One Core i7 box with 8GB ram and a second database slave (for replication) with only 1GB of ram. I was running node.js, mongodb and redis all on the same box. I

One Core i7 box with 8GB ram and a second database slave (for replication) with only 1GB of ram. I was running node.js, mongodb and redis all on the same box. I believe it could have handled a lot more, because it wasn't really breaking a sweat on CPU or memory usage, and my dataset was small enough to fit into memory. Perhaps "scaled" was the wrong word, as I wasn't using multiple nodes horizontally, but you'd only have to add more database servers to scale as the load contribution of node.js was absolutely tiny. If you were doing huge numbers you could probably also put a load balancer in front of a couple node.js application servers, each connecting to the same mongodb replica set (which you could easily scale by just throwing in more nodes). But if you're doing an app like this one, you couldn't use nginx as a load balancer because it won't proxy WebSockets.