JPH0324697B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of the Invention) The present invention relates to a real memory management method in a data processing device, and in particular to a real memory management method for multiple sequential programs being executed simultaneously when the system is running out of real memory. The present invention relates to a resource management method that enables efficient real memory allocation.

(Prior Art) In a virtual memory type information processing system that operates with multiple programs, the problem is how to allocate real memory to each program on the real memory. In a virtual memory system, the processing unit
The CPU performs processing while accessing real memory that contains part of the program to be executed.
The entire program is stored on the auxiliary storage device, and the entire program is not accessed all the time. Therefore, sections with a high probability of being accessed are selected for each page, which is the management unit of the program (this is called the working set). (call) into real memory that can be directly accessed by the CPU. This allows many programs to be executed simultaneously with a small amount of real memory, increasing the efficiency of using real memory resources.

If the information that the CPU attempts to access is not in real memory (at the time of a page fault), the contents of the desired logical page of the program are transferred from the auxiliary storage device to the appropriate physical page of real memory.
Prior to this transfer, if there is no free real memory already, the old information that was in the page of the real memory to which the new information is to be transferred is transferred to the auxiliary storage and saved. If the amount of real memory given to a program is smaller than the amount of real memory required by the program (working set size), the probability that the desired program's logical page is not in real memory increases, and the above transfers will occur more frequently. It will be done. For this reason,
The program waits for transfer and is no longer executed. To prevent this, normally when a page fault occurs, the contents of real memory being used by other programs are transferred to auxiliary storage to create free real memory, and this is allocated to the program to increase the working set size. to reduce the number of page faults.

In a multiprogramming system, a large number of programs are processed simultaneously by time sharing, but when the above transfer starts and the program becomes unexecutable, the CPU executes another program. However, if the total amount of real memory required by all programs (sum of working set sizes) exceeds the total amount of real memory in the system, real memory requests will occur one after another each time control is transferred to each program.
Since programs compete for real memory, the above-mentioned transfer to create free memory and transfer of necessary logical page contents from auxiliary storage to real memory occur frequently in all programs. As a result, all programs have a large amount of waiting time for transfer, and the execution of the programs does not proceed (thrashing state).

Conventionally, in order to prevent this thrashing, all the contents of the real memory used by the program with the lowest execution priority are forcibly transferred to the auxiliary storage device, regardless of the amount of real memory that is insufficient, to free up real memory. (swap out), and then allocate the insufficient amount of real memory to other programs with higher execution priority. Swap-out is performed on a program-by-program basis regardless of the amount of real memory that is insufficient, so the amount of free real memory is always greater than the amount of real memory that is insufficient, resulting in unused real memory and effective use of memory resources. The drawback was that it was not possible. Furthermore, since the program with the lowest execution priority is forcibly deprived of all real memory by swap-out, the program cannot be executed at all. As a result, even if other programs cannot be executed due to waiting for input/output, etc., this program cannot be executed, and in the end, the CPU of any program is not used and is wasted.

(Objective of the Invention) In order to eliminate these drawbacks, the present invention has the following objectives: When the total amount of real memory required by each program exceeds the total amount of real memory of the system, () one of the programs belonging to the lowest execution priority is Select the actual memory in use for the missing pages from , and transfer the contents to the auxiliary storage device to create free memory and give it to other programs that lack real memory, but leave the rest as is and do the above. The program to be taken memory is made executable, and () in the event of a page fault of the program to be taken memory, that program is given high priority by not being given more memory than the amount of real memory currently in use by that program. () To prevent thrashing from occurring in high-execution-priority programs by preventing real memory used by programs from being taken away; If other higher execution priority programs run on the CPU,
By giving these remaining system resources to a program only when the auxiliary storage control unit (channel) is not in use, the use of system resources for handling page faults, which occur frequently in that program, is reduced to the execution of high-priority programs. Avoid causing disturbances.

This is achieved using a resource management method that combines a memory schedule and a CPU schedule, and will be explained in detail below using the drawings.

(Structure and operation of the invention) FIG. 1 is a block diagram showing the structure of an embodiment according to the method of the present invention, in which 1 is a real memory (MM) in which a program P is stored during execution, and P ₀ to _{P o} are Program name 2 is the program storage area on the real memory that stores the program, PA ₀ ~ _{PA o} are the areas that store part of each program P ₀ ~ _{P o} ,
3 is a memory management table that manages the amount of real memory in use (working set size) WS for each program P ₀ to _{P o} and the pointer PTR to its mapping control table 5; 4 is the amount of free real memory E and its physical Free real memory management table that manages the page number PPAGE, 5 is the logical page number LPAGE of the program and the physical page number of the real memory
A mapping control table that manages for each program a flag A indicating whether or not the correspondence of PPAGE and the contents of the logical page are already in a physical page on real memory; 6 is a mapping control table that manages programs P ₀ to P _o for each execution priority PR; 7 is a memory control mechanism that controls memory and transfers data; 8 is a processing unit (CPU) that is an execution unit for executing programs; 9 is an execution priority management table for each program P ₀ to P _o 10 is a mapping control mechanism that converts between the logical page number LPAGE and the physical page number PPAGE and detects page faults; 10 is a page selection mechanism that determines which physical page of real memory is given when the program P requests real memory; 11 is a page-out control mechanism that transfers the contents of the physical page of the real memory to a logical page on the auxiliary storage device 16 to make the real memory free; 12 is the auxiliary storage device 16;
a page-in control mechanism that transfers the contents of the logical page of the upper program to a free real memory and makes the physical page containing the contents of the logical page of the program P desired by the processing unit (CPU) 8 accessible; 13 is an execution priority control mechanism that changes the execution priority of a program; 14 is a CPU control mechanism that controls the entire processing device 8 and transfers data; and 15 is an auxiliary device that controls the auxiliary storage device and transfers data. Storage control device, 16 is a program
An auxiliary storage device in which all logical pages from P ₀ to P _o are stored; 17 is a data transfer path that transfers data between the real memory (MM) 1, the processing unit (CPU) 8, and the auxiliary storage device 16; be. Note that the value of flag A in the mapping control table 5 is 1 if the content of the logical page already exists in any physical page, and 0 if there is none.

Next, the operation will be explained. The processing device 8 proceeds with the processing while accessing PA _i in the program storage area 2 on the real memory. In the virtual memory system, the entire programs P ₀ to _{P o} are stored on the auxiliary storage device 16, and the contents of the logical pages of each program P are transferred to the program storage area 2 on the real memory 1 via the data transfer path 17 when needed. is copied to the physical page on PA _i and becomes accessible. When the processing device 8 accesses the program, it does so using the logical page number LPAGE _j , but this logical page number LPAGE _j is once accessed by the mapping control mechanism 9.
passed to. The mapping control mechanism 9 activates the memory control mechanism 7 via the CPU control mechanism 14,
The flag A _j and the physical page number PPAGE _j in the entry for the logical page number LPAGE _j of the mapping control table 5 are obtained through the data transfer path 17 . In this way, the mapping control mechanism 9 determines the logical page number according to the mapping control table 5.
Convert to physical page number PPAGE _j corresponding to LPAGE _j . Further, the physical page number LPAGE _j is transferred from the mapping control mechanism 9 to the memory control mechanism 7 via the CPU control mechanism 14 and the data transfer path 17.
passed to. Next, this memory control mechanism 7 retrieves the data and transfers the desired page to the processing device 8 through the data transfer path 17 to complete the access. If the value of the flag A _j of the entry corresponding to the logical page LPAGE _j of the program received by the mapping control mechanism 9 is 0, that is, the contents of the desired logical page LPAGE _j are not found in any physical page (page fault). , the mapping control mechanism 9 is the CPU
The control mechanism 14 is notified of the occurrence of a page fault.

For example, the processing when a page fault occurs and a real memory request is made during the execution of program P _i will be explained step by step.

() From mapping control mechanism 9 to CPU control mechanism 1
4 is notified of the occurrence of a page fault,
The CPU control mechanism 14 activates the page selection mechanism 10.

() The page selection mechanism 10 is the CPU control mechanism 14
The free memory entry E of the free memory management table 4 on the real memory is accessed and checked via the data transfer path 17. If the value of E is not 0, that is, if there is free real memory, select one physical page registered in that table and set its physical page number.
The page-in control mechanism 12 is activated by notifying the page-in control mechanism 12 of PPAGE _k and the logical page number LPAGE _j in which the page fault has occurred.

() Free memory entry E is 0 in ()
If there is no free memory, the page selection mechanism 10
is the data transfer path 1 via the CPU control mechanism 14.
7, accesses and checks the entries PR _{0 to PR} Select as. Furthermore, by looking at the mapping control table 5 pointed to by PTR ₁ of the P ₁ entry in the memory management table 3, one page PPAGE _k is selected from among the physical pages being used by the program P ₁ to be taken up, and the page The page-out control mechanism 11 is notified of each program P ₁ to be taken up from memory and the selected physical page number PPAGE _k , and the page-out control mechanism 11 is activated. However, if the program P _i that caused the page fault is one of the programs that belong to the lowest execution priority,
The program itself is selected as the program to be memory taken. Moreover, at the same time as starting the page-out control mechanism 11 in this way, the execution priority control mechanism 13 is also notified of the program name P ₁ to be taken up from memory, and the execution priority control mechanism 1
Start 3.

() When the real memory 1 has a low usage rate and is not in a memory shortage state, the execution priority control mechanism 13 sets the memory takeover target program P ₁ to the lowest execution priority PR _X in the system where no program is set. In order to set , the name of the program to be taken up from memory P ₁ is registered in the PR _X entry of the execution priority management table 6 via the CPU control mechanism 14 and the data transfer path 17 . Therefore, the program targeted for memory takeover runs only when all other programs are not running, and the page fault of that program occurs while the program is running, that is, when all other programs are running on the processing unit 8 and Since this occurs while O is not being used, an increase in the processing device 8 and I/O load due to frequent page faults does not affect the execution of other programs.

() The page-out control mechanism 11 activates the contents of the physical page PPAGE _k notified in () to the memory control mechanism 7 and the auxiliary storage control device 15, and transfers the contents from the real memory 1 to the auxiliary storage device 16 through the data transfer path 17. Transfer to. After this transfer is completed, the physical page number PPAGE _k that became free due to this process and the logical page number LPAGE _j where the program P _i caused the page fault are transferred in the same way as when there is already free memory (). Then, the page-in control mechanism 12 is notified and the page-in control mechanism 12 is activated.

() The page-in control mechanism 12 activates the memory control mechanism 7 and the auxiliary storage device 15 via the CPU control mechanism 14, thereby transferring the program from the auxiliary storage device 16 through the data transfer path 17.
P _i causes a page fault and makes a request to copy the contents of logical page LPAGE _j to the real page PPAGE _k , making it accessible. Furthermore,
Data transfer path 17 via CPU control mechanism 14
Through mapping control table 5 logical page number LPAGE _j entry to physical page number
PPAGE _k is set, and the flag A _j of the entry is set to 1 to indicate that the contents of the logical page already exist in the real memory. Next, the page-in control mechanism 12 notifies the CPU control mechanism 14 that the desired page has become accessible, and completes all page fault processing.

FIG. 2 shows a flowchart of these processes.

An embodiment of the present invention has been described above, but the gist of the present invention is that when the total amount of real memory required by each program exceeds the total amount of real memory of the system, the program belonging to the lowest execution priority is Selects the actual memory in use for the missing pages from one program and transfers its contents to the auxiliary storage device to create free memory and give it to other programs that lack real memory, but gives the rest. If the above program to be taken memory is executable as it is, and if a page fault occurs in the program to be taken memory, by not giving memory in excess of the amount of real memory being used by that program, it will be possible to execute the program to be taken up by a high-priority program. This prevents real memory from being taken and suppresses the occurrence of thrashing in high-priority programs, and sets the program targeted for memory take-up as the only one with the lowest execution priority in the system, allowing other programs with higher execution priorities to processing device 8,
Only when the auxiliary storage control unit (channel) 15 is not used, these remaining system resources are given to the above program, and the use of system resources for handling page faults that occur frequently in that program is reduced to high execution priority programs. Avoid causing execution blockage.

This is a comprehensive resource management method that combines a memory schedule and a processing unit schedule.

(Effects) As explained above, according to the present invention, even if the memory load is high and the working set size of a program with the lowest execution priority is not satisfied, all real memory used by that program can always be swapped out. Resource allocation to the lowest execution priority program so that it uses only the remaining memory, excluding the real memory used by the high priority program, and runs when all other high priority programs are not using the processing unit. This eliminates the need for the program to run at all, which is the case with conventional methods, and improves processing efficiency.

In particular, programs that have low execution priority and require a large amount of real memory to the extent that running them will inevitably lead to a memory shortage state, which would be difficult to run using conventional control methods, can now be executed using this control method. Furthermore, even if the system falls into a thrashing state due to a high memory load, real memory will be taken up in order from the program with the lowest execution priority to the amount that other programs lack, and the program with the lowest execution priority will always take up the remaining memory used by other programs. Since it is executed using only resources, the low execution priority program is automatically stopped, and when the memory load decreases, real memory is automatically given to the low execution priority program and restarted. In addition, system operators have to stop low-priority programs to prevent interference with the execution of high-priority programs, or restart low-priority programs when the system memory load drops again. Another advantage is that no intervention is required.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing the configuration of an embodiment according to the method of the present invention, and FIG. 2 is a flowchart of the processing of the present invention. 1...Real memory, 2...Program storage area,
3... Memory management table, 4... Free real memory management table, 5... Mapping control table,
6... Execution priority management table, 7... Memory control mechanism, 8... Processing device, 9... Mapping control mechanism, 10... Page selection mechanism, 11... Page out control mechanism, 12... Page in control mechanism,
13... Execution priority control mechanism, 14... CPU control mechanism, 15... Auxiliary storage control device, 16... Auxiliary storage device, 17... Data transfer path.

Claims (1)

[Claims] 1 In a virtual memory information processing system that operates with multiple programs and each program is executed according to its execution priority, when allocating real memory between multiple sequential programs that are being executed simultaneously, each A mechanism that allocates all real memory to each program in order of program execution priority, and when the amount of remaining real memory is less than the amount of real memory required by the next program to be allocated, only the remaining amount of real memory is given. Then, the execution priority of the program whose allocated real memory is less than the required real memory is set to the lowest execution priority in the system, and the real memory that occurs due to real memory competition due to lack of real memory is set to the lowest execution priority in the system. A resource management method characterized by having a mechanism for suppressing a state in which a program cannot be executed (thrashing) caused by frequent transfer of memory contents between a storage device and an auxiliary storage device.