MovableBlog: One Nine
February 19, 2003
It's very bad of me to quote without attribution here—okay, from top of the thread to the bottom, and using first names only, it's Ryan, Keith, Peter, and Erick—but this quote represents what I love about the sfu.unix newsgroup. Very geeky, and very picky about things sometimes. And often very off-topic. (This is part of a thread discussing my alma mater's consideration of removing all Microsoft products from the lab computers and going open-source.)
>>>(single point of failure) inter-campus link at DC in all of 2003, it
>>>will have operated at roughly "one nine" service levels for the entire
>>"One nine". Useful phrase, that. Covers the whole huge range between 9%
>> and 99%.
> Actually no, in this context it refers to the number of 9s after the
>decimal point (with %99. being assumed). The public telephone network typically
>strives for 5 9s reliability (i.e. 99.99999 uptime). One 9s reliability is
>considerably less reliable ...
Wrong, five 9's reliability is 99.999% uptime. The correct way to look at this is the number of 9's after the decimal point with "0." being assumed. One nine is thus 0.9, or 90% uptime.
This allows "nines" to be used as fractional units on a log-scale like decibels
"Zero nines" would mean no uptime and "half a nine" would be about 68% uptime.
"Five and a half nines", 99.99968% uptime. Works nicely, doesn't it?
I've been lurking in the newsgroup for at least 6 years now. Also, the sfu.test newsgroup has some of the funniest rejoinders to test messages I've ever seen.