XML processing performance in database systems depends on static optimizations such as XML query rewrites, cost-based optimizations such as choosing appropriate XML indices, and the efficiency of runtime tasks like XML parsing and serialization. This paper discusses some of the runtime performance aspects of XML processing in relational database systems using Microsoft^® SQL Server^TM 2005’s approach as an example. It also motivates a non-textual storage as the preferred choice for storing XML natively. A performance evaluation of these techniques shows XML query performance improvements of up to 6 times.