NINETY DAYS
NINETY INCIDENTS
NINETY PERCENT
YOU PAID FOR ALL FIVE NINES BUT YOU’LL ONLY NEED THE EDGE
NINETY DAYS
NINETY INCIDENTS
NINETY PERCENT
YOU PAID FOR ALL FIVE NINES BUT YOU’LL ONLY NEED THE EDGE
@gkrnours @0xabad1dea Ooh ooh, I can answer this!
Given that 100% is hard to reach and 0% is not, we can use a simpler version of the logit function and take log₁₀(1÷(1−.89)) = ~0.96 which means it's still a tad below 1 "nine".
(The full logit function is log₁₀(x÷(1−x)) and is useful when both 0% and 100% are hard/impossible to reach.)
Nine fives
@0xabad1dea Once upon a time, at a small entertainment company that rhymes with 'misery', I was responsible for the operations of the Github Product.
It came as a vmdk. No documentation.
I was told "Backup" was as simple as snapshotting the VM.
It ran gunicorn.
There was no tuning, monitoring or built-in telemetry.
They refused to give us an extension on licensing when they forgot to cash our check, and shut it off from remote.
Every call was with techbros.
I cannot stress enough, how happy this makes me.
@slembcke to confirm what someone else said while I was asleep, it’s an independent third party tracker called the “missing” github status page. https://mrshu.github.io/github-statuses/
they track ten subsystems separately, which are each in the 96% to 99.9% range, then overlay all those incidents on the same timeline to arrive at 90% overall.
I do not have any particular stance on whether github’s own tracker or this third party one is more fair and accurate, beyond “90-90-90 is hilarious” and observing only one of them has a contractual stake in possibly sometimes downplaying issues a little bit.
@0xabad1dea There's a 996 joke in here somewhere, I swear.
Edit: 996: one 9 of uptime, 9-0 incidents, and the third nine in "90 days" is upside down lol. We can workshop it.
Somebody just has to type "continue" into the right co-pilot chat at Github.
@0xabad1dea the green bits are 100%
It’s 99.999% but we’re not telling you the timeframe.