This book doesn't deal with true high-frequency trading, where it is more about execution than anything else. The book IS ten years old when I write this, so high frequency trading has taken on a different meaning, so no false advert here.
That said, it is a great treatment of the practical issues of handling large, heterogeneous financial data sets and their statistics. I haven't seen their methodology and framework anywhere else, although there are some really good treatments of irregularly spaced financial data (Hautsch, Engle).
The authors are prolific in this area, in particular, the use of tick data to build better volatility models and the use of seasonality (business time scale) and stochastic time (see intrinsic time). They also present a good way to use higher frequency homogeneous data to effectively filter historical volatility computations that makes them more robust when the data is interpolated or sparse. The best part is that they bring everything together for use in multivariate cases and for forecasting/trading.
Overall, this is a great book, that doesn't have many peers (if any). I can't recommend it enough.
Minor downsides:
(1) I also agree with the other reviewers on the notation, although it doesn't bother me that much personally.
(2) Would be nice to see some type of flowchart for an implementation of the methods in Ch. 6 and later, like they did in Ch. 4.
(3) No explicit mention of duration and/or point processes, although it is implicit in many of their techniques. This one might be a little unfair because one can't expect the authors to survey the entire body of literature.