In many cases, SSAS works efficiently with default settings right out of the box. However, when you have large databases, substantial number of concurrent users, insufficient resources on your server, or when best practices are not followed during SSAS database design, you can run into performance bottlenecks and problems. In these scenarios, you need to know what to measure and how to measure them, what’s normal for your environment (benchmark), and you need to have some amount of historical measurements to be able to see the events that lead to a certain bad performance/failure point. Once you have this data, you can improve your server’s performance by addressing the problem(s).
This is a tour de force of an article, absolutely worth reading if you plan on dealing with Analysis Services at some point. Even if you don’t build your own tool, you’ll learn a lot about what drives SSAS performance and what indicates that there might be a problem.