Transitioning of SQLOS to SQLPAL

​The SQLOS​

Prior to SQL Server 2005, SQL Server runs as a user mode process like any other Windows application and it depends on Windows OS to manage hardware resources

​. The hardware resource management includes the following;

• ​
  ​Processor scheduling
• Memory management
• Network 
• Disk I/O​

In SQL Server 2005, the hardware resource management part has been incorporated into SQL Server as a separate component known as SQLOS (SQL Server Operating System). The objective of this initiative was to improve SQL Server performance by assigning database own tasks manage itself, so that SQL Server can perform better. The SQLOS then acts as a non-preemptive schedule to service database own tasks by cooperating with Windows OS. As a result, SQL Server knows how to manage the hardware resources better than Windows OS. 

The SQLPAL

SQLPAL-SQL Platform Abstraction Layer, is an extended version of SQLOS to facilitate platform independent behaviour with the introduction of SQL Server on Linux initiative. To achieve this objective Microsoft has used a research project called, Drawbridge. Drawbridge uses essentials Windows components inside SQLPAL to interact with SQL Server and then SQLPAL will interact with the underlying OS. Essentially, SQL Server on Linux still thinks SQL Server is working on Windows. See below Figure;

The SQLPAL interact with underlying OS (Windows or Linux) through its host extension implementation. So host extension has the actually platform specific implementation for Windows OS as well as Linux OS. The biggest advantage of this innovative approach is, you did not need to rewrite SQL Server where more than 20+ years of development happened on Windows and contains thousands of millions of code. 

​How about the performance?

The obvious doubt is, adding an additional abstraction layer (SQLPAL) will negatively impact the performance?​ The testing shows that SQL Server performs almost the same against both the OS. In fact, the performance of SQL Server 2017 is far better than SQL Server 2016. 

​Need to know more: Click here​

Cheers!