Before you can evaluate the performance advantage that reference pattern services offer, you must understand some facts about how the operating system handles the data your program references. The system divides the data into 4096-byte chunks; each chunk is called a "page". For the processor to execute an instruction, the page that contains the data that the instruction requires must reside in central storage. Central storage contains pages of data for many programs — your program, plus other programs that the system is working on. The system brings a page of your data into central storage when your program needs data on that page. If the program uses the data in a sequential manner, once the program finishes using the data on that page, it will not immediately use the page again. After your program finishes using that page, the system might remove the page from central storage to make room for another page of your data or maybe a page of some other program’s data. The system allows pages to stay in central storage if they are referenced frequently enough and if the system does not need those pages for other programs.

The process that the system goes through when it pauses to bring a page into central storage is called a "page fault". This interruption causes the system to stop working on your program (or "suspend" your program) while more of your program’s data comes into central storage. Then, when the page is in central storage and the system is available to your program again, the system resumes running your program at the instruction where it left off.

Reference pattern services can change the way the system handles your program’s data. With direction from reference pattern services, the system moves multiple pages into central storage at a time. By bringing in many pages at a time, the system takes fewer page faults. Fewer page faults mean possible performance gains for your program.