Wednesday, November 15, 2017
02:00 PM - 02:50 PM
|Level: ||Technical - Advanced|
It is tempting to wish that performance issues can be crushed with the overwhelming compute power of modern server systems, but reality can still bite back. Software not architected for important hardware characteristics, specifically that total capability is distributed over very many processor cores, likely on a system with non-uniform memory access may fall far short of expectations. While modern database systems are very sophisticated, the core engine employs a set of methods based on reasonable expectations for general transaction processing and data warehouse patterns such that a typical OLTP or DW should run well out-of-the-box. But many applications use the database in imaginative and unanticipated patterns which do not work well with the standard rules. By understanding in detail how the database engine works, it is possible to design an application to overcome almost any such problems.
Joe has been a database consultant for nearly 20 years, specializing in SQL Server and storage system performance with some work in other database systems and application servers. Joe is best known for reverse engineering the SQL Server cost-based query optimizer, pioneering the use of statistics-only shell database in SQL Server, and establishing a quantitative approach to performance analysis.