JPH0954658A - Parallel secondary storage device and its access method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To speed up the data input and output of the parallel secondary storage device without any partiality of an access load quantity. SOLUTION: When a processor 14 makes a request to write, a data position determining device 20 calculates the ratio of the load quantity of an original storage device and a secondary storage device 16 with a minimum load quantity according to information on the load quantities of the respective secondary storage devices 16 of an access load management device 26 and determines the secondary storage device 16 with the minimum load quantity as a write destination data position '2', the data are written to the original secondary storage device 16. When there is no free area at the physical location of the secondary storage device 16 with the minimum load quantity corresponding to the logical position of the data, the secondary storage device 16 with the small load quantity is selected next.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a parallel secondary storage device and an access method thereof, and more particularly, for example, data processed by a plurality of processors is stored in a secondary storage device such as a plurality of magnetic disk devices, The present invention relates to a parallel secondary storage device for reading desired data from a secondary storage device or writing data processed by a processor to the secondary storage device and an access method thereof in accordance with the above request.

[0002]

2. Description of the Related Art In recent years, the speeding up of computers has been rapidly advanced due to the speeding up of processors and the improvement of parallel processing technology accompanying the progress of semiconductor technology. Along with this, it has become possible to process large amounts of data on computers, and the number of applications that handle large amounts of data are also increasing. Therefore, in order to perform the data processing at high speed, it is not enough to increase the processing speed of the processor, and it is important to realize the data input / output performance corresponding to it.

The following methods are conventionally known as methods for speeding up data input / output to / from the secondary storage device. The first is a method of increasing the access speed of the secondary storage device itself. The access speed of the secondary storage device, particularly the magnetic disk device, is mainly determined by the time for moving the head (seek time) and the time for the disk to rotate to a desired data position (rotation waiting time). So conventionally,
The rotation speed of the disc has been improved. The rotation speed of recent high-speed magnetic disk devices reaches, for example, 7200 rpm or more.

The second method is to improve the access speed of the secondary storage device by improving the access method. For example, the document "Proceedings of the 13th ACM Symposium
on Operating Systems Principles `` The Design a
nd Implementation of a Log-Structured File System
In the file system called LFS shown in (pp.1-pp.15) ”, when the data on the secondary storage device is updated, as shown in FIG. Instead of writing to, append and write after the end of the series of data to invalidate the old data. In other words, since all writing is done in an additional form, as for writing, the head is moved backward from the end of the series of data, no unnecessary movement of the head is performed, and the seek time is greatly shortened. And the access speed is greatly improved.

Third, by connecting a plurality of secondary storage devices in parallel and processing a plurality of access requests at the same time,
This is a method for increasing the access speed of the entire secondary storage device. For example, the document “Information Processing (Vol.34 No.5)”
"Recent Secondary Storage: Disk Array (pp.642-pp.6
In the disk array called RAID shown in (51)), a series of data is divided into fixed units (blocks), and the data is cyclically distributed and stored in a plurality of disks as shown in FIG. As a result, in FIG. 2, for example, when there is an access request from the processor for the data from the first block to the eighth block, the data from the first block to the fourth block or the data from the fifth block to the eighth block is transferred. Since data can be read from the secondary storage device or written to the secondary storage device in parallel, the speed of data input / output to / from the secondary storage device is significantly improved.

As described above, there are various conventional methods for increasing the speed of the secondary storage device, but the above methods are not parallel, and it is possible to use the three methods together. .

[0007]

For example, in a parallel computer, a plurality of access requests often occur at the same time. Therefore, in particular, a disk array for connecting a plurality of secondary storage devices in parallel, which is the third technique, is used. It is mainly used. In the third method, when consecutive blocks are accessed, the access loads on the respective secondary storage devices (for example, magnetic disks) in the disk array are often the same, and all secondary storage devices are accessed. On the other hand, the accesses are simultaneously executed, which has the effect of matching the number of secondary storage devices. However, when the data requests cause uneven distribution of access loads to some of the secondary storage devices in the disk array, the access time of the secondary storage device in which the access requests are concentrated becomes a bottleneck. Further, if the number of secondary storage devices to be connected is increased in order to improve parallelism, it may be difficult to make all secondary storage devices have the same level of access even if the access size is relatively large. Therefore, it becomes easy for access requests to concentrate on some secondary storage devices.

That is, in the third conventional method, when the access load from the processor is unbalanced, the access time of some secondary storage devices becomes long, resulting in a data input / output speed increase. There was a problem that improvement could not be achieved sufficiently. Therefore, a main object of the present invention is to provide a parallel secondary storage device and its access method that can improve the data input / output speed.

[0009]

SUMMARY OF THE INVENTION A first invention is a plurality of storage media for storing data to be processed by a plurality of processors, and an access load management which is connected to the storage media and manages an access load amount from the processor to each storage medium. Means, a data position management means connected to the storage medium for managing the data position on the storage medium, and information on the load amount for each storage medium notified from the access load management means in response to an access request from the processor. A parallel secondary storage device comprising data position determining means for determining a storage medium for writing data.

According to a second aspect of the present invention, in a parallel secondary storage device provided with a plurality of storage media for storing data to be processed by a plurality of processors, in response to a data write request from the processor, the storage in which the data existed before the write request. This is an access method for a parallel secondary storage device, which compares an access load amount of a medium with an access load amount for another storage medium and, when the ratio is equal to or more than a predetermined value, writes data to a storage medium having a small access load amount.

[0011]

In the first aspect of the present invention, the data position determining means selects a storage medium having a small load as the storage medium to be written, thereby preventing the access load from being biased to a part of the storage medium, and thus the parallel secondary storage device as a whole. It is possible to give an access load on average.

According to the second aspect of the present invention, when a data write request is issued, the load amount is compared based on a comparison between the access load amount of the storage medium existing before the write request and the access load amount to another storage medium. When the ratio is greater than or equal to a predetermined value, the data is written to the storage medium having a small access load. Further, when the access load amount is less than or equal to a predetermined value, data is written to the storage medium existing before the write request and the other storage medium, and when there is a data read request, the one with the smaller access load amount is selected. A storage medium is selected and reading is performed. Therefore, the access performance is further improved.

Since there is a possibility that a storage medium different from the storage medium in which the data was stored before writing is selected as the write destination, the order of the logically continuous data and the physical location of the data are present. That is, the difference in the order of the physical data may increase with the execution of the data access, which may deteriorate the access performance. However, in these inventions, for example, the logical order of the data to be written and the data before and after the data is written. The position of the data to be written can also be determined so as not to impair the relationship between the order and the physical order on the storage medium.

[0014]

According to the present invention, since the access load can be prevented from being biased to a part of the storage medium, the parallelism of input / output can be improved and the access performance can be further improved. . In addition, since it is possible to maintain the relationship between the logical order and the physical order of data, it is possible to suppress the performance deterioration even with respect to access requests of various sizes.

The above and other objects, features and advantages of the present invention will become more apparent from the following detailed description of embodiments with reference to the accompanying drawings.

[0016]

FIG. 3 is a functional block diagram showing an embodiment of the present invention. A parallel secondary storage device 10 of this embodiment is accessed by a parallel processor 12. Parallel processor 1
2 includes a plurality of processors 14. Further, the parallel secondary storage device 10 includes a plurality of secondary storage devices 16 and a control device 18 that controls writing and reading of data in the secondary storage device 16. The secondary storage device 16 is, for example, a magnetic disk, and the control device 18 is, for example, a disk controller.

When an access request for data read or data write (including data change) to the secondary storage device 16 is generated due to the arithmetic processing of the execution application on the processor 14, the access request is accessed from the processor 14. It is transmitted to the request processing device 20. Specifically, the access request is provided with either read / write, a logical position of access data, an access size, a processor number requesting access, or the like. Upon receiving the access request from the processor, the access request processing device 20 requests the data position determining device 22 to access the data.

In the case of data reading, the data position determination device 22 determines the secondary storage device from the logical position of the data requested to be read, based on the information in the data position management table of the data position management device 24. Get the physical data location on 16. Table 1 shows an example of the data position management table.

[0019]

[Table 1]

The information on the physical data position includes the secondary storage device number in which the data exists and the information on the position where the data on the secondary storage device is stored.
Then, the data position determination device 22 determines the secondary storage device 16 from which the data is read according to the secondary storage device number, and the secondary storage device 1 is determined for the corresponding control device 18.
A read request is issued with the physical data position on 6. In response to the read request, the control device 18 notifies the access load management device 26 of the read request, the access size, and the like. In addition, the control device 18
When the processing for the immediately preceding access request to the secondary storage device 16 is completed, the data is read from the designated physical data position by the access size, and the data is determined by the access request processing device 20 via the data position determining device 22. Send the read data. After that, the control device 18
Notifies the access load management device 26 that the processing of the previous access request has been completed. On the other hand, the access request processing device 20 sends the sent read data to the processor 14 that has made the access request.

In the case of data writing, the data position determination device 22 determines the secondary storage device number in which the data requested to be written is stored based on the information in the data position management table of the data position management device 24. obtain. The data position determination device 22 further uses the access load management device 26 to load information (load amount) of each secondary storage device 16 at that time.
Then, the load amount of the secondary storage device for the above-mentioned secondary storage device number is compared with the load amount of the secondary storage device having the smallest load amount among the secondary storage devices, and the ratio is, for example, twice. If not, the secondary storage device corresponding to the secondary storage device number obtained from the data position management device 24 is determined as the write destination secondary storage device. At this time, the write data position on the secondary storage device is also the position before the write request obtained from the data position management device 24. That is, writing is performed at the original position.

On the other hand, if the ratio of the load amount of the secondary storage device in which the data was previously stored and the load amount of the secondary storage device having the smallest load amount is, for example, twice or more, the load amount The smallest secondary storage device is determined as the write destination secondary storage device. Also, based on the free area management table of the data position management device 24, the free area corresponding to the logical position of the data to be written is checked, and if there is a free area larger than the requested access size, the physical position of the corresponding free area. Is determined as the data position to be written. Table 2
Shows an example of the free area management table.

[0023]

[Table 2]

The free space management table is prepared in the data position management device 24 for each secondary storage device 16 individually. If there is no free area, a new area is separately secured on the secondary storage device and that area is set as the write destination data position. Data position determination device 22
Issues a write request together with the physical data position on the secondary storage device to the control device 18 corresponding to the write destination secondary storage device 16 thus determined. When the write destination is changed, the data position management device 24 is notified of the new write destination secondary storage device number and information regarding the data position. Therefore, the data position management device 24 rewrites the column of the corresponding logical position in the data position management table to the notified new position.

In response to the write request, the control device 18 notifies the access load management device 26 of the write request, the access size, and the like. When the processing for the immediately preceding access request to the secondary storage device 16 is completed, the control device 18 also ends the processor 1
4 writes the write data transmitted from the access request processing device 20 and the data position determining device 22 to the designated physical data position. When the data writing is completed, the control device 18 notifies the access load management device 26 that the processing of the previous write request is completed.

The above-described embodiment is constructed by using a dedicated circuit device such as a gate array. However, according to the present invention, the access request processing device 20, the data position determination device 22, the data position management device 24, and the access load management device 26 of the embodiment of FIG. 3 can be realized by a general-purpose CPU or microprocessor. In this case, it operates according to the flowcharts shown in FIGS. 4 and 5.

In the first step S1 of FIG. 4, the CPU or microprocessor (not shown) refers to the data position management table of Table 1 shown above to store the data requested to be written. Get the secondary storage device number. In the next step S2, the load amount of the secondary storage device with the number obtained in step S1 at that time is obtained. Then, in step S3, the secondary storage device having the smallest load amount among the secondary storage devices is designated or set as the write destination secondary storage device.

In the next step S4, the ratio of the load amounts of the two secondary storage devices set in steps S2 and S3 is calculated. Then, in step S5, it is determined whether the calculated ratio is "2" or more. However, the reference value for this determination may be appropriately changed depending on the specifications of the secondary storage device. If "NO" is determined in step S5, it means that the load amount of the original secondary storage device in which the data existed at that time is not more than twice the minimum load amount. First, in step S6, the CPU or the microprocessor determines the data position of the original secondary storage device as the write destination data position. In other words, if the load amount of the original secondary storage device does not exceed twice the minimum load amount, the secondary storage device is not changed and the data requested to be written exists in the secondary storage device before that. I tried to write to the storage device. This is to avoid fragmentation, that is, a situation in which data is discretely stored in a plurality of secondary storage devices. Therefore, in the next step S7, the write destination data position and the write request are sent to the control device 18 corresponding to the original secondary storage device.

When "YES" is determined in the previous step S5, the CPU or the microprocessor operates as shown in FIG.
In step S8 shown in FIG. 3, the physical area corresponding to the logical position of the write-requested data in the secondary storage device with the minimum load set in step S3 is referred to by referring to the free space management table shown in Table 2 above. Check whether there is free space at the target position. If it is determined in step S9 that there is a free area, the next steps S10 and S
At 11, the write-requested data is written to the secondary storage device, which has the minimum load amount at that time.

When "NO" is determined in the step S9, that is, when there is no free area in the logically valid physical position in the secondary storage device having the smallest load amount, in the next step S12, The CPU or the microprocessor searches for the secondary storage device having the next smallest load. Then, in the same manner as in step S4, in step S13, the ratio between the load amount of the secondary storage device in which the data existed before the write request at that time and the next smallest load amount is calculated, In step S14, it is determined whether the ratio is "2" or more. Also in this case, the judgment reference value of "2" can be appropriately changed.

If "NO" is determined in the step S14, in the next step S15, a write area required for the data requested to be written is newly secured in the secondary storage device having the smallest load amount. , The new area is determined as the write destination data position. Therefore, in step S15, the write destination data position and the write request are transmitted to the control device 18 corresponding to the secondary storage device.

In other words, when there is no free area in the secondary storage device with the minimum load amount, it is checked whether or not there is a free area in the secondary storage device with the next smallest load amount. If there is no free area in the next secondary storage device, steps S8 and S9 are executed for the secondary storage device with the next smaller load amount. Then, when the ratio of the load amounts is not “2” or more, the data is written in the new area of the secondary storage device (specified in step S3) having the smallest load amount.

Further, the load amount of the original secondary storage device does not exceed twice the minimum load amount, and the secondary storage device having the minimum load amount has an empty area at a physically valid physical position. In some cases, data is written to and read from the secondary storage device in which the data existed before the write request and the secondary storage device with the smallest load among the secondary storage devices excluding the storage device. Sometimes, if the data is read from the secondary storage device with the lower load, the load distribution becomes easier and the access load amount can be further averaged.

That is, in step S17 of FIG. 6, it is determined whether or not the secondary storage device 16 determined by the access load management device 26 to have the smallest load amount has a free area corresponding to the logical position of the data to be written. , Based on the free space management table of the data position management device 24. There is a free area in step S18 ("YE
S "), the empty area and the data position of the original secondary storage device where the data existed are set as the write destination data position in step S19.

In step S18, "NO", that is,
If it is determined that the secondary storage device 16 with the smallest load has no free area at the logically valid physical position, in step S6, as described above (in FIG. 4), the original The data position of the secondary storage device is determined as the write destination data position, and in the next step S7, the write destination data position and the write request are sent to the control device 18 for the original secondary storage device.

When the processor 14 requests the reading of the data, in step S20 of FIG. 7, the reading is requested by referring to the data position management table shown in Table 1 as in the case of the writing request. The secondary storage device number for storing the stored data is obtained. Then, step S21
In, it is determined whether the number is plural (for example, two). If "NO" is determined in the step S21, that is, if the data to be read is stored in one secondary storage device, in the step S23, the control device 18 corresponding to the number is stored. Then, the read data position and the read request are transmitted.

On the other hand, "YES" in step S21.
That is, when it is determined that the data is stored in the two secondary storage devices 16, in the next step S22, the secondary storage device having the smallest load amount is selected from the secondary storage devices of that number. Then, in step S23, as described above, the read data position and the read request are transmitted to the control device 18 corresponding to the selected secondary storage device.

Further, as shown in FIG. 8, by setting an empty area in the secondary storage device 16 in advance, the probability that the empty area is judged as "YES" in step S9 of FIG. 5 increases, It becomes easy not to lose the relationship between the logical order and the physical order of data.

[Brief description of drawings]

FIG. 1 is an illustrative view showing a method of writing data in a conventional LFS.

FIG. 2 is an illustrative view showing an arrangement example of data in a disk array which is a conventional technique.

FIG. 3 is a functional block diagram showing an embodiment of the present invention.

FIG. 4 is a flowchart showing the operation of the embodiment of the present invention.

FIG. 5 is a flowchart showing an operation following FIG.

FIG. 6 is a flowchart showing an operation at the time of writing data according to another embodiment.

FIG. 7 is a flowchart showing an operation at the time of reading data in the embodiment of FIG.

FIG. 8 is an illustrative view showing a secondary storage device of still another embodiment.

[Explanation of symbols]

 10 ... Parallel secondary storage device 14 ... Processor 16 ... Secondary storage device 18 ... Control device 22 ... Data position determination device 24 ... Data position management device 26 ... Access load management device

Claims (7)

[Claims] 1. A plurality of storage media for storing data to be processed by a plurality of processors, access load management means connected to the storage media and managing an access load amount from the processor to each storage medium, connected to the storage media. And a data position management unit that manages the data position on the storage medium, and a memory that writes data based on the load amount information for each storage medium notified from the access load management unit in response to an access request from the processor. A parallel secondary storage device comprising data location determining means for determining a medium. 2. The data position determining means further determines a storage medium to be written and a write position on the storage medium based on a logical order and a physical order of data.
The parallel secondary storage device according to claim 1. 3. A parallel secondary storage device comprising a plurality of storage media for storing data to be processed by a plurality of processors, the storage medium having the data before the write request in response to a data write request from the processor. The access method of the parallel secondary storage device, comprising: comparing the access load amount of 1) with the access load amount of another storage medium, and writing the data to a storage medium having a small access load amount when the ratio is equal to or more than a predetermined value. 4. The parallel storage system according to claim 3, wherein the writing position on the storage medium is determined based on the relationship between the logical order of the data and the data before and after the data and the physical order on the storage medium. Next storage device access method. 5. Each storage medium is divided into a plurality of continuous segments in advance, and if there is a reusable free area in a segment in which a large amount of data is adjacent to the data in a logical order, priority is given to it. 5. The method for accessing a parallel secondary storage device according to claim 3, wherein the data is written in the empty area by writing. 6. In response to a data write request from the processor, the data from the processor is written in a storage medium in which the data existed and other storage mediums other than the recording medium, and a plurality of the recording media are written. 6. The access method for a parallel secondary storage device according to claim 3, wherein the data is read from the recording medium having the smaller access load amount. 7. The access method for a parallel secondary storage device according to claim 3, wherein the storage medium includes a predetermined free area for writing the data.