  Woe is
me!
>
>I guess I should stick with antique kernels, then.

You should certainly stick to Linux, rather than statistics. With your
arithmetic, 90 sec. downtime per quarter only gets you three nines,
whereas it's really four, and damn near five. 

[Reliability or availability in engineering is often expressed as N
nines e.g. 4 nines is 99.99% reliable / available. Below is a table of
nines against seconds downtime per year, for a 365.25 day year:


Avail. %    Seconds

99          315576
99.9        31557.6
99.99       3155.76
99.999      315.576
99.9999     31.5576
99.99999    3.155760002
99.999999   0.315575998
99.9999999  0.031557599

]

So for a given Linux installation, you can have 100% uptime, depending
on when you count. But 99.99999% uptime allows for 3 seconds of downtime
per year - not going to be doing much maintenance there.

Of course in any real engineering project, availability of the system is
what's measured, so you have lots of redundant systems, such that the
failure of any one is not critical. This is especially important in e.g.
aviation, a field where uptime is a very meaningful word.


Niall