Read an Excerpt
ForewordForeword
It was a pleasure to be asked to write the foreword to this new book, which is remarkable for two reasons:
- People who have spent five years developing a product are normally more than ready to move on to the next release once the product is finally ready for release. Indeed, long before a new version gets into customers' hands, the developers are normally already working on the next release. So, for the actual developers to spend the considerable time that this book must have taken to write a lengthy, detailed book on it is very rare.
- In my years as an industry analyst with The OLAP Report, and much earlier as a product manager, I have rarely come across developers who are prepared to provide such chapter and verse information on exactly how a product works. Even under NDA, few software vendors are prepared to volunteer this level of inside information.
But why should this be of interest to anyone who isn't an OLAP server developer? Why should a mere user or even an application developer care about what exactly happens under the hood, any more than ordinary car drivers needs to know the details of exactly how their car's engine management system works?
There are some good reasons why this is relevant. Analysis Services is now by far the most widely used OLAP server, which inevitably means that most of its users are new to OLAP. The OLAP Surveys have consistently found that the main reason for the choice is price and the fact that it is bundled with SQL Server, rather than performance, scalability, ease of use, or functionality.
This is not to say that Analysis Services lacks these capabilities; just thattypical Analysis Services buyers are less concerned about them than are the buyers of other products. But when they come to build applications, they certainly will need to take these factors into account, and this book will help them succeed. Just because Analysis Services is perceived as being a low-cost, bundled product does not mean that it is a small, simple add-on: particularly in the 2005 release, it is an ambitious, complex, sophisticated product. How it works is far from obvious, and how to make the most of it requires more than guesswork.
Many of the new Analysis Services users will have used relational databases previously, and will assume that OLAP databases are similar. They are not, despite the superficial similarities between MDX and SQL. You really need to think multidimensionally, and understand how Analysis Services cubes work.
Even users with experience of other OLAP servers will find that they differ from each other much more than do relational databases. If you start using Analysis Services without understanding the differences and without knowing how Analysis Services really works, you will surely store up problems for the future. Even if you manage to get the right results now, you may well compromise the performance and future maintainability of the application.
The OLAP Surveys have consistently found that if there is one thing that really matters with OLAP, it is a fast query response. Slow performance is the biggest single product-related complaint from OLAP users in general, and Analysis Services users are no different. Slow query performance was also the biggest technical deterrent to wider deployment.
Many people hope that ever improving hardware performance will let them off the hook: If the application is too slow, just rely on the next generation of faster hardware to solve the problem. But results from The OLAP Surveys show that this will not work—the rate of performance complaints has gone up every year, whether actual query performance has improved or not. In an era when everyone expects free sub-second Web searches of billions of documents, books, and newsgroup postings, they are no longer willing to wait five or ten seconds for a simple management report from a modest internal database. It is not enough for an OLAP application to be faster than the spreadsheet or relational application it replaced—it must be as fast as other systems that we all use every day.
The good news is that fast query performance is possible if you take full advantage of the OLAP server's capabilities: The OLAP Survey 6 found that 57% of Analysis Services 2005 users reported that their typical query response was less than five seconds. This was the traditional benchmark target query time, but in the new era of instant Web searches, I think the new target should be reduced to one second. This is a tough target, and will require application developers to really know what they are doing, and to take the time to optimize their systems.
This is where this book comes in. The authors—who have been involved with Analysis Services from its earliest days, long before it was called Analysis Services—have documented, in detail, what really happens inside Analysis Services 2005, right down to the bit structure of data records. Along the way, numerous controllable parameters are described, with helpful information about how they cause memory or other computer resources to be used.
This book is not intended to teach new users how to use Analysis Services 2005; it is for technically competent implementers who want to make the most of Analysis Services by understanding how it really works, as described by those who really know, unlike other books written by external authors who sometimes have to speculate. If you are new to Analysis Services, you probably need to start with a "how do I?" book or course, rather than a "what happens inside?" book like this one.
Nigel Pendse
Editor of The OLAP Report
Author of The OLAP Survey
© Copyright Pearson Education. All rights reserved.