JP2943693B2 - Sort work file space management method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a sort processing method for sorting a large amount of data which cannot be stored in a main storage device by using a work file, and more particularly to a sort work file space management for performing a sort processing at high speed. About the method.

[0002]

2. Description of the Related Art Conventionally, as an internal processing procedure of a sort program for performing a sort process using a work file on a magnetic disk drive, sorting is performed while inputting records, and a sorted record sequence called a "string" is processed. In general, a pre-sort phase for writing a plurality of files to a file and a collation phase for merging each string are performed separately.

In the collation phase, when the result of string merging is output to the work file again, an area (hereinafter, also referred to as an "available sector") which has been completed and has become reusable is used. This eliminates the need to secure a new work file space on the magnetic disk device.

However, there is no guarantee that such an available sector is secured in a continuous area on the magnetic disk drive. Therefore, especially when the data amount is large and merging is repeated many times, the magnetic disk drive is not used. Movement of the head frequently occurs, which may cause performance degradation.

Therefore, in order to prevent a decrease in performance when using an available sector, for example, Japanese Patent Laid-Open Publication No.
Japanese Patent Application Laid-Open No. 5604 proposes a system that can improve the performance by using the last available sector to reduce the moving distance of the head of the magnetic disk drive.

[0006]

However, in the conventional method described in Japanese Patent Laid-Open No. 5-165604 described above, when output processing to a work file is performed immediately after input processing from a work file, It is effective, but if the available sector is created in a distant area instead of a continuous area, it takes time to move,
There is a problem that it is not effective when output processing is performed continuously.

In the conventional method described in Japanese Patent Application Laid-Open No. 5-165604, even if output processing to a work file is performed immediately after input processing from a work file, the processing is performed immediately before the processing. If the input / output processing of the magnetic disk device was performed by the input / output processing of another process, the head of the magnetic disk device was moved by the input / output processing of the other process, and the output was performed immediately after the input processing. Even when the processing is performed, the head must be moved again, and there is a problem that the performance is deteriorated.

SUMMARY OF THE INVENTION Accordingly, the present invention has been made in view of the above-mentioned problems, and has been made by minimizing the moving distance of the head of a magnetic disk drive and shortening the time required for input / output processing of a work file. Another object of the present invention is to provide a sort work file space management method capable of improving the performance by speeding up the sorting of a large amount of data using a work file.

[0009]

SUMMARY OF THE INVENTION In order to achieve the above-mentioned object, the present invention provides a sort processing method for sorting a large amount of data that cannot be stored in a main storage device by using a work file. A work file for temporarily storing the result, a work file output means for outputting a sorted record string which is an intermediate result of the sorting process from the main storage device to the work file, and a work file output means for outputting the work file to the work file. A work file input means for reading the sorted record sequence into the main storage device again to merge the sorted record sequence, and a sorted record sequence output on the work file being input to the main storage device. A table for managing an area on the work file that has become reusable, wherein the work file input means controls the area. Each time a read-out record sequence is read into the main storage device, an available sector management table for registering the position of the read area and the sorted record sequence output on the work file are merged, and the merge result is re-read. An available sector selection unit that is called by the work file output unit when outputting to the work file, selects an area on the work file to be output next from the available sector management table, and notifies the work file output unit of the area; And a cylinder position obtaining means for obtaining the current cylinder position of the head of the magnetic disk drive, wherein the work file input means performs input / output each time the sorted record sequence is read into the main storage device. After the input processing is performed, the cylinder position The cylinder position of the head of the magnetic disk drive obtained by the obtaining unit is registered in the available sector management table, and the available sector selecting unit selects the cylinder position when selecting an area on the work file to be output next. The current cylinder position of the head of the magnetic disk drive is obtained by the obtaining unit, and the obtained current cylinder position is compared with the cylinder position of each area registered in the available sector management table, and the closest A sorting work file space management method characterized by selecting an area at a cylinder position.

According to the present invention with the above-described configuration,
Since the available sector selection means selects the available sector in which the moving distance of the head of the magnetic disk device is the minimum as the output destination, the output of the magnetic disk device is continuously performed, or the head of the magnetic disk device is moved by another process. Even in the case where the magnetic disk drive has been lost, the time required for moving the head of the magnetic disk drive can be minimized.

[0011]

Next, embodiments of the present invention will be described in detail with reference to the drawings.

FIG. 1 is a block diagram showing a configuration of an embodiment of the present invention. Referring to FIG. 1, the embodiment of the present invention includes a sort processing unit 1, a main storage device 2, and a magnetic disk device 3.

The sort processing means 1 includes an available sector selection means 11, a work file output means 12, a work file input means 13, a cylinder position acquisition means 14,
have. The main storage device 2 has an available sector management table 21, and the magnetic disk device 3
It has a work file 31.

The sort processing means 1 performs output processing to the work file 31 on the magnetic disk drive 3 by the work file output means 12 in the collation phase of the sort processing, and performs the output processing on the magnetic disk drive 3 by the work file input means 13. The input process from the work file 31 is performed.

FIG. 2 is a diagram showing a configuration of the available sector management table 21 shown in FIG.

Referring to FIG. 2, the available sector management table 21 includes a file descriptor 211, which is positional information necessary for accessing a specific area in a file on the magnetic disk device 3, and an offset address in the file. 212 as a component, and also includes a physical cylinder position 213 in the magnetic disk drive 3 as a component.

Here, the "file descriptor" is a number for specifying a file on which the program performs input / output processing with respect to the operating system, and is notified from the operating system to the program when the file is opened. The “offset address in the file” indicates a position from the beginning of the file. Further, the "physical cylinder position" is used as information for the available sector selecting means 11 to select an area in which the moving distance of the head of the magnetic disk drive 3 is minimized when outputting to the work file 31. It is. That is, when an output process is about to be performed, an available sector having a cylinder position closest to the current cylinder position of the head of the magnetic disk drive 3 is selected as an output destination area.

As a method for selecting an available sector, for example, there is a method according to the following procedure. That is, first, at the time of input processing from the work file 31, the cylinder position in the input request to the magnetic disk device 3 issued at that time is stored in the available sector management table 21. At the time of a request for output processing to the work file 31, first, the current cylinder position of the head of the magnetic disk drive 3 is obtained, and the available sector of each of the first and second entries in the available sector management table 21 is obtained. The difference between the cylinder position and the current cylinder position of the head is compared, and the one with the smaller difference is selected. Then, the cylinder position of the selected entry,
The cylinder position of the available sector of the third entry in the available sector management table 21, the cylinder position of the available sector of the fourth entry, and so on are repeated in order, and the entry with the smaller difference is selected. The available sector of the entry that minimizes the difference between the cylinder position of the head of the magnetic disk drive 3 and the cylinder position of the available sector is selected. If an entry with a cylinder position difference of “0” is found during the search, that is, if an available sector in the same cylinder is found, the search is stopped at that time and the available sector of the entry is selected. I do.

FIG. 3 is a flowchart for explaining the sort processing in the embodiment of the present invention.

The operation of the embodiment of the present invention will be described below with reference to FIG. However, in FIG. 3, the sort start processing, the presort processing (step 310), and the sort end processing are parts that are not directly related to the present invention.
The description here is omitted.

In FIG. 3, steps 322, 3
23, 325 and 326 are executed by the work file input means 13 and steps 333, 334 and 33 are executed.
7 and 338 are executed by the work file output unit 12, the process of step 336 is executed by the available sector selection unit 11, and the processes of steps 324 and 335 are executed by the cylinder position acquisition unit 14.

Referring to FIG. 3, the presort processing (step 310) is completed, and the collation processing (step 3) is completed.
After the start of step 20), when a record input request is made (step 321), the record is fetched from the work file buffer on the main storage device 2 (step 32).
6). However, if there is no record in the work file buffer (step 322), the work file input means 1
3 inputs one block from the work file 31 to the work file buffer (step 323), and the cylinder position obtaining means 14 obtains the cylinder position of the area 32 in the work file containing the input one block (step 323). 324), the available sector management table 21 using the obtained cylinder position, file descriptor, and offset address in the file as available sector position information.
After that, the record is taken out from the work file buffer (step 326).

The record thus obtained is merged (step 330), and when a record output request is made (step 332), the record is transferred to a work file buffer on the main storage (step 338). ). However, if the work file buffer is full when attempting to transfer (step 33
3), the work file output means 12 outputs the contents of the work file buffer to the work file 31 before transferring the record (step 337). At this time, the work file output unit 12 requests an available sector from the available sector selection unit 11 (step 334), and the available sector selection unit 11 obtains the current cylinder position of the head of the magnetic disk device 3 by the cylinder position obtaining unit 14. (Step 335), referring to the cylinder position registered in the available sector management table 21, select an available sector that has the closest cylinder position and minimizes the moving distance of the head of the magnetic disk drive 3 (Step 335). 336), the area of the selected available sector is set as the output destination of the contents of the work file buffer.

In this way, the record input processing and the output processing are repeated until the merging of all the records is completed (step 341). Note that the record input processing and the output processing are not performed alternately. Particularly, at the final merge stage (step 331) in the collation processing (step 320), there is no output processing to the work file and only the input processing. Are repeated, and no available sector is registered in the input processing during this time.

FIG. 4 is a flowchart for explaining a cylinder position obtaining process according to the embodiment of the present invention. Although a specific method for obtaining the cylinder position information by the cylinder position obtaining means 14 shown in FIG. 1 depends on the operating system or the like and cannot be specified, one method will be described with reference to FIG.

In this method, the number of the sector (the unit area of the input / output of the magnetic disk device, which is determined independently of the data file management method) is the number of the last input / output sector. It is assumed to be recorded in the system management area. Under such a premise, the cylinder position obtaining means 14 first obtains the sector number m of the last input / output (step 401) and obtains the number n of sectors per cylinder of the magnetic disk drive 3. (Step 402), m / n
The cylinder position of the head is calculated according to the formula (round down the decimal point) (step 403).

Here, if the available sector is determined only by the difference between the sector numbers without calculating the cylinder position, it is possible that the cylinders are distant even in the available sector having the closest sector number. Calculate and determine the cylinder position.

[0028]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Next, in order to more specifically describe the above-described embodiment of the present invention, with reference to the drawings, an embodiment in which the embodiment of the present invention is applied and sort processing is actually performed will be described. And will be described in detail. In this embodiment,
One merge in the collation phase in the latter half of the sort process will be described.

FIG. 5 is a block diagram for explaining an embodiment of the present invention. Referring to FIG. 5, the present embodiment includes:
It comprises a magnetic disk device 51, a main storage device 53, and a sort processing means 54.

Since the sort processing means 54 has the same configuration as the sort processing means 1 shown in FIG. 1, the description is omitted here.

In the work file area 52 on the magnetic disk device 51, three strings of a string (# 1) 521, a string (# 2) 522 and a string (# 3) 523 are output.

The string (# 1) 521 and the string (# 2) 522 are each composed of three blocks and are output to the cylinder (# 1) 511 of the magnetic disk drive 51. The string (# 3) 523 is composed of four blocks and is output to the cylinder (# 2) 512 of the magnetic disk device 51.

In this embodiment, it is assumed that each block has a size of one sector. For example, reference numerals 5211, 5212 and 5213 shown in FIG.
21 and the sector position of each block is 1, 2 and 3. Note that this relationship is established between the string (# 2) 522 and the string (# 3) 5
The same applies to 23.

It is assumed that each block includes two records. For example, the codes ac and de shown in FIG. 5 are records in the block, and one character of the code indicates one record (contents of a key). In the following description, each block is referred to as “block ac”, “block de”, and the like.

The main storage device 53 has a sort memory 531
And an available sector management table 532. The sort memory 531 has the number of strings, that is, three input buffers 5311, 5312, and 5313, and an output buffer 5314, and reads one block from each string into the input buffer.

The merge process is performed on records in the input buffer, and when an output record is determined, it is buffered in the output buffer 5314 and output. Also, all records in one input buffer are transferred to the output buffer, and the next block is read when the input buffer becomes empty.

Next, the operation of this embodiment will be described. 6 to 9 are diagrams for explaining the sorting process in one embodiment of the present invention, and the same components as those shown in FIG. 5 are denoted by the same reference numerals.

In the following description and the corresponding drawings, in order to distinguish between a record before input and a record of an output result after merge processing, the former is represented by a lowercase letter, and the latter is represented by an uppercase letter. In the present embodiment, the cylinder position and the sector position in the cylinder are registered in the available sector management table 532. At this time, the sector position 1 in the cylinder 1 is expressed as "position 1-1". I do.

The operation of this embodiment will now be described with reference to FIGS. 6 to 9 showing the time-series steps in the sorting process.

First, with reference to FIG. 6, the operation up to reading the first block of each string will be described.

Referring to FIG. 6, the work file input means includes the first blocks ac, ad and a of each string.
b is read into the input buffers 5311, 5312 and 5313 in order from the strings # 1, # 2 and # 3,
Positions 1-1, 1-4 and 2-1 of the read block
Is registered in the available sector management table 532.

Next, with reference to FIG. 7, the operation from merging the data read as shown in FIG. 6, outputting the merged data, and reading the second block from each string will be described.

Referring to FIG. 7, the six records read as shown in FIG. 6 are merged and output as three blocks AA, AB and CD of two records each.

The first block AA is output to the latest input position (current cylinder position) 2-1. The second block AB has no available sector in the same cylinder # 2. Is output to

When the second block AB is output,
The input buffer 5313 of the string # 3 becomes empty, and the next block df is read from the position 2-2.

Further, when the first record C of the third block CD is transferred to the output buffer, the input buffer 5311 of the string # 1 becomes empty, and the next block de is read from the position 1-2.

The third block CD is output to the latest input position (current cylinder position) 1-2. Thereafter, the input buffer 5312 of the string # 2 becomes empty, and the next block dg is moved to the position 1-5. Read from.

Next, with reference to FIG. 8, the operation of merging and outputting the data read as shown in FIG. 7, and further reading the third block from each string will be described.

Referring to FIG. 8, the six records read as shown in FIG. 7 are merged and divided into three blocks DD, DE, and FG each having two records and output.

The first block DD is output to the latest input position (current cylinder position) 1-5, and the second block DE is output to the available sector 1-1 in the same cylinder # 1.

When the second block DE is output,
The input buffer 5311 of the string # 1 becomes empty, and the next block hj is read from the position 1-3.

Further, when the first record F of the third block FG is transferred to the output buffer, the input buffer 5313 of the string # 3 becomes empty, and the next block hi is read from the position 2-3.

The third block FG is output to the latest input position (current cylinder position) 2-3, after which the input buffer 5312 of the string # 2 becomes empty, and the next block hk is moved to the position 1-6. Read from.

Finally, the operation until the merge is completed will be described with reference to FIG.

Referring to FIG. 9, the six records read as shown in FIG. 8 are merged and output as three blocks HH, HI and JK of two records each. The third fourth block ZZ is output.

The first block HH is output to the latest input position (current cylinder position) 1-6, and the second block HI is output to available sectors 1-3 in the same cylinder # 1.

When the second block HI is output,
The input buffer 5313 of the string # 3 becomes empty, and the next block ZZ is read from the position 2-4.

The third block JK is output to the latest input position (current cylinder position) 2-4, and the last block ZZ is output to the available sector 2-2 in the same cylinder # 2.

FIGS. 10 and 11 show, for comparison with the present embodiment, the same data as those used in the present embodiment sorted by the conventional method described in the above-mentioned Japanese Patent Application Laid-Open No. 5-165604. FIG. 10 is a diagram for explaining an operation when a process is performed, and the same components as those shown in FIGS. 5 to 9 are denoted by the same reference numerals.

In the conventional method described in Japanese Patent Laid-Open No. Hei 5-165604, available sectors are registered in the form of a stack. Select the sector for which input processing has been performed.

The operation up to reading the second block of each string is exactly the same as in FIGS. 6 and 7, and a description thereof will be omitted.

FIG. 10 shows an example of merging the data read as shown in FIG. 7 and outputting the merged data.
FIG. 9 is a diagram for explaining an operation up to reading a second block, and corresponds to FIG.

Referring to FIG. 10, blocks DD, DE, and FG are output in the same order as in FIG. 8, but the second block DE is not the available sector 1-1 in the same cylinder 1 but the latest one in the available sector. Input position 2
8 is different from FIG.

FIG. 11 is a diagram for explaining an operation until the merge is completed, and corresponds to FIG.

Referring to FIG. 11, blocks HH, HI, JK and ZZ are output in the same order as in FIG. 9, but the last block ZZ is output to position 1-1 first registered in the available sector management table. This is different from FIG.

As described above, in the conventional method shown in FIGS. 10 and 11, the cylinder position must be moved ten times at the time of input / output processing for the work file. According to the method of the present invention shown in FIG. 6, the cylinder position needs to be moved seven times, and the movement of the head of the magnetic disk drive can be reduced.

While the embodiments and examples of the present invention have been described above, the present invention is not limited to such embodiments and examples.

For example, the method of obtaining the cylinder position information includes a method of calculating the cylinder position from the last sector number of the input / output operation recorded in the system management area of the operating system as described above. A method of detecting the state of the head of the magnetic disk device and acquiring the sector number and the cylinder position as numerical values can also be adopted.

According to such a method, the cylinder position of the head can be accurately determined even if there is a head movement that cannot be recognized by the system, such as a head movement by firmware. Further, by directly obtaining the cylinder number instead of calculating the cylinder number from the sector number, the processing of calculating the cylinder number can be omitted, and the processing can be speeded up.

[0070]

As described above, according to the present invention,
In a sort processing method for sorting a large amount of data that cannot fit in the main storage device using a work file, an available sector at a cylinder position closest to the current cylinder position of the head as an output position is always determined by an available sector selection unit. Since the selection is made, the movement of the head of the magnetic disk device during the input / output processing for the work file in the collation phase is minimized, and the time required for sorting can be reduced.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of an embodiment of the present invention.

FIG. 2 is a diagram showing a configuration of an available sector management table according to the embodiment of the present invention.

FIG. 3 is a flowchart illustrating a sorting process according to the embodiment of the present invention.

FIG. 4 is a flowchart illustrating a cylinder position obtaining process according to the embodiment of the present invention.

FIG. 5 is a block diagram for explaining one embodiment of the present invention.

FIG. 6 is a diagram for explaining a sorting process in one embodiment of the present invention (first stage).

FIG. 7 is a diagram illustrating a sorting process according to an embodiment of the present invention (second stage).

FIG. 8 is a diagram illustrating a sorting process according to an embodiment of the present invention (third stage).

FIG. 9 is a diagram for explaining a sorting process in one embodiment of the present invention (fourth stage).

FIG. 10 is a diagram for explaining an operation when the same data as that used in one embodiment of the present invention is sorted by a conventional method (third stage).

FIG. 11 is a diagram for explaining an operation when the same data as that used in one embodiment of the present invention is sorted by a conventional method (fourth stage).

[Explanation of symbols]

REFERENCE SIGNS LIST 1 sort processing means 11 available sector selection means 12 work file output means 13 work file input means 14 cylinder position acquisition means 2 main storage device 21 available sector management table 211 file descriptor 212 offset address in file 213 cylinder position in magnetic disk drive 3 Magnetic Disk Unit 31 Work File 32 Area on Work File 51 Magnetic Disk Unit 511, 512 Cylinder Position in Magnetic Disk Unit 52 Work File Area on Magnetic Disk Unit 521, 522, 523 Strings in Work File 5211, 5212, 5213 Blocks 5221, 5222, 5223 constituting string # 1 Blocks 5231, 5232, 5233, 5234 constituting string # 2 String # 3 Sorting Block 53 main storage device 531 for forming memory 5311,5212,5313 input buffer 5314 outputs the buffer 532 Abbaye Reconfigurable sector management table 54 sorting means

──────────────────────────────────────────────────続 き Continuation of front page (58) Field surveyed (Int.Cl. ⁶ , DB name) G06F 12/00

Claims (2)

(57) [Claims] In a sort processing method for sorting a large amount of data that cannot be stored in a main storage device using a work file, a work file for temporarily storing intermediate results of the sort processing, A work file output means for outputting a sorted record sequence as an intermediate result from the main storage device to the work file; and a main file unit for sorting the sorted record sequence again for merging the sorted record sequences output on the work file. A work file input means for reading into the storage device, and a table for managing an area on the work file which can be reused by inputting the sorted record sequence output on the work file to the main storage device And
Each time the sorted record sequence is read into the main storage device by the work file input means, an available sector management table for registering the position of the read area is merged with the sorted record sequence output on the work file. When the merged result is output to the work file again, it is called by the work file output means, selects an area on the work file to be output next from the available sector management table, and outputs the selected area to the work file output means. comprising a Abbaye Reconfigurable sector selection means for notifying, the cylinder position acquiring means for acquiring the current cylinder position of the head of the magnetic disk device, wherein the working file input means, source <br/> over preparative from the working file Required to perform I / O every time a read record sequence is read into main storage Along with the position information, the cylinder position of the head of the magnetic disk drive acquired by the cylinder position acquisition means after the input processing is performed is registered in the available sector management table, and the available sector selection means outputs When selecting an area on the work file, the current cylinder position of the head of the magnetic disk drive is acquired by the cylinder position acquisition means, and the acquired current cylinder position is registered in the available sector management table. A sort work file space management method, which compares the cylinder position of each area with the selected area and selects the area at the closest cylinder position. 2. A sorted record sequence output on a work file in a sort processing method using a work file to sort a large amount of data that cannot be stored in a main storage device .
Work that can be reused because the
Available for storage management of the area on the business file
Every time a sorted record sequence, which is an intermediate result of the sorting process, is read from the work file into the main storage device, the sector management table and the position information necessary for performing input / output and the information on the area where the input process was performed are performed. Register the cylinder position of the head of the magnetic disk drive in the available sector management table .
Means for merging sorted record strings read into the main storage device and outputting the merged result to the work file again, the current cylinder position of the head of the magnetic disk device, and the available sector management. Means for comparing the cylinder position of each area registered in the table and selecting the area located at the closest cylinder position.