JPH08339341A - Selection method for main storage page - Google Patents

Abstract

PURPOSE: To shorten the time needed for a page acquisition processing, etc., when an input/output buffer is assigned to a disk storage that performs the fast transfer of data as for a computer which contains the disk storage of a different type from a main storage composed of pages of different sizes. CONSTITUTION: At the point of time a disk array device 111 or an ordinary disk storage 20 are enabled to transfer the data requested by an OS, when requests an OS the OS is requested to acquire an input/output buffer, the OS uses with preference the page of a large size included in its resource area 302BR when the device 20 that issued the buffer acquisition request performs the fast transfer of data and also the size of the requested buffer is larger than the large page. In other cases, the OS uses the small pages of the area 302BR to secure the requested buffer.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention should allocate an area to a buffer used by an operating system (OS) for inputting / outputting data to / from an external storage device when the main storage is composed of pages of different sizes. Regarding how to select a page.

[0002]

2. Description of the Related Art A general computer system comprises a processor and a secondary storage device. The secondary storage device that is mainly used is a magnetic disk device.

At present, the capacity growth rate of the disk storage device is extremely high, but the performance of the magnetic disk device accompanied by mechanical operation is not as high as the processor performance growth rate. A disk array has been proposed as a method for solving the problem. "A by D. Patterson, G. Gibson, and RH Kartz
Case for Redundant Arrays of Inexpensive Disks (RA
ID), in ACM SIGMOD Conference, Chicago, IL ", PP.
109-116 (June 1988) (hereinafter referred to as the first reference)
In RAID, the data is distributed to a plurality of disk drives to shorten the access time to the data stored in the disk, and the redundant data called parity or ECC is also stored to improve reliability. Disk array configuration technology is introduced. In other words, array disks perform parallel I / O to multiple disk drives, so reading or writing data is faster, and even when a disk drive fails, parity and data other than the failed disk drive are used. Can recover failed disk drive data.

More specifically, in the above literature, the RAID levels are classified into a plurality of RAID levels according to the data storage method.
Among them, the RAID level that is often used in products is R
AID1, RAID3, and RAID5. RAID1
Is mirroring, and the reliability of a disk failure is improved by storing the same data in two disks. When reading, the speed is higher than that of a single disk by reading from whichever of the two disks is faster. RAID 3 can improve data transfer performance by striping single data bit by bit or byte by byte. RAID 5 is a level at which striping is performed on a plurality of disks in units of input / output blocks. RAID 3 strips single I / O requests in small units, whereas RAID 5 strips in larger units. The main purpose of RAID3 is to speed up I / O for a single user.
The main purpose of AID5 is to execute input / output of multiple users in parallel. Therefore, RAID 3 is used for multimedia and scientific / technical calculations that require particularly large-scale data input / output. RAID 5 is used for online transaction database processing that provides services for many users.

On the other hand, the OS has a virtual memory function.
The virtual memory function temporarily outputs data and programs in the main memory that are not needed for the time being to an external large-capacity disk device, and immediately uses them in a program that requires memory (main memory). This is a function that makes it look like a large capacity memory. This operation is called paging or swapping. The OS acquires a virtual storage area and a main storage area of a required capacity when inputting / outputting data to / from a disk device or a disk array device. The unit of this acquisition is called a page. In other words, the main memory is composed of a plurality of unit areas called pages. The paging and swapping are performed in units of this page. It is necessary to acquire multiple pages when inputting / outputting a large amount of data that exceeds the page size. Further, the OS also performs the page fixing process when the disk is input / output. The page fixing is a process of preventing the main memory allocated to the virtual memory from being stripped by another program until the input / output ends. The paging and swapping are operations performed independently of disk input / output. If you do not fix the page,
There is a problem that the area is allocated to another program while transferring the data to the main memory. In order to avoid this, a page fixed bit is provided in the main memory management table managed by the OS, and O is set while this bit is set.
S has a mechanism to prevent the page from being stripped.
This page fixing is also managed for each page, so when a large number of pages are newly assigned to any program,
Fixed processing across these pages is required.

[A Simulating Ba by J. Bradley Chen et al.
sed Study of TLB Performance ", 19th International S
ymposium on COMPUTER ARCHTECTURE, PP. 114-123 (199
In 2) (hereinafter referred to as the second reference), when the user allocates a memory area used as a work area, the page size managed by the TLB of the processor is made variable, and a plurality of applications are accessed from the processor. The results of simulating instruction and data hit rates are described. It is stated that when the page size is changed, the instruction hit ratio of the processor changes, and the optimum page size exists depending on the application.

It is described that the processing performance of the processor changes depending on whether or not the cache is hit when the instruction or data of the application is referred to. This change is considered from the page size. Different applications have different reference characteristics for instructions and data. The larger the page size is, the higher the processor processing performance becomes, as the instruction and the data access are sequential with respect to the memory address and refer to the large-capacity area. This is because the processor always converts the virtual memory address to the real memory address when referring to / updating the instruction and the data referred to by the instruction, so that the larger the page size is, the smaller the number of times of this conversion processing is required.

[0008]

However, current commercial computers use a memory composed of a single page size, for example, a 4 KB page size. Therefore, considering the processing speed from the matters described in the above-mentioned second reference, it is desirable to use the main memory having a page of a size already used in the past, for example, a size larger than 4 KB, for example, 16 KB. But,
Among the programs running on such a computer, there are programs that normally operate only on a computer having an original size of 4 KB pages. Therefore, in order to introduce a page of a large size into a computer, in addition to a page of a size already used, for example, 4 KB, a page of a larger size, for example, 16 KB, is mixed in the main memory. It is desirable to configure.

In the prior art, data area management at the time of input / output of a disk is performed by, for example, "design and implementation of UNIX 4.3BSD",
Pages 198 to 200, Maruzen Co., Ltd., 1991 (hereinafter referred to as the third reference). Here, for example, when a user reads data from a disc, first, a user area for storing the data is secured. After that, when a disk input request is issued to the OS, the OS issues an input request to the disk device. When the disk reads the requested data, an interrupt indicating the completion of the data reading from the disk device is generated to the OS controlling the host that issued the request. A buffer for temporarily storing signal data is acquired inside the OS. This buffer is secured in a part of the memory area managed by the OS. The memory area managed by the OS is also called a resource area, and a certain amount is secured when the system is activated, and the memory area is secured or released according to the necessity of processing. Upon this interruption, the disk storage device notifies the OS of the total amount of read data,
The S determines the size of the buffer secured in the main memory according to the notified data amount.

When a normal disk storage device and an array disk device are connected to a computer using a main memory having pages of different sizes, there are the following problems.

As mentioned above, different sizes, eg 4
In a computer having a main memory including pages of KB and 16 KB, as pages forming this buffer area,
When the size of this buffer is large, specifically, for example, 16 KB or more, the smaller the total number of pages constituting this buffer, the shorter the time required for page securing processing and page fixing processing for those pages. Expected to be small. Therefore, when the size of the buffer required by any of the disk storage devices is larger than the page size of 16 KB or larger, it is desirable to allocate the larger page or page of 16 KB to this buffer area. However, since the total number of 16 KB pages is limited, unlimited allocation of 16 KB pages to data transfer from any disk storage device would result in 1
6KB pages are expected to run out. That is,
As described on page 198 to page 200 of the third reference, the input / output buffer area secured by the user and the input / output buffer temporarily secured in the resource area by the OS at the time of input / output of the disk or the like Is not the same as. As described above, the resource area is a memory area managed by the system and shared by requests from many users. Therefore, it is desirable to select the size of the page allocated in the buffer requested by the disk storage device so that this resource can be effectively used.

That is, the data transfer rate of an external device such as a disk varies greatly from several megabytes / second to several hundred megabytes / second depending on the apparatus. For a buffer of the same size, the lower the data transfer rate of the disk storage device that uses it, the longer the buffer is occupied. As a result, conversely, it becomes impossible to allocate a large page to the buffer requested by the disk storage device having a high transfer rate, and as a result, a smaller size page is allocated.

As a result, a large number of Y pages are allocated to the disk storage device, which is originally expected to transfer data at high speed, and the time required to secure these pages and fix the pages increases. .

[0014]

According to the present invention, of a plurality of disk storage devices connected to a main memory, a buffer requested by a disk storage device having a relatively high data transfer rate has a predetermined size. When the value is, for example, the maximum page size or more, the page of a relatively large size is preferentially used, and the buffer requested by the disk storage device having a relatively low data transfer rate has a size of a predetermined value or more. Even in case of, preferentially use the smaller size page.

When the requested buffer size is smaller than the above-mentioned predetermined value, for example, the maximum page size, the smallest size page is used for the buffer regardless of the buffer size.

[0016]

When a large-scale data is input / output between a device capable of high-speed data transfer, such as a disk array device, and a main storage device, a large page size area is used to realize the high-speed data transfer. It is possible to reduce the time required for processing such as area reservation processing and area release processing, which is an overhead, and the characteristic of high-speed data transfer is not impaired.

[0017]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The main memory page selection method according to the present invention will be described in detail below with reference to the embodiments shown in the drawings.

FIG. 1 is a schematic block diagram of a computer system to which the main memory page selection method according to the present invention is applied. 10 is a CPU, 101 is a main storage device, 111 is a disk array device, and is composed of a plurality of disk devices 112. 20 is a conventional, single drive, disk storage device. In this embodiment, the disk array device 111 is a disk storage device that can transfer data at high speed, and 20 is a disk storage device that transfers data at lower speed.

A main program used in this embodiment is shown in the main memory 101 of the figure. These programs include the user program 102 and the OS 209.
The OS further includes a virtual memory management program 103, a main memory management program 104, and an input / output management program. The user program 102 indicates a program other than the OS and issues an input / output request to the secondary storage device. The virtual memory management program 103 and the main memory management program 104 are both necessary programs for realizing the virtual memory function. The virtual memory management program 103 reserves / releases a virtual memory area in response to a memory reservation request from the user program 102. The main memory management program 104 secures / maintains a main memory which is hardware /
Open. The input / output management program 110 is a program for controlling input / output of a secondary storage device such as a disk device.

The input / output area management program 106 in the virtual memory management program 103 is a user program 102.
Is issued when a virtual storage area acquisition request is issued to issue a disk array I / O request. Page size selection program 108 in main memory management program 103
Is a system having a plurality of main memory page sizes, and when high-speed input / output is required like the disk array device 111, a main memory area having a large page size is allocated, and otherwise, a small page size is used. Performs an operation to allocate a main memory area.

Generally, there are a plurality of types of disk arrays depending on the storage method of data and parity, but in the present embodiment, any of them can be used. The main types of them are: RAID 1 is mirroring, and the same data is stored in two disks to improve reliability against disk failure. When reading, the speed is higher than that of a single disk by reading from whichever of the two disks is faster. RAID
In No. 3, by striping single data bit by bit or byte by byte, the data transfer performance can be improved.
RAID 5 is a level at which striping is performed on a plurality of disks in units of input / output blocks. RAID 3 strips single I / O requests in small units, whereas RAID 5 strips in larger units. The main purpose of RAID 3 is to speed up the input / output of a single user, while the main purpose of RAID 5 is to execute the input / output of a plurality of users in parallel. Therefore, the RAID 3 is used for multimedia processing and scientific / technical calculation for processing images and the like which require particularly large-scale data input / output. RAID 5 is used for online transaction database processing that provides services for many users. Among the above, the type in which this embodiment is particularly effective is RAID3. RAID 3 has a feature that it is easy to speed up the transfer performance among the disk array types because RAID 3 always strips a single data to a plurality of disk devices.

(Outline of Allocation of Virtual Memory and Real Memory) FIGS. 2 and 3 show page structures of the virtual memory area 301 and the main memory area 101. FIG. 2 shows a page structure to the virtual storage area 301 and the main storage area 101 when inputting / outputting from the disk array device 111. Further, FIG. 3 shows a virtual storage area 30 when inputting / outputting from the disk storage device 20 which transfers data at a lower speed than the disk array device 111.
1 and the page structure to the main storage area 101.

In this embodiment, the virtual storage area 301 is divided into a plurality of small pages of the same size. In individual terms, the size of this page is 4KB. Main storage area 1
01 is divided into a small page having this size and a large page having a size larger than these, for example, 16 KB. The reason why the main storage area 101 has a large page is that the use of a large page can reduce the total number of pages to be managed by the OS. The main storage area 101 includes a small page in order to allow a user program using a conventional small page to be executed by this computer system. The virtual storage area 301 is divided into an area 301A that can be used by the user program and an area 301B that can be used by the OS. Similarly, the main storage area 101 is also divided into an area 302A that can be used by the user program and an area 302B that can be used by the OS.

2 and 3, reference numerals 307 and 308 denote two input / output buffer areas provided in the virtual storage area 301 and requested by the user program. In this embodiment,
According to the request from the user program, a buffer area of the size designated by the request is secured. Also, 3
Reference numerals 03 and 304 denote input / output buffer areas on the main storage area allocated to these two virtual input / output buffer areas. These buffer areas are formed in the area 302A released to the user program in the main storage area. In this embodiment, the virtual input / output buffer 307
Is smaller than the actual large page size of 16 KB,
Assuming that the size is 8 Kbytes, for example, a real I / O buffer area 303 including two real 4 KB pages is allocated to the virtual I / O buffer 307. Assuming that the virtual input / output buffer 308 has a size of a real large page of 16 KB or more, for example, 64 KB, the virtual input / output buffer 30.
A real input / output buffer area 303 composed of four real 16 KB pages is allocated to the area 8.

As described above, in this embodiment, when the size of the virtual input / output buffer requested by the user is large, an actual input / output buffer area consisting of a large page is allocated. In this way, in this embodiment, when the size of the buffer requested by the user program is large, the number of pages forming the actual buffer area allocated to it is reduced. Virtual storage area 3
01 and the main storage area 302 are secured / released in units called pages. Therefore, in order to secure a large-scale area, it is necessary to perform a page acquisition process of securing the number of bytes / page size (times). Therefore, the smaller the number of pages, the shorter the time required to acquire or fix the pages forming the input / output buffer requested by the user program.

Another input / output buffer 3 shown in FIGS.
09A, 309B, 310A, 310B perform the input / output operation requested by the user program.
11 or a normal disk storage device 20 executes,
An example of an input / output buffer area secured by the OS when a request is made to the OS when data can be transferred to the OS is shown.
These buffers are the actual input / output buffers 30 in which the data read from the disk storage device by the data requested by the user program are assigned to the user program.
OS temporarily before transferring data to 3 or 304
Used for storage in 209. The buffers 309A, 309B or 310A, 3 in these OSs
10B is the resource area 30 in the OS usage area 302B.
Reserved to 2BR.

The input / output buffers 310A and 310B are secured by the OS when the array disk storage device 111 executes the input / output operation requested by the user program and then when the data can be transferred to the OS when the OS requests the OS. did,
An example of an input / output buffer area managed by the OS is shown. In the present embodiment, when the OS secures the buffer requested by the disk storage device 111 or 20, when the disk storage device is an array disk device capable of high-speed data transfer, the OS requests the disk storage device. It is determined whether or not the size of the buffer is equal to or larger than the size of the large page, and if the requested size is equal to or larger than the size of the large page, the large page is preferentially used for the buffer (I / O buffer in FIG. 2). 310B). If the requested size is smaller than the large page size, the small page is preferentially used (I / O buffer 310 in FIG. 2).
A). On the other hand, the disk storage device that requested the buffer
When the disk device is a normal disk device that cannot transfer data at high speed, the small page is preferentially used for the buffer regardless of whether the buffer requested by the disk storage device is larger than the size of the large page (( (Figure 3 I / O buffers 309A, 309B).

In general, since the buffer area 310 configured by 16 KB pages has a smaller number of pages than the buffer area 309 configured by using 4 KB pages, the time required for the acquisition processing and the fixed processing of those pages is small. I'm done. In this embodiment, these large pages are preferentially used in the buffer area requested by the array disk capable of executing high-speed data transfer. As a result, the time taken to start high-speed data transfer by the array disk device is prevented from increasing. Originally, the resource area 302BR is a memory area secured when the system is started up, and there is only one for the system, and a plurality of users and input / output requests share and use this area. . Therefore, it is necessary to use a certain amount of memory efficiently. Since this area is smaller than the main storage area 302A available to the user, the number of large pages is also finite. Therefore, in this embodiment, the disk storage device to which the large page is allocated is limited so that such a limited large page is not occupied for a long time due to low-speed data transfer. This prevents the data transfer rate of the entire system from decreasing.

(Details of Input / Output Processing) FIG. 4 shows an example of an operation of inputting data from the disk array device 111 or the normal disk storage device 20 of the user program 201.

In step 202, before reading the file 207A in the normal disk storage device 20 or the file 207B in the disk array device 111, the OS is requested to acquire the input / output buffer area on the virtual storage. In response to this request, the virtual memory allocation program 105 in the OS causes the input / output management program 106
However, for example, the area 307 or 308 shown in FIG. 2 is acquired.

FIG. 5 shows the virtual storage area management list 407. This list shows the virtual memory management program 1
03 is used to allocate virtual pages to the buffer requested by the user program. This list includes a list 401 showing unused areas and a list 40 showing areas in use.
It consists of two. The list has one data structure per page. The data structure is a pointer 403 to the next page, a page number 404, and a page size designation 40.
6 and the like are stored. The data structure of the list 401 showing the unused area and the list 402 showing the area in use is the same.

The page size designation 406 is performed in the main storage area 3
Specify the page size to allocate 02. The main memory allocation program 107 can determine which page size is allocated by referring to this field. When a new virtual storage area is secured, it is realized by removing the requested page from the unused area list and adding it to the used area list.

Returning to FIG. 4, in step 203, a file open command is issued to the OS in preparation for reading the file 207. O for this open command
S performs a file open process, and the OS notifies the user process 102 of information such as where the file 207A or 207B is normally stored in the disk storage device 20 or the disk array device 111.

In step 204, the file 20 is actually
A read request for storing 7A or 207B in the buffer area 303 or 304 acquired in step 202 is issued to the OS. At this time, in response to this request, the OS issues an input / output command for instructing execution of the read operation requested by the file request to the disk array device 111.
Alternatively, it is issued to the normal disk storage device 20.

The array disk storage device 111 or the normal disk storage device 20 that executed this command
When the requested data can be transferred to the OS, O
A data transfer request interrupt is sent to S to request buffer acquisition. The OS temporarily acquires the buffer 310 or 309 in the resource area 302BR in the OS by a method described in detail later. Thus, the array disk storage device 111 or the normal disk storage device 20 can transfer data to the acquired buffer 310 or 309. When the data is stored in the buffer 310 or 309, the OS displays the area 307 or 30 acquired by the user.
Data transfer to 8 is started. However, at this point in time, a page fault occurs because the main storage area of the area 307 or 308 has not been acquired, and the main storage allocation program 107 is requested to acquire the main storage area, and the main storage area is acquired by the method described in detail later. With the above operation, the user can refer to the data in the array disk storage device 111 or the normal disk storage device 20.

In step 205, the resources acquired for the disk input / output by the file close are released.

(Details of Selection of Page Size for Buffer Area in OS) In this embodiment, the page used in the buffer when the OS acquires this buffer 309 or 310 after step 204 described above. Is selected according to whether the device that has read the data is an array disk device that performs high-speed data transfer or an ordinary disk storage device that does not. The details will be described below with reference to FIG.

FIG. 9 shows the main memory allocation program 107.
The processing flow of is shown.

The main memory allocation program 107 is activated by the data transfer request interrupt 701 or page fault 710 issued from the disk storage device after step 204 described above.

In step 702, main memory area allocation acceptance / analysis processing triggered by this interrupt is performed.
Then, the page size selection program 108 is activated.

In this program 108, step 70
At 9, it is determined whether or not this interrupt requests the reservation of the resource area.

The issuer of this interrupt can be analyzed based on the disk device identifier (system ID) stored as the status at the time of the data transfer request interrupt. When the issuer of this interrupt is the disk storage device, it is determined that this interrupt requires the reservation of the resource area.

As a result, when this interrupt requests the reservation of the resource area, in step 703 it is judged whether the disk storage device which issued this interrupt is a device capable of high speed transfer. This judgment is made based on the identifier of the interrupt issuing source and the table of the secondary storage device managed by the OS shown in FIG. This table contains the system I assigned to the various disk storage devices in the system.
The type 902 and the data transfer rate 903 of the device are stored corresponding to D (901). Step 7
In 03, specifically, it is determined whether the data transfer rate of the interrupt issuing source is a predetermined value, for example, 10 or more,
If it is 10 or more, it is determined to be a high-speed transfer device. Instead of this determination method, it may be determined whether the interrupt source is an array disk storage device based on the disk storage device type 902, and if so, a high speed transfer device determination method. When the interrupt source is a device capable of high-speed transfer, the process proceeds to step 708. Otherwise, go to step 706.

In step 708, it is further determined whether the allocation request has a large capacity. This determination is made based on the data transfer amount specified by the disk storage device that is the source of the interrupt request in association with the interrupt request. In this embodiment, the data transfer amount is a predetermined value. For example, when the page size is larger than the large page size, it is determined that the large capacity transfer request is made.

If it is determined that the input / output request has a large capacity, the process proceeds to step 704. Otherwise, step 7
Proceed to 06. Step 704 is executed at the time of input / output to / from the disk array, and searches a large page size free list 503 (FIG. 7) described later in order to secure a large page size in the main storage area. If there is a space here, the process proceeds to step 705 and the in-use list 50 to be described later.
4 (FIG. 7), it is assigned for a device capable of high-speed transfer. In a system having two or more page sizes, the effect of this embodiment can be obtained by selecting a page other than the smallest page size.

In step 706, basically, a small page size empty list 501 (FIG. 7) described later is searched. If a small page size vacancy is found, it is added to the in-use list 502 (FIG. 7) described later to secure the area.
By this processing, a large page size is preferentially allocated at the time of input / output to / from a disk array device or the like capable of high-speed transfer, so that the load of the page allocation processing can be reduced. Further, even if a large-capacity input / output request is made, a small page size is allocated to a device having low performance, so that the resource area can be used efficiently.

In step 711, it is determined whether the request is a main memory allocation request due to a page fault. This determination can be made by referring to the interrupt status, as in the resource area determination. In step 712, it is determined whether to allocate the main memory to the large page size or the small page size. This determination can be made by referring to the free virtual storage area management list in FIG. The pointer 402 in FIG. 5 indicates a list of pages from which virtual memory is acquired. From the page fault interrupt status, it can be determined which virtual memory page caused the page fault. Based on this information, the virtual memory list is searched from the pointer 402 to find the page in which the page fault has occurred. In the list, there is a field 406 describing page surge, and it is described whether a large page size or a small page size is assigned to this field. When a virtual storage area is newly allocated to this field, the size of the virtual storage area is determined, and if the area is larger than a predetermined value, a large page size allocation request is stored. Otherwise, a small page size allocation request is stored. To store. That is, since the main memory area is acquired after the virtual memory area, when allocating the virtual memory area, the virtual memory allocation program 105 sets the main memory page size to be allocated in advance in the field 4
It is stored in 06. By referring to this field, the judgment in step 712 in FIG. 9 can be made. As a result, if it is a large page size acquisition request, step 704.
If it is a small page size acquisition request, step 706.
To execute. By this processing, depending on the capacity of the input / output buffer (input / output buffers 303 and 304 in FIG. 4) acquired by the user, it can be allocated to the large page size if it is larger than a predetermined value and to the main storage area of small page size if it is smaller. It will be possible. Details of main memory page management will be described later.

(Management of Main Memory Page) FIG. 6 shows a main memory management list group 1105. Pointer list 1
100 includes pointers to two main memory management lists.
One is a pointer 1101 to a user area main memory management list 1103 that can be secured by the user, and the other is a pointer 1102 to a resource area list 1104. The resource area is an area which is not directly allocated by the user but is allocated by the OS when it is necessary for system operation. 1101 and 1102 are OS main memory allocation /
When releasing, switch according to the application.

FIG. 7 shows the main storage area management list 1103 that can be assigned by the user. 501 is a pointer to a free list in the small page size area, 502
Stores a pointer to the in-use list of the small page size area. Reference numeral 503 stores a pointer to a free list in the large page size area, and reference numeral 504 stores a pointer to a busy list in the large page size area. The list has one data structure per page. The list data structure stores a pointer 505 to the next page, a page fixed bit 506, a page number 507, and the like. The data structures of the lists 501 and 503 indicating unused areas and the lists 502 and 504 indicating areas in use are the same. When a new main storage area is to be reserved, the unused area list 5
This is realized by removing the requested pages from 01 and 503 from the list and adding them to the in-use area lists 502 and 504. In the busy lists 502 and 504, the OS does not strip a page for which the page fixing bit 506 is set.

FIG. 8 shows the main storage area management list 1104 of the resource area. The structure is the same as in FIG.
Reference numeral 1201 stores a pointer to a free list of the small page size area, and 1202 stores a pointer to a busy list of the small page size area. Reference numeral 1203 stores a pointer to a free list in the large page size area, and 1204 stores a pointer to a busy list in the large page size area. The list has one data structure per page. List data structure is pointer 1 to the next page
205, page fixed bit 1206, page number 120
7 etc. are stored. A list 1201 showing unused areas,
1203 and lists 1202, 12 showing areas in use
The data structure of 04 is the same. To newly secure the main storage area, the requested pages are removed from the unused area lists 1201 and 1203, and the used area list 1
It is realized by adding to 202 and 1204. Page fixed bit 12 in the busy list 1202, 1204
The OS does not strip the page on which 06 is set.

FIG. 10 shows a flow of page fixing processing. The page fixing prevents the OS from stripping pages in the main storage area during data transfer to a secondary storage device such as a disk device. In step 801,
Search the page that should be fixed. Step 8
In 02, the page fixed bit 506 of the in-use list of the main storage area is turned on. The page fixing is completed by performing this operation for the acquired buffer area.

[0052]

According to the present invention, software control for improving the data transfer rate of a high speed disk array device becomes possible.

[Brief description of drawings]

FIG. 1 is a schematic configuration diagram of a computer system that executes page selection processing according to the present invention.

FIG. 2 is a diagram showing a relationship between a virtual storage area and a main storage area.

FIG. 3 is a diagram showing a relationship between a virtual storage area and a main storage area.

FIG. 4 is a flowchart of a user program.

FIG. 5 is a diagram showing a virtual storage area management list.

FIG. 6 is a diagram showing a pointer to a main storage area management list.

FIG. 7 is a diagram showing a user area main memory management list.

FIG. 8 is a diagram showing a management list of a main storage area of a resource area.

FIG. 9 is a flowchart of a page size selection program of the main memory management program.

FIG. 10 is a flowchart of a page fixing process for the main storage area.

FIG. 11 is a diagram showing a secondary storage device management table managed by the OS.

Claims (3)

[Claims] 1. A main storage device comprising a plurality of pages having different sizes, and a plurality of secondary storage devices, responding to a data transfer request requested from any one of the secondary storage devices. , A buffer area used for data transfer between the secondary storage device and the main storage device is secured in the main storage, and the secondary storage device of any one of the above and the main storage device is secured via the secured buffer region. In a computer system that transfers data to and from a main storage device, when the transfer speed of any of the above secondary storage devices is relatively high, the size of the plurality of pages constituting the main storage device is relatively large. Large empty pages are used in the buffer area in preference to relatively small empty pages, and the main storage device is configured when the transfer speed of any of the above secondary storage devices is relatively low. Compound Of the page, a small page of relatively size, main memory page selection method used in the buffer region in preference to large free pages relatively size. 2. When the transfer speed of any one of the secondary storage devices is relatively high and the size of the buffer area of each of the secondary storage devices is smaller than a predetermined value, the main storage device is configured. Of the plurality of pages to be used, a free page having a relatively small size is used for the buffer area, the transfer speed of any one of the secondary storage devices is relatively high, and the size of the buffer area is larger than a predetermined value. When the size is small, among the plurality of pages forming the main storage device, a relatively small-sized empty page is used in the buffer area in preference to a relatively large-sized empty page. When the transfer rate of the secondary storage device is relatively low, the secondary storage device does not depend on whether or not the size of the buffer area is equal to or larger than a predetermined value, and a plurality of main storage devices are configured. Within the page Method of selecting the main memory page of claim 1 wherein the use of small pages relatively size in the buffer area. 3. The main memory page selecting method according to claim 2, wherein the predetermined value is equal to a maximum page size.