JPH0448355A - Distributed paging control system - Google Patents

Abstract

PURPOSE:To shorten the overhead caused by the replacement of pages by providing an extension page map table within its own computer and a distributed paging control part in its own computer to store the pages transferred from another computer into an idle memory block of a memory provided in its own computer. CONSTITUTION:When a distributed paging control part 115 included in its own computer 11 drives out the pages stored in a memory 112 of the computer 11, the presence of an idle memory block is discriminated in a memory 122 of another computer 12 by reference to a memory block control table 114 contained in the computer 11. In such a case, the control is carried out to drive out the pages to the memory 122 of the computer 12. If it is discriminated that a request is received to drive out the pages to the memory 122 by reference to an extension page map table 113 of the computer 11, a request is given to the computer 12 to transfer the corresponding pages. Then the pages transferred form the computer 12 are stored in an idle memory block of the memory 112 of the computer 11. Thus the overhead caused by the replacement of pages can be shortened.

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial Application Field] The present invention provides a method for connecting multiple computers that perform virtual memory management using a paging method to a high-speed line network (for example, 1.5 Mbp
s l5DN (Integrated Service
Digital Network.

The present invention relates to a distributed paging control method for controlling a paging method in an information processing system connected via a rk) *).

(Prior art) - In an information processing system having a computer that performs virtual memory management using a paging method, a disk device installed in its own computer is used as a backing store, and this disk device and memory (main memory bag f ), a paging control method is adopted in which pages are replaced (swapped) between

Conventionally, in this type of paging control method, when pages are frequently replaced, pages are frequently replaced between the disk device and memory in a computer, and access to the disk device is required each time. Ta.

[Problem to be solved by the invention]

In the conventional paging control method described above, when pages are frequently replaced, pages are frequently replaced between the disk device and memory in a computer, and access to the disk device is required each time. A few + m5ec depending on mechanical access time (especially seek time)
The disadvantage is that it generates some overhead.

Measures are taken to reduce such overhead, such as using a disk device with a short access time, but as long as mechanical access is involved, there is a limit to reducing the overhead.

In view of the above points, an object of the present invention is to provide a distributed paging control method that is a paging control method that can shorten the time required for page replacement and reduce the overhead associated with page replacement. It is in.

[Means to solve the problem]

The distributed paging control method of the present invention is applied to an information processing system in which multiple computers, each of which manages virtual memory using a paging method, are connected via a high-speed line network. A memory block management table within a computer that manages information indicating the usage status of memory blocks within a computer, and information indicating memory blocks in which virtual storage pages are stored on a disk device provided in one computer. Free memory blocks are created in the memory of other computers based on the extended page map table in the managed 100-calculator and the memory block management table in the 100-calculator when evicting pages stored in the memory of the 100-calculator. If it is determined that the page exists, the page is controlled to be evicted from the memory in the other computer, and when it is determined that the page evicted from the memory in the other computer is requested based on the reference to the extended page map table in the computer. Distributed paging control unit within the 100-computer that requests other computers to transfer the page and controls the storage of the page transferred from the other computer in a free memory block of the memory within the 100-computer (the above-mentioned control In this case, a memory block transfer mechanism, etc. is used).

[Effect]

In the distributed paging control method of the present invention, a memory block management table within a computer manages information indicating the usage status of memory blocks in memories provided in one computer and other computers, and an extended page map within a computer. The table manages information indicating the memory blocks in which virtual storage pages are stored on the disk device provided in one computer, and the distributed paging control unit within the computer is stored in the memory within the computer. If it is determined that there is a free memory block in the memory of another computer based on the reference to the memory block management table in the 100-computer, the page will be controlled to be evicted from the memory of the other computer. If it is determined that the page ejected to the memory of another computer is requested based on the reference to the extended page map table of Controls storage in free memory blocks in the computer's memory.

〔Example〕

Next, the present invention will be explained in detail with reference to the drawings.

FIG. 1 is a block diagram showing the configuration of an embodiment of the distributed paging control method of the present invention.The distributed paging control method of this embodiment is a type of computer in which virtual memory management using a paging method is adopted. The present invention is applicable to an information processing system including a workstation,
A network 3 includes two workstations 11 and a workstation 2, and a 1,5Mtl SISDN network 13 that connects them.
It is.

Each workstation 11 and 12 has a C
PU (Central Processing U)
nit) 111 and 121 and memories 112 and 1
22, extended page map tables 113 and 123, memory block management tables 114 and 124, distributed paging control units 115 and 125, memory block transfer mechanisms 116 and 126, and disk devices 117 and 127. ing.

Memories 112 and 122 are each divided into a plurality of memory blocks of a single length (the same size as a page). Memory] The usage status of each memory block in 12 or 122 is determined by the distributed paging control unit 115 or 12.
5, the memory block management table 114 or 1
24 in real time.

FIG. 2 is a diagram showing an example of the structure of the extended page map table 113 or 123. The extended page map table 113 or 123 stores a page number 21 indicating a sequential number given to each page when the virtual storage area provided in the disk device 117 or 127 is divided into pages, and a certain page. The work stage address 22 indicates whether the memory block is within the workstation 11 or the workstation 12, the block number 23 indicates the number of the memory block in which the page is stored, and whether the page is located within the workstation 11 or 12. - The page not present bit 24 indicates whether the page exists in the memory in the local workstation (if the page does not exist in the memory in the local workstation, it is set to on, and if it does, it is set to off). It is configured.

FIG. 3 is a diagram showing an example of the configuration of the memory block management table 114 or 124. Memory block management table 11
4 or 124 is information for managing the memory 112 in the workstation 11 or the information for managing the memory 112 in the workstation 12
The workstation number 31 (the identification number of the workstation 11 or the identification number of the workstation 12 is set) to indicate whether the information is for managing the memory 122 in the memory 112 or 122, and the information for each memory block in the memory 112 or 122. It consists of a block number 32 indicating the number, and an empty bit 33 indicating whether each memory block is empty (unused status) (if not used, it is set to on).

FIGS. 4(a) to 4(e) show the distributed paging control unit 115.
or 125 is a flowchart showing an example of the processing. This process includes an evicted page determination step 401, an empty memory block presence/absence determination step 402, an evicted destination determination step 403, a page etc. transfer step 404, a table update step 405, and a page replacement step 406.
, free bit off step 411 , requested page storage position acquisition step 421 , free memory block presence/absence determination step 422 , page replacement step 423 , free memory block selection step 424 , transfer request step 425 , free focus on Step 431 and
It consists of a table update step 441.

Next, the operation of the distributed paging system of this embodiment configured as described above will be explained. In addition, here, the fourth
The operation when the processes shown in FIGS. (a) to (e) are performed will be described.

Now, when the memory 112 in the workstation 11 is full and there is a request from the CPU III for a page that does not exist on the memory 112, the distributed paging control unit 1
15 performs the following processing (see FIG. 4(a)).

First, LRU (Least Recently Us
(step 401). In this case, the page is stored in memory block #n (n is a positive integer). Page #
Assume that A (A is a positive integer).

Next, with reference to the memory block management table 114, it is determined whether there is a free memory block (unused memory block) in the memory 122 in the workstation 12, which is another workstation (step 402).
).

If it is determined in this step 402 that there are free memory blocks in the memory 122, one of them is allocated to page #
It is determined that A is to be evicted (step 403). In this case, the memory block to be evicted is memory block #m (m is a positive integer) in the memory 122.

Once the ejection destination of page #A is determined, the page #A stored in memory block #n in the memory 112 and the ejection destination (memory block #m in the memory 122)
is transferred to the workstation 12 via the 1.5 Mbps IsDN network 13 (page transfer is performed by controlling the memory block transfer mechanism 116) (step 404).

Additionally, the empty bit 33 for the block number 32 in the memory block management table 114 indicating the memory block #n in the memory 112 is updated to ON, and the page number 21 indicating the page #A in the extended page map table 113 is updated.
Workstation address for 22. The block number 23 and page absent bit 24 are updated to information indicating the workstation 12, information indicating the memory block #m, and ON (step 405).

Note that if it is determined in step 402 that there is no free memory block in the memory 122, the above processing (processing in steps 403 to 405) is not performed and the connection between the memory 112 and the disk device 117 is changed as in the conventional technology. The page is replaced (step 406).

When information indicating the page #A and the expulsion destination (memory block #m) is sent from the workstation 11 via the 1.5 Mbps ISDN line 13, the memory block transfer mechanism 126 in the workstation 12 transfers the page #A to the memory. 122 (stored under control from the distributed paging control unit 115).

At the same time, the distributed paging control unit 125 in the workstation 12 controls the free bit 33 for the block number 32 in the memory block management table 124 indicating the memory block #m in the memory 122, as shown in FIG. 4(b). is turned off (step 411).

As described above, page #A is
When there is a request for page #A from the CPU 11 after the page has been evicted to memory block #m of the memory 122 in the workstation 11, the distributed paging control unit 115 in the workstation 11 performs the following processing (fourth Figure (C
)reference).

First, with reference to the extended page map table 113, page #
Storage location of A (memory 12 in workstation 12
(Step 4)
21), this page #A is the disk device 117 in the workstation 11 which is the own workstation.
The processing when the data is stored in is similar to the processing in the conventional technology.

Next, with reference to the memory block management table 114, it is determined whether there is a free memory block in the memory 112 (
step 422).

If it is determined in this step 422 that there is no free memory block in the memory 112, the page is replaced in the same manner as the process shown in FIG. 1.5Mb for workstation 12 after creating memory block #1 (where 1 is 0 or a positive integer).
A request is made to transfer the contents of memory block #m (page #A) via the pslSDN network 13 (step 425).

Further, if it is determined in step 422 that there is a free memory block in the memory 112, one of the free memory blocks is selected.In this case, memory block #l
) (step 424), the 1.5 Mbps ISDN network 1 is selected for the workstation 12.
A request is made to transfer the contents of memory block #m via 3 (step 425).

When the transfer request in step 425 is made, the memory block transfer mechanism 126 in the workstation 12 transfers the contents of memory block #m of the memory 122 by 1.5 Mb.
It is transferred to the workstation 11 via the pslSDN network 13 (transferred based on control from the distributed paging control unit 125).

At the same time, the distributed paging control unit 125 in the workstation 12 controls the free bit 33 for the block number 32 in the memory block management table 124 indicating the memory block #m in the memory 122, as shown in FIG. 4(d). is turned on (step 431).

Contents of memory block #m (page #A) from workstation 12 via 1.5Mbpsl SDN line 13
When the data is sent, the memory block transfer mechanism 116 in the workstation 11 stores the contents in the memory block #l in the memory 112 (storing based on the control from the distributed paging control unit 115).

At the same time, the distributed paging control unit 115 in the workstation 11 controls the free bit 33 for the block number 32 in the memory block management table 114 indicating the memory block #l in the memory 112, as shown in FIG. 4(e). is updated to OFF, and the workstation address 22 .corresponding to page number 21 indicating page #A in the extended paging map table 113 is updated. Block number 2
3 and the page fault bit 24 are updated to information indicating the workstation 12, information indicating the memory block #l, and OFF (step 441), thereby causing the series of information shown in FIGS. 4(a) to (e) to be updated. Processing ends.

〔Effect of the invention〕

As explained above, the present invention eliminates the need for page replacement in an information processing system in which multiple computers (workstations, etc.), each of which manages virtual memory using a paging method, are connected via a high-speed line network. In this case, the old page (the page that should be evicted) should be evicted to the disk device in the own computer (transferred to a free memory block in the memory of another computer via a high-speed line network and stored, and the old page mentioned above should be evicted). By controlling the transfer of the old page from the memory block in the other computer to one computer when it is needed again (data transmission between computers (page transfer, etc.) via a high-speed line network) Compared to conventional technology, which always requires mechanical access time such as seek time when replacing a page, the time required to replace a page can be easily shortened. This has the effect of reducing the overhead associated with page replacement.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing the configuration of an embodiment of the present invention, FIG. 2 is a diagram showing an example of the configuration of the extended page map table in FIG. 1, and FIG. 3 is a memory block management diagram in FIG. 1. A diagram showing an example of the structure of a table. FIGS. 4(a) to 4(e) are flowcharts showing an example of the processing of the distributed paging control unit in FIG. 1. In the figure, 11.12...workstation, 13.----
-1,5MbpslSDN network, 21...Page number, 22...Workstation address, 23.
...Front number, 24...Page not present bit, 31...Workstation number, 32...
...Block number, 33... Empty focus, 111 121 ・CPU, 112.122 ・Memory, 113.123 ・Extended paging map table, 114.1
24.Memory block management table, 115.125.Distributed paging control unit, 116.126.Memory block transfer mechanism, 117, 127.Disk device.

Claims (1)

[Claims] In an information processing system in which a plurality of computers, each of which manages virtual memory using a paging method, are connected via a high-speed line network, memory in the memory provided in the own computer and other computers. A memory block management table within the local computer that manages information indicating block usage status, and a self-computing memory block management table that manages information indicating the memory blocks in which virtual storage pages are stored on the disk device installed in the local computer. Determines that free memory blocks exist in the memory of other computers based on the extended page map table in the computer and the memory block management table in the computer when evicting pages stored in the memory of the computer. control to eject the page to the memory in the other computer, and if it is determined that the page ejected to the memory in the other computer is requested based on the reference to the extended page map table in the own computer, the other computer and a distributed paging control unit within the own computer that requests the transfer of the page from the other computer and controls the storage of the page transferred from the other computer in a free memory block of the memory within the own computer. A distributed paging control method.