p. 203

2-2-3 Memory Management

Memory management is a management function for the effective use of a computer’s memory areas. Memory management is classified as either real memory management or virtual memory management.

(1) Real memory management

Real memory management is a management function for the effective use of main memory areas. A program to be executed must be recorded in main memory but there is a limit to the storage capacity of memory. Therefore, this method uses an auxiliary storage device.

(1) Partitioning method This is a method that divides main memory into partitions for management. It is classified as either fixed partitioning method or variable partitioning method.

A: Fixed partitioning method This is a method that divides the usable area of main memory into several partitions of a fixed size. This method reads (roll-in) a program or data from an auxiliary storage device when it is necessary, and writes (roll-out) the program or data back to the auxiliary storage device when it is no longer necessary.

p. 204

B: Variable partitioning method This is a method that partitions the usable memory area of main memory according to the capacity required by a program.

When the variable partitioning method is applied, many small areas of garbage (i.e., unused areas) are generated because a partition remains unchanged after the execution of a program is completed. This is called fragmentation. In this case, garbage collection (or memory compaction) is executed to combine these unused areas into one large area. · 1 question

Garbage collection can also be expected to have an effect whereby the potential for memory leak to occur is reduced. A memory leak is caused by a bug or such other problem in a program (e.g., an OS or an application) that uses memory (i.e., main memory), and areas that

p. 205

were secured are not released during operation. If a memory leak occurs, usable memory areas in main memory decrease, and serious problems may be caused.

(2) Overlay method (Segment method) This is a method that is used in order to execute a program that is larger than the usable area of main memory. It divides the program into multiple segments that are not executed concurrently and transfers segments to and from an auxiliary storage device upon execution.

(3) Swapping In multiprogramming and such other environment, swapping refers to interrupting execution of the current program and transferring it to an auxiliary storage device in order to execute programs or such other processes of a higher priority. As soon as main memory has available space (or when execution is restarted), the program that was transferred to the auxiliary storage device is retrieved and processing is restarted from where it was interrupted. The activity to transfer from main memory to auxiliary storage is called swap-out, and the retrieval activity is called swap-in. If swapping occurs frequently, much time (i.e., overhead) that is not related to a program is required, and processing efficiency is drastically decreased.

(2) Virtual memory management

Virtual memory management is a virtual memory system that implements a concept called “virtual memory” by establishing a logical memory space (a virtual memory space or a logical

p. 206

address space) that is larger than main memory and executing programs that are stored in it. (In reality, only programs in main memory can be executed, so the section to be executed is loaded (i.e., transferred) from auxiliary storage to main memory).

Virtual memory management is classified as shown below according to the difference in the management method of memory areas.

In a virtual memory system, it is necessary to convert between the virtual memory address (logical address) and the main memory address (physical address). This conversion is called address conversion and is performed by hardware called a DAT (Dynamic Address Translator).

p. 207

In the segment method, address conversion is performed using the base address method, which references a segment address table in which the starting points (i.e., physical addresses) of segments are recorded. On the other hand, in the paging method, addresses conversion is performed by a DAT referencing a page address table in which the physical address of each page is stored.

In this case, when a page that is not recorded in the page address table (or not recorded in main memory) is referenced, an interrupt called a page fault occurs from the DAT. When a page fault occurs, the page for reference is loaded into main memory, and the page address table is updated. If main memory has no free space at this time, a page (i.e., page for replacement) from main memory is selected, a page-out is performed, and then a page-in is performed in order to put the referenced page into the free space. If page faults frequently occur, page-in and page-out increase, and a state called thrashing occurs in which processing efficiency decreases. Therefore, in order to minimize the occurrence of page faults, it is necessary to perform a page-out for pages that have a low probability of referral after this point. The method that is used to select such pages is called a page replacement algorithm, and there are three typical ones. · 2 questions

p. 208

The apparent memory capacity of main memory can be increased with the virtual memory system. However, if the capacity of main memory is too small, thrashing occurs and processing efficiency decreases. This means that for running applications that will not run because of insufficient memory, increasing the size of virtual memory is only a temporary solution. In such cases, it is necessary to consider increasing the usable memory areas by adding memory or stopping the use of unnecessary applications.