I always find it funny that half the people/companies advocating these approaches can't even keep a simple WordPress blog up when faced with a bit of traffic.
Surely there must be an algorithm for that?
Surely there must be an algorithm for that?
And BOOM, suddenly you find out that the large instance on AWS, cannot handle the traffic, you are befuddled its not like the server is getting million hits/sec. You thought a large server would be enough to handle a single wordpress blog, and you realize that's not the case.
Now you go back and start fixing your self hosted wordpress blog, you feel sad that you missed out on audience because the server crashed, your work gets derailed because now you also have to maintain the blog server.
I agreed I too would have been irritated to see the link not working, but its not that you switch to an x-large instance each time you post a link on HN. Sometimes you just want somethings to work. But alas that's not the case.
I'll use myself as an example: I'm a self-taught developer who has built and managed high-profile, high-usage websites/web applications. I probably could not pass one of your algorithm-based FizzBuzz tests, but if I was hiring and you suggested that a site should scale simply because you threw it on a large AWS instance, I would not consider your candidacy further regardless of your ability to pass a FizzBuzz test.
You want to focus on the product, fix bugs, look after feature request, produce high quality content. And then your blog crashes and things get derailed.
That's precisely what you said:
"You thought a large server would be enough to handle a single wordpress blog..."
> when you are swamped with so many things you don't want to spend too much time on things like maintaining your blog.
A competent developer should be able to set up a load balanced WordPress blog (with Nginx/PHP-FPM, Varnish and a separate database server) in a day or less. There are a number of well-written tutorials on this.
If they pass that, we don't give them any more programming tests. We give them a hands-on sysadmin assault course problem, where they have to SSH into an EC2 box that has been carefully broken by our devious chief sysadmin, find out what's wrong, and fix it. They do it in a tmux session, with some kind of audio link (Skype, Hangout, etc), so one of our guys can watch and talk to them as they do it. I've listened to several people running through it, and it sounds terrifying.