Neat article on Amazon.com scalability. Link.
My scalability story:
A late-mid-90s startup. A lightweight broadcast application tied to real-time notification of Olympics competition results, done as a demo of our product (which was actually pretty whizzy, and still would be today). Imagine zillions of users all getting notifications that Nadia Mumblinsky had just won her fifteenth gold medal in the Javelin Catch. Or something.
Anyway, we get some co-lo space at best.com, configure like 8 servers with 100Mbit connections to whatever massive pipe the co-lo has, set up the monitoring software. With our whizzy lightweight messaging we figure that each server can handle like 5,000 connections at once (pretty good for Pentium-200s in that day). We’re not really sure what the load is going to be. We’re ready to drop bucks in the local computer stores and build more servers if we have to. We wait, *cringing* in anticipation . . .
High-water mark is like in the low dozens of users at once. The slowest, measiliest workstation we had at the company could have handled the load. It’s depressing.
Anyway, one of our sales guys is talking to a customer a few months later. “You know,” he remarks, “Our software is so sophisticated, well, with just eight servers we were ready to handle twenty thousand connections at once, in real time!”
“Wow,” says the customer. Or something like that.
“You know how many we got? Over thirty.”
That’s why I’ll never be in sales.
[Wafarfarites: I might have some details wrong. Still a good story 🙂 ]