I liked this book so much I read it twice so that I'd get the maximum benefit from its recommendations. The authors clearly have strong real world experience so their recommendations seem highly relevant. I also like the way the book is organized into 50 rules that you can absorb in small chunks. I was also impressed with the quality of the writing - clear, grammatically correct, and very meaty without a lot of fluff or overly technical jargon.
With regard to the specific recommendations in the book, I was very pleased to see "Don't Overengineer the Solution" as Rule 1. In my experience, this is by far the biggest problem in software today and is the biggest cause of not only scalability problems, but performance, maintainability, and extensibility problems as well.
My only small quibble with the book is the lack of a discussion of object relational mapping technologies. As the authors note several times, the database is usually the most difficult component to scale. I think that is absolutely true. Given that fact, I think the database needs to be fully tuned and optimized and this includes the data access code. As other authors have noted (and my own experience confirms), ORM generally performs and scales very poorly. Not to mention it often represents "overengineering". Therefore, I expected at least some discussion of that issue.
With that small criticism aside, I highly recommend this book.