Here is a blog post to once and for all dispel the difference between scalability and performance!

Let's imagine on non computer task. Let's say the your business is to carry goods from point A to point B on the back of a donkey. Business is good, so you seem to be getting more and more requests to carry such goods. Soon enough you realize that you need a donkey that performs better! So you kill your donkey, and you get a body builder donkey! As business grows, his body builder donkeys in enough either! So you kill him, and instead you replace him with a really expensive Sylvester Stallone donkey. Business is getting even crazier, you kill the Sylvester Stallone donkey, and you get an Arnold Schwarzenegger donkey and you pump him with steroids! By now you're donkey is really performant, but also really expensive!

What is important to realize here, is that the same purpose could've been achieved by numerous cheaper donkeys, by scaling your operation amongst the numerous cheaper donkeys. This means, a scalable architecture breaks down its tasks into equivalent activities, that any cheaper donkey can pick. The process of finding a stronger and stronger donkey is aiming for better performance! The equivalent of replacing weaker servers with stronger servers.

The process of distributing, distributable load across multiple servers, is referred to as scaling out your operation.

As you can see, scalability is a very different animal than performance. The end result is perhaps the same you're able to support more business, and you're able to support more users hitting your system. It is how you solve the problem, by providing better performance, or better scalability - is that also differentiates the cost. Numerous less powerful servers are almost always cheaper than a single very high performance supercomputer.