ols.wtf / blog / pizza-24-7-the-pursuit-of-100-uptime

Pizza 24/7: the pursuit of 100% uptime

Oliver Leaver-Smith <oliver@leaversmith.com> on 2023-10-17 09:53:06

This is a first-hand anecdote I often tell when talking about SLOs, SLAs, and system reliability. Someone asked that I write it down, so here it is. It's all about my favourite pizza shop as a student

I would finish work quite late at night in the city centre, and lived about a 25 minute walk away. About half way along my route home was a pizza shop called "Pizza 24/7". As you might imagine from the name, this establishment was "always open" and so was perfect for me when I was coming past about 10pm, 2am, 4am, etc. I'd ring up when leaving work, and collect my hot and fresh medium pepperoni to eat on the rest of the walk home.

Any time I would ring after around 1am, it would sound like the guy was sleeping. But he dutifully woke up and made me the pizza. I don't want to think about what extras he put in for the inconvenience.

One night I called him, around 3am to request the usual. No response. I called again. The phone was picked up and then he hung up without so much as a greeting. I called a third time and couldn't get through. He'd unplugged the phone or turned it off.

What did I do? I went elsewhere and got some chips. I am a user, I have no loyalty. Not only did I stop calling for pizza in the small hours, I stopped calling for his pizza at "normal" hours too.

If he'd sold his pizza as "midday til midnight" or something like that from the beginning then I would have never tried to call outside those times. But because the option was there, I was hurt when it wasn't honoured. A few months later, the shop was closed.

So what is the lesson? "24/7 pizza" is like five-nines availability. It sounds great on paper, but without severe investment and business structures in place, it's just words. It's not attainable. And worst of all, your users lose confidence in what you're selling.

Much better to go with an SLO that is fair and reasonable than to chain yourself to an unhealthy and personally detrimental way of running a system. And it's important that those making decisions understand just how much effort it takes to go from 99.9% to 99.999% uptime

I walked past the pizza shop the other week. It's changed its name and while it was closed, it was obviously still in business. It makes me happy that they're back. I might just give them another chance.