JP2003173294A - Disk device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To increase the efficiency of using a cache to enhance its hit rate by controlling the way of prefetching so that data from the same area on a disk will not exist in the cache at the same time. <P>SOLUTION: When prefetching is started or continued while the range of prefetching is limited, a cache data management table is used to check whether data in the range to be prefetched exists on the cache; if data within the range to be prefetched is determined to exist on the cache as effective data, and if it is determined that the data existing on the cache will not be overwritten even when untransferred data in the range to be prefetched is transferred to the cache, then the read gate of a disk formatter is controlled for opening or closing to prevent the transferred data in the range to be prefetched from being transferred to the cache. <P>COPYRIGHT: (C)2003,JPO

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention improves the efficiency of use of the cache area and the hit rate by preventing data from the same area on the disk from existing in the cache at the same time during pre-reading. The present invention relates to a technique for improving the performance of a disk device.

[0002]

2. Description of the Related Art When a hard disk receives a read request from a host, it temporarily stores read data from the disk in the cache and transfers the read request data from the cache to the host. However, if the read request data has been transferred to the cache (cache hit), the request data is transferred from the cache to the host without accessing the disk. In this case, since access to the disk is unnecessary, it is possible to save both the access time to the host request data on the disk and the time required to transfer the host request data from the disk to the cache. Therefore, the performance of the disk device can be greatly improved by increasing the hit rate.

In order to improve the cache hit rate, it is necessary to transfer data other than the host request data from the disk to the cache in advance (hereinafter, prefetch). Until now, the distance between the head position during pre-reading and the host request data has been the sector read-out method.
A method has been adopted in which the prefetching is continued up to the host request data when it is within r Track, and the method of devoting the overhead time and rotation waiting time of the host to the prefetching.
For a disk device that dedicates waiting for rotation to prefetch, for example,
There is a "magnetic disk device" (JP-A-6-95810).

[0004]

In the previous read-ahead method, the read-ahead range is determined only by the positional relationship between the current read-ahead position and the host request data and the time allocated for the read-ahead. Therefore, depending on the access pattern, a plurality of data read from the same area on the disk may exist in the cache, which causes a problem that the cache usage efficiency decreases.

Further, since the number of sectors per track increases due to the improvement in recording density and the amount of data that can be transferred to the cache increases within a fixed time, the cache capacity does not increase, so the cache capacity is insufficient with respect to the read-ahead amount. Tend to do.

In view of the above, the present invention controls the prefetching method so that data from the same area on the disk will not exist in the cache at the same time, thereby improving the efficiency of use of the cache and the hit rate. In addition, the present invention, when the host request data is distributed in the cache,
By connecting the distributed host request data so that they can be transferred to the host, the efficiency of data transfer from the cache to the host is improved.

[0007]

In order to achieve the above object, in the disk device of the present invention, when the read-ahead is limited and the read-ahead is started or continued, data from the same area on the disk is simultaneously present in the cache. The way to look ahead is controlled so that nothing happens.

[0008] That is, the present invention, as a first aspect,
A disk, a head that accesses the disk, a cache that temporarily stores the data read from the disk, a cache data management table that manages the data on the cache, and a cache control unit that controls the data transfer via the cache. And a signal processing unit that converts analog data read from a disk into digital data, or that converts digital data from a host into analog data, and the gate is opened and closed to transfer data from the cache to the signal processing unit or to perform signal processing. A disk device that implements a disk formatter that controls the transfer of data from the cache to the cache, and when the prefetch is started or continued by limiting the prefetch range, the data in the prefetch target range is cached using the cache data management table. Existing on If it is determined that at least a part of the data in the prefetch target range exists in the cache as valid data, and the untransferred data in the prefetch target range is transferred to the cache, the transferred data still remains. When it is determined that the data is not overwritten, the device further comprises means for controlling the opening / closing of the read gate of the disk formatter so as not to transfer the transferred data within the prefetch target range to the cache. We provide disk devices.

Further, when it is determined that the data existing in the cache is overwritten when the untransferred data is transferred to the cache within the prefetch target range, the opening / closing of the read gate of the disk formatter is controlled. Then, there is provided a disk device having means for preventing only data in an area which is not overwritten among data existing in the cache from being transferred to the cache.

Further, in the above-mentioned disk device, the distance between the head position during pre-reading and the host request data is Sector.
Within the s Per Track, when pre-reading is continued from the head position during pre-reading to the address of the host request data, the pre-reading method is controlled so that data from the same area on the disk will not exist in the cache at the same time. ing.

Further, in the above disk device, when the rotation waiting is used for the prefetching and the rotation waiting is used for the prefetching of the preceding data of the host request data, the data from the same area on the disk does not exist in the cache at the same time. It controls how to look ahead. Further, the disk device receives a read-ahead command during pre-reading, and when the pre-reading is newly started in the next command, the rotation wait that occurs is used for the pre-reading of the preceding data, and further the pre-reading of the preceding data is performed. When it is judged that the rotation wait still occurs even if the waiting time is allotted, when the rotation wait is allotted to the read-ahead of the subsequent data, the read-ahead method should be adopted so that the data from the same area on the disk may not exist in the cache at the same time. Have control.

As a second mode, the disk device of the present invention searches for all partially hit (half hit) areas when a full hit does not occur as a result of hit search using the on-cache data management table. When all of the host request data exists in the area, it is considered as a full hit, and means for calling the host request data distributed on the cache in the host request order and transferring it to the host is provided. We provide a disk device that does.

[0013]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 shows an example of a disk device to which the present invention is applied. A disk device to which the present invention is applied includes a program ROM 101 that implements a read-ahead control program, and a control processor 10 that reads and executes a control program on the ROM.
2. A cache control unit 103 that controls data transfer between the host and the cache and between the cache and the disk, a cache 104 that temporarily writes read request data / write request data (the cache 104 stores data read from the disk). A cache data management table 105 for managing information of the above), a servo control unit 106 and a servo control unit 10 for performing control for moving the head to a designated position when reading and writing data.
According to the instruction of 6, the voice coil motor (VCM) 107 for moving the head, the motor driver 108 for controlling the rotation of the disk, the selector 109 for selecting only the signal of the specified head from the magnetic signals read from the head, and the signal sent from the selector. The signal processing unit 110 for converting the received analog data into digital data or converting the digital data sent from the cache control unit 103 into analog data, and the read gate is opened and closed to read the signal sent from the signal processing unit 110. Cache data 1
Disk formatter 11 that transfers the write data transferred from the cache 104 to the signal processing unit 110 by opening and closing the write gate.
1. An interface control unit 112 for exchanging commands and data.

Next, the operation in the read command processing will be described.
Description will be given based on the configuration of FIG. The host sends read command data to the disk via the bus (host interface 113) connected to the disk. FIG. 2 shows the structure of a command issued during the read process. The read command is code 2 indicating the read command
1. Logical block address of the data to read (hereinafter,
LBA) 22 and the host request data transfer number 23 indicating the size of the read request data.

When the disk device receives the read command, it uses the cache data management table 105 on the cache 104 to search whether the requested data exists on the cache 104. FIG. 3 shows the configuration of the cache data management table 105. The cache data management table 105 includes a start LBA 31 of data written in the cache 104, a start address 32 in the cache memory, and a data size 33. When performing a hit search, the cache data management table 105 is used to check whether or not all the host request data exists in the cache 104, and if there is, the host starts from the cache memory start address 32 of the data in the cache. Calculate the start address of the request data on the cache 104,
The host request data is read from the start address and transferred to the host.

When the host request data does not exist in the cache 104, the logical address conversion program installed in the program ROM 101 is used to convert the LBA into a physical address indicating the arrangement on the disk, and then the servo control is performed. The position of the physical address is detected using the unit 106,
The opening and closing of the read gate of the disk formatter 111 is controlled to transfer the data sent from the signal controller 110 to the cache controller 103. Cache control unit 1
03 writes the transferred data in the cache 104, and further registers information about the data written in the cache 104 in the cache data management table 105. Then, the host request data is transferred from the cache 104 to the host.

The read data registered in the cache data management table 105 is deleted from the cache data management table 105 when the area on the disk as the transfer source is updated by writing. Therefore, in the cache data management table 105, only valid read data among the read data read in the cache 104 is registered.

FIG. 4 shows a flow of a method of controlling the prefetching method so that data from the same area on the disk does not exist in the cache 104 at the same time when the prefetching target range can be limited at the time of starting or continuing the prefetching.

When the pre-reading target range is determined (step 41) when the pre-reading is started or continued, the disk device according to the present invention uses the cache data management table 105 to store the data within the pre-reading target range in the cache 104 (in the cache). It is checked whether it has been transferred (step 42). The prefetch target range can be indicated by the LBA and data size. In addition, the cache data management table 1
Since the start LBA 31 and the data size 33 of the disk area which is the transfer source of the data existing in the cache 104 are registered in 05, it is possible to judge whether the data within the prefetch target range exists in the cache 104.

If the data within the prefetch target range exists in the cache 104 (step 43), the total amount of data existing in the cache 104 among the data within the prefetch target range is calculated, and among the data within the prefetch target range, The total amount of untransferred data is calculated (step 44). When all the untransferred data are written in the cache 104, it is checked whether the data in the cache is overwritten (step 45). If it is determined that the data on the cache is overwritten, the data is treated as data that does not exist in the cache 104 (data that has not been transferred to the cache) (step 46), and steps 44 to 45 are repeated. If it is determined that the data on the cache is not overwritten even if all the untransferred data in the prefetch target area is written to the cache 104, the start address and the data size of the area that is the transfer source of the data on the cache are set in the disk formatter 111. (Step 47), prefetching is started (step 48).

If data within the prefetch target range does not exist in the cache 104 (step 43), prefetch is immediately started (step 48). During pre-reading, the disk formatter 111 opens and closes the read gate according to the preset start address and size of the transfer source area of the transferred data, and transfers only the untransferred data within the pre-read target range to the cache 104. .

FIG. 5 shows the case where the distance between the head position X51 during pre-reading and the start address Y52 of the host request data of the read command received during pre-reading is within Sectors Per Track. Address Y5
An example in which the prefetching is continued up to 2 is shown. In this case, the prefetch target range 53 is the range from X51 to Y52. It is assumed that it is known by searching the cache data management table 105 that the data in the area B54 has already been transferred to the cache 104 (area (B) 57). In addition, even if the data of the area D55 and the area E56 are read into the cache 104 through the processing of the steps 41 to 48, the data of the area B54 (B)
It is assumed that 57 is known not to be overwritten.
The disk device according to the present invention starts prefetching when the start address and size of the area B54 are set in the disk formatter 111. When the disk formatter 111 identifies the start address of the area B54, it closes the read gate for the preset data size and stops the data transfer to the cache 104.

After that, the read gate is opened again and the data transfer to the cache 104 is restarted.
Through this series of prefetch processing, only the data of the first half area D55 and the second half area E56 of the initially set prefetch target range, excluding the area B54, is transferred to the cache 104.

Further, FIG. 6 shows an example of pre-reading when the rotation wait is used for pre-reading and the pre-reading of the preceding data, and the data within the pre-reading target range has been transferred to the cache 104. The data in the area A62 in the preceding data prefetch target range 61 is searched for in the cache data management table 105,
It is known that the data has been transferred to 04, and through the processing of steps 41 to 48, the data (A) 67 of the area A62 is not overwritten even if the data of the area D64 and the area E65 are read into the cache 104. It is assumed that Further, the predicted head position 63 after seek can be calculated from the head position immediately after receiving the read command and the start address of the host request data (the method of devoting the waiting for rotation to the prefetch of the preceding data is not within the scope of the present invention). For the sake of omitting the description), the prefetch target range of the preceding data can be obtained before starting the prefetch.

The disk device according to the present invention has an area A62.
When the start address and the data size of are set in the disk formatter 111, prefetching is started. When the disk formatter 111 identifies the start address of the area A62, the disk formatter 111 closes the read gate for the set data size and stops the data transfer to the cache 104. Then, after that, the read gate is opened again to restart the data transfer to the cache 104. Through this series of pre-reading processing, only the data in the areas D64 and E65 in the initially set pre-reading target range 61 excluding the area A62 is transferred to the cache 104.

Further, in FIG. 7, when the read command is received during the pre-reading, the pre-reading is stopped, and the pre-reading is newly started in the received read command, the rotation waiting time that occurs is pre-readed for the preceding data. Further, when it is determined that the waiting for rotation still occurs even when the prefetch of the preceding data is determined, when the prefetch control method that is used for extending the prefetch to stop the rotation waiting time is adopted, the prefetch target range of the preceding data and the subsequent An example is shown in which at least a part of the prefetch target range of data already exists in the cache 104 as valid data.

The cache data management table 105
By searching for the following data prefetch target range 701
Then, it is assumed that it is known that the data in the area A 702 and the data in the area C 704 in the preceding data prefetch target range 703 have already been transferred to the cache 104. Further, even if all the data in the areas F707, G705, H708, and I706 are transferred to the cache 104 through the processing of steps 41 to 48, the data (A) 709 in the area A702 and the area C704 on the cache 104,
(C) It is assumed that 710 is known not to be overwritten.

The disk device according to the present invention has an area A70.
When the start address and size of area 2 and area C 704 are set in the disk formatter 111 in the order of A and C, prefetching is started. When the disk formatter 111 identifies the start address of the area A 702, it closes the read gate and stops the data transfer to the cache 104 for the set data size. Then, after that, the read gate is opened again to transfer the data in the area G705. Area G
When the transfer of the data 705 is completed, a seek is performed, and after the seek, prefetching is started from the beginning of the track.

When the disk formatter 111 identifies the start address of the area C704, the disk formatter 111 closes the read gate and stops the data transfer to the cache 104 for the data size. Then, after that, the read gate is opened again to transfer the data in the area I706 to the cache 104.

By the above series of pre-reading processing, the following data pre-reading target range 701 and the preceding data pre-reading target range 7
Area F70 of Area 03 except Area A702 and Area C704
Second half of 7, area G705, area H708 and area I7
The data of 06 will be transferred to the cache 104.

Further, in the disk device according to the present invention, when the host request data is distributed in the cache 104, the distributed host request data can be linked and transferred to the host.

In FIG. 8, the host request data is cache 1
In the case of being distributed over 04, the flow of the method that enables the host request data to be transferred from the cache 104 by linking the distributed data is shown. When receiving a read request from the host, the cache data management table 105 is used to check whether the read request data exists in the cache 104 (step 81). It is checked whether the host request data are all present in one continuous data area on the cache 104 (hereinafter referred to as a full hit) (step 82),
If a full hit occurs, the read request data is transferred from the continuous area to the host (step 83).

Further, the cache data management table 105
Is used to check whether the host request data partially exists in one continuous data area on the cache 104 (hereinafter referred to as a half hit) (step 84). If a half hit occurs, all half hit areas are checked. As a target, it is checked whether or not all the host request data is included in the area (step 85). If all the host request data is included in the area, the host request in each half-hit area is included. The address information of the data is set in the register of the cache control unit 103 in the order of LBA (step 86).
Then, the host request data is cached 104 in the order set in the register of the cache control unit 103.
It is read out and transferred to the host (step 87).

Further, the cache data management table 105
To see if a half-hit occurs (step 8
4) If there is no half-hit area, or if there is a half-hit area but all the host-requested data is not included in the area, the host-requested data is read from the disk and transferred to the cache 104. Then
The host request data is transferred from the cache 104 to the host (step 88). FIG. 9 shows an example in the case where all the host request data are stored in the cache 104 and all the host request data are not included in one continuous data area. The area (A) 91 and the area (D) 93 in FIG. 9 are areas where the host request data is half-hit. The host request data in the area (A) 91 is the area (a) 92, and the host request data in the area (D) 93 is the area (d) 94. It is assumed that the transfer source end address of (a) 92 and the transfer source start address of (d) 94 are continuous.

FIG. 10 shows the configuration of registers in the cache control unit 103. Data is set in the register in the order of reading data, and the start address 1001 and end address 1 of the data area on the cache are set.
003 and pointer address 100 in the data area
2 can be set.

The cache control unit 103 uses the host request data (a) in the half hit areas (A) 91 and (D) 93.
92 and (d) 94 are identified, and the cache area information (start address 1001, cache pointer address 1002, end address 1003) of the host request data is set to
Set the registers in the order of (a) and (d). Next, the cache control unit 103 reads the data in the area (a) 92 from the cache pointer address and transfers it to the host in the order set in the register, and then reads the data in the area (d) 94 from the cache pointer address. Transfer to host.

[0037]

According to the disk device of the present invention, the use efficiency of the cache is improved and the hit rate is improved by controlling the prefetching method so that the data from the same area on the disk do not exist in the cache at the same time. Therefore, the performance can be improved.

[0038]

[Brief description of drawings]

FIG. 1 is a diagram showing a configuration example of a magnetic disk device to which the present invention is applied.

FIG. 2 is a diagram showing a configuration of a read command issued by a host.

FIG. 3 is a diagram showing a configuration of a cache data management table.

FIG. 4 is a flow showing a method of controlling a prefetching method so that data from the same area on the disk will not exist in the cache at the same time when the prefetching target range can be limited when starting or continuing prefetching.

FIG. 5 is a case where the distance between the head position during pre-reading and the start address of the host request data of the read command received during pre-reading is within Sectors Per Track, and pre-reading continues until the start address of the host request data. FIG. 6 is an explanatory diagram showing an example of a method for preventing data from the same area on the disk from being read into the cache in the case of doing so.

FIG. 6 shows an example of a method for preventing data from the same area on the disk from being read into a cache when a rotation wait occurs when accessing host request data and the rotation wait is used for prefetching of preceding data. Explanatory drawing.

FIG. 7: When a wait for rotation occurs when accessing host request data, and when the wait for rotation is used for pre-reading of preceding data and pre-reading of subsequent data, data from the same area on the disk is not read into the cache. Explanatory diagram showing an example of a method.

FIG. 8 is a flow showing a method in which, when host request data is distributed on a cache, linking the distributed data allows the host request data to be transferred from the cache.

FIG. 9 is an explanatory diagram showing an example in which host request data is distributed on a cache.

FIG. 10 is a diagram showing a configuration of a register of a cache control unit.

[Explanation of symbols]

101: Program ROM 102: control processor 103: cache control unit 104: cache memory 105: cache data management table 106: Servo control unit 107: Voice coil motor (VCM) 108: Motor driver 109: Selector 110: Signal processing unit 111: Disk formatter 112: Interface control unit 113: Host interface 21: Command code 22: Logical block address (LBA) 23: Number of host requested data transfers 31: Data start LBA 32: Start address in cache memory 33: Data size

   ─────────────────────────────────────────────────── ─── Continued front page    (72) Inventor Manabu Nishikawa             2880 Kozu, Odawara City, Kanagawa Stock Association             Company Hitachi Ltd. Storage Division F-term (reference) 5B005 JJ13 MM11 NN12 NN14                 5B065 BA01 CA15 CC08 CE12 CH05

Claims (5)

[Claims] 1. A disk, a head for accessing the disk, a cache for temporarily storing data read from the disk, a cache data management table for managing data on the cache, and a data transfer via the cache. A cache control unit for controlling, a signal processing unit for converting analog data read from the disk into digital data, or for converting digital data from the host into analog data, and data from the cache to the signal processing unit by opening / closing the gate. A disk device that implements a disk formatter that controls transfer or data transfer from the signal processing unit to the cache, and uses the cache data management table to start the read ahead when starting or continuing the read ahead with a limited read ahead range. The range data is Even if data that has not been transferred to the cache within the prefetch target range is transferred to the cache, it is determined that at least a part of the data in the prefetch target range exists as valid data in the cache. When it is determined that the data existing in the cache will not be overwritten, the opening / closing of the read gate of the disk formatter is controlled so that the data already existing in the cache out of the data of the read-ahead target range is not transferred to the cache. However, when it is determined that the data existing in the cache is overwritten when the untransferred data is transferred to the cache within the prefetch target range, the opening / closing of the read gate of the disk formatter is controlled, Of the data existing in the cache, only the data in the area that is not overwritten is cached. Disk device and controls not to transfer to the shoe. 2. The disk device according to claim 1, wherein
If the distance from the head position during pre-reading to the host request data is within Sectors Per Track and the pre-reading continues from the head position during pre-reading to the host request data address, data from the same area on the disk A disk device characterized by controlling the prefetching method so that they do not exist in the cache at the same time. 3. The disk device according to claim 1, wherein:
A disk device characterized by controlling the prefetching method so that data from the same area on the disk will not simultaneously exist in the cache when the rotation waiting is used for the prefetching of the preceding data of the host request data. 4. The disk device according to claim 1, wherein when a pre-reading read command is received and a new pre-reading is started in the next command, and the rotation wait that occurs is used for the pre-reading of the preceding data, Furthermore, if it is determined that the rotation wait still occurs even if the rotation waiting time is used for the prefetch of the preceding data, the data from the same area on the disk is simultaneously present in the cache when the rotation waiting is used for the prefetch of the subsequent data. A disk device characterized by controlling the way of pre-reading so as not to perform. 5. A disk, a head for accessing the disk, a cache for temporarily storing data read from the disk, a cache data management table for managing data on the cache, and data transfer via the cache. A cache control unit for controlling, a signal processing unit for converting analog data read from the disk into digital data, or for converting digital data from the host into analog data, and data from the cache to the signal processing unit by opening / closing the gate. A disk device that implements a disk formatter that controls data transfer or data transfer from the signal processing unit to the cache, and if a full hit does not occur when a hit search is performed using the data management table on the cache, a partial hit occurs. Search all areas, Strike requested data regarded as full hit if it contains all the half hit area, the disk apparatus characterized by enabling transferred to the host reads the host request data of the half-hit region in the host request order.