Mongostat used Vs dirty

I think MongoStat is a powerful debugging tool, that can help in production.

In the course video, when they are explaining MongoStat they say when dirty is quite larger than used, they conclude that working set is quite larger than RAM. How can we make that inference?

His actual statement : Dirty is less than Used , it means that we don’t have enough space to accommodate all the data set that we need. This is confusing. He says that when dirty and used are very near to each other, it means queries are faster. This is also confusing. Why should it necessarily be?