The Relational Database Engine

You can look at the relational engine in two ways: you can study it at the core, or the relational kernel, or you can study it at a higher level with all its surrounding components that the kernel hands functionality off to. Studying the kernel at any lower level is beyond the scope of this book. A good analogy, however, is what we need to study in a steam locomotive engine, the pistons and drive shaft being the kernel, and the rest of the engine, comprising the coal burners, wheels, knobs, pulleys, levers (not to mention the engineer and the horn), representing a higher level. Let’s look at the kernel in all its functional glory.

Note 

At the very core of the SQL Server relational engine, the kernel is the same code used for each version of SQL Server, from the Mobile Edition (which is embedded into your average Windows Mobile device) to the Enterprise Edition parked on the shared or clustered disks of a TerraServer. This is the reason you can write code for a handheld and execute it with very little modification on any other version of SQL Server 2005.

At the kernel level, you have a highly optimized chunk of code that picks up inbound T-SQL statements from the Net-Libraries and optimizes it into native code for the most effective execution plan. (In many respects, the relational engine is like a compiler that preprocesses code into object code before compiling it into machine code. It is no different from the preprocessing algorithms used in many different kernels, aside from database systems.)