JP2008140459A - Hybrid disk storage device and disk write method applied to the same - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To improve reliability of data by compensating a weak point of using a nonvolatile memory. <P>SOLUTION: In a disk 11 on which data is to be read/write by a head 12, a specific region (backup region 11) being different from a region in which data specified from a host is to be written is secured previously. A CPU20 writes data stored in a nonvolatile memory 22 which is used for writing data to be written in the optical disk 11 more preferentially than in the disk 11 in a backup region 111 of the disk 11. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a hybrid disk storage device having a non-volatile memory, and in particular, the data in the non-volatile memory is specified in advance on the disk separately from the area on the disk to which the data is originally written. The present invention relates to a hybrid disk storage device for writing to an area and a disk writing method applied to the device.

  A disk storage device such as a magnetic disk device (hard disk drive) uses a disk as a storage medium. Access to the disc involves mechanical operations. For this reason, access to the disk is slower than access to a memory (for example, a semiconductor memory). Therefore, the disk storage device generally has a cache memory called a disk cache in order to speed up access to the disk storage device from a host such as a personal computer using the disk storage device. Is.

  In recent years, disk storage devices that use part or all of rewritable nonvolatile memory as cache memory have been proposed (see, for example, Patent Documents 1 and 2). As described above, by using a part or all of the nonvolatile memory as the cache memory, it is possible to prevent the data existing only in the cache memory from being lost even when the power is shut off during use of the apparatus, for example.

In particular, Patent Document 2 describes that after the power of the apparatus is shut down, the data in the cache memory (nonvolatile memory) is written back to the area on the disk (actual device) where the data is to be written at the next power-on. Has been. Japanese Patent Application Laid-Open No. 2003-228561 describes that scheduling of this write-back is performed so that data in a cache management table that is referred to in order to read data in the cache memory is read in the order of track addresses on the disk. By this scheduling, the time required for the seek operation associated with the write back is reduced.
JP 2006-4407 A Japanese Patent No. 3183661

  However, a nonvolatile memory cannot stably store data compared to a disk storage medium such as a magnetic disk, and the stored data changes in a relatively short period (about several months in an extreme example). It is known. That is, the nonvolatile memory is less reliable than the disk storage medium in that the data is stably retained for a long period of time. For this purpose, for example, a function for refreshing data stored in the nonvolatile memory is incorporated in the controller of the nonvolatile memory itself.

  In the disk storage device described in Patent Document 2, the data in the nonvolatile memory (cache memory) is written back to an area on the disk where the data is to be written when the power is turned on next time after the device is turned off. . Therefore, if the power supply to the apparatus is repeatedly turned off / on in a relatively short period of time, it is possible to compensate for the above-described weak points of the nonvolatile memory. However, when the device is operated continuously for a long time, it is difficult to compensate for the weaknesses of the nonvolatile memory.

  Also, if the area on the disk (real device) where the data in the non-volatile memory (cache memory) is to be written is distributed on the disk, it is scheduled to be read out in the order of track addresses on the disk. However, it is difficult to shorten the time required for the seek operation associated with the write back.

  The present invention has been made in view of the above circumstances, and its purpose is to secure the data stored in the nonvolatile memory in advance on the disk different from the area on the disk where the data should be originally written. By writing to a specific area, high-speed access is possible, but it compensates for the weakness of using non-volatile memory, which is inferior to disk storage media in terms of maintaining stable data for a long period of time. An object of the present invention is to provide a hybrid disk storage device capable of improving reliability.

  According to one aspect of the present invention, a hybrid disk storage device including a rewritable nonvolatile memory is provided. In this hybrid disk storage device, a disk in which a specific area different from the area to which the data is to be written is stored in advance for storing data stored in the nonvolatile memory, and requested by the host Data writing means for writing data to be written to the disk to the nonvolatile memory in preference to data for the disk, and writing data stored in the nonvolatile memory to the specific area of the disk Specific area operating means.

  According to the present invention, the data stored in the non-volatile memory is stored in a specific area (that is, the data stored on the disk different from the area on the disk to which the data is to be written (original area)). It is written in a specific area on the disk which is highly reliable in that it is stably held for a long period of time. This makes it possible to compensate for the weakness of using a non-volatile memory that is inferior to a disk storage medium in terms of stably holding data for a long period of time, and to improve data reliability. In addition, the data stored in the non-volatile memory is written to a limited area (specific area) different from the original area on the disk, so the data written to this specific area must be read. When this occurs, it is possible to perform high-speed access compared to reading data written in a distributed manner in the original area.

  Embodiments in which the present invention is applied to a hybrid magnetic disk apparatus will be described below with reference to the drawings. FIG. 1 is a block diagram showing the configuration of a hybrid magnetic disk drive (HDD) according to an embodiment of the present invention. In FIG. 1, for example, the upper disk surface of a disk (magnetic disk medium) 11 forms a recording surface on which data is magnetically recorded. A head (magnetic head) 12 is arranged corresponding to the recording surface of the disk 11. The head 12 is used for data writing (data recording) to the disk 11 and data reading (data reproduction) from the disk 11. It is assumed that the lower disk surface of the disk 11 also forms a recording surface, and a head similar to the head 12 is arranged corresponding to the recording surface. In the configuration of FIG. 1, an HDD including a single disk 11 is assumed. However, it may be an HDD in which a plurality of disks 11 are stacked.

  A plurality of concentric tracks (not shown) are arranged on each recording surface of the disk 11. In each track, servo areas in which servo information is recorded are arranged in advance at regular intervals and discretely. The servo information includes position information necessary for control to position the head 12 on the target track. This position information includes a cylinder code (cylinder number) indicating a cylinder (track) position on the disk 11 in which servo information is written, and a position error signal for detecting a deviation of the head 12 from the cylinder position. . An area sandwiched between adjacent servo areas is called a user data area. In the user data area, a plurality of data sectors used for recording / reproducing user data are arranged. In addition, a backup area 111 for backing up data stored in the nonvolatile memory 22 described later is secured in a specific area (first specific area) on the disk 11. The backup area 111, that is, the first specific area on the disk 11 has the same capacity as the nonvolatile memory 22.

  The disk 11 is rotated at high speed by a spindle motor (SPM) 13. The head 12 is attached to the tip of an actuator (carriage) 14. The actuator 14 has a voice coil motor (VCM) 15 serving as a drive source for the actuator 14. The actuator 14 is driven by the VCM 15 to move the head 12 in the radial direction of the disk 11. Thereby, the head 12 is positioned on the target track.

  The SPM 13 and the VCM 15 are driven by drive currents (SPM current and VCM current) respectively supplied from a motor driver (motor driver IC) 16. That is, the motor driver 16 includes an SPM driver and a VCM driver (not shown) that drive the SPM 13 and the VCM 15 respectively.

  The head 12 is connected to a read / write channel (read / write IC) 17 through a head amplifier circuit (head IC) (not shown). The read / write channel 17 is a device for performing signal processing related to read / write, processing for analog / digital conversion (A / D conversion) of a read signal amplified by a head amplifier circuit, and processing for decoding read data. Then, processing for extracting servo information from the data after A / D conversion is executed.

  The motor driver 16 and the read / write channel 17 are connected to the system LSI 18. The system LSI 18 is an LSI in which a disk controller (HDC) 19 and a CPU 20 are integrated on a single chip. The system LSI 18 is connected to the ROM 21 and the nonvolatile memory 22. The system LSI 18 may further include at least one of the read / write channel 17, the ROM 21, and the nonvolatile memory 22. The ROM 21 stores a control program (firmware program) in advance.

  The nonvolatile memory 22 is, for example, an electrically rewritable flash memory. The nonvolatile memory 22 can maintain the stored data even when power is not supplied. On the other hand, data stored in a cache memory that is conventionally applied using, for example, a DRAM (Dynamic RAM) is lost when power supply to the cache memory is cut off. That is, the non-volatile memory 22 is suitable for storing data for a long time unlike the cache memory. Therefore, in the present embodiment, the nonvolatile memory 22 is used to store data that does not exist on the disk 11. The nonvolatile memory 22 is also used for storing a part of data existing on the disk 11 in the same manner as the cache memory in order to improve the access speed.

  The HDC 19 constitutes an interface between the HDD and a host such as a personal computer using the HDD. The HDC 19 has a function of dividing the user data transferred from the host into a predetermined size corresponding to the data sector, encoding the divided user data, and sending it to the write channel 17. The HDC 19 also performs access to the nonvolatile memory 22. This access control, in particular the control of how data is stored in the nonvolatile memory 22, is executed by the CPU 20 in accordance with a control program stored in the ROM 21.

  The CPU 20 functions as a main controller of the HDD by executing a control program stored in the ROM 21. For example, the CPU 20 performs seek control for moving the head 12 to the target track and positioning control for positioning the head 12 at the target position of the target track based on the servo information detected by the read / write channel 17. The CPU 20 also controls access to the nonvolatile memory 22 by the HDC 19, for example, how data is stored in the nonvolatile memory 22 by the HDC 19.

  The nonvolatile memory 22 is less reliable than the disk 11 in that the nonvolatile memory 22 holds data stably for a long period of time. Therefore, in the present embodiment, at least data stored only in the nonvolatile memory 22 is appropriately backed up to the backup area 111 of the disk 11 to compensate for the weak point of the nonvolatile memory 22. This backup process is executed by the CPU 20 based on a control program stored in the ROM 21. Hereinafter, this backup processing will be described with reference to the flowchart of FIG.

  The CPU 20 monitors the access frequency by monitoring the access command given from the host to the HDD, and starts the backup operation when the access frequency becomes lower than a certain level (step S1). Here, it is determined that the access frequency is low when the next access command is not transmitted even after a predetermined time has elapsed since the most recent access command was transmitted from the host.

  In the present embodiment, the conditions for starting the backup operation are the same as the conditions for starting the conventionally known self-test operation (self-diagnosis operation). Here, as described above, the backup operation is started when it is determined that the access frequency is low, that is, when the HDD is in a state where the access from the host to the HDD is not adversely affected. This backup operation may be part of the self-test operation.

  Now, the backup operation is performed as follows. First, the CPU 20 extracts data stored only in the nonvolatile memory 22 using the HDC 19 (step S2). In the present embodiment, a nonvolatile memory management table 130 (see FIG. 13), which will be described in detail later, is used to manage data stored in the nonvolatile memory 22. In the management table 130, the data (state (1) data) stored only in the non-volatile memory 22, or the data (state (state (1)) stored not only in the non-volatile memory 22 but also in the disk 11. Management information (entry information) indicating at least whether the data is any one of 2) to (4). The states (1) to (4) will be described later.

  When the data stored in the nonvolatile memory 22 is also stored in the disk 11, the data stored in the nonvolatile memory 22 is necessary for effectively using the area of the nonvolatile memory 22. Depending on, it may be overwritten with new data. On the other hand, data stored only in the nonvolatile memory 22 needs to be managed so as not to be overwritten with other data. Therefore, data stored only in the nonvolatile memory 22 is identified from other data by the nonvolatile memory management table 130 as described later. Therefore, data stored only in the nonvolatile memory 22 can be extracted by referring to the nonvolatile memory management table 130. The nonvolatile memory management table 130 is stored in a specific area in the nonvolatile memory 22 or another nonvolatile memory provided independently of the nonvolatile memory 22.

  The CPU 20 functions as a backup area (specific area) operation means, and copies the extracted data (as backup target data) to the backup area 111 on the disk 11 using the HDC 19. The operation including this light is controlled (step S3). Thereby, the backup of data from the nonvolatile memory 22 to the backup area 111 is completed.

  Then, the CPU 20 registers the completion of data backup from the nonvolatile memory 22 to the backup area 111 in association with the data in the nonvolatile memory management table 130 (step S4). This backup completion may be registered in a management table for managing data stored in the disk 11 (the backup area 111 of the disk 11). That is, it is only necessary that a mark indicating the completion of backup (backed up) is attached in association with data in either the nonvolatile memory 22 or the disk 11. This mark may be attached to the data itself of either the nonvolatile memory 22 or the disk 11. However, it is safer to apply a configuration in which the above mark is attached in association with the data of both the nonvolatile memory 22 and the disk 11. However, in the present embodiment, the mark is added in association with the data in the volatile memory 22 for simplification of description.

  Further, the condition for starting the backup operation (start timing of the backup operation) is not limited to the above example. For example, the backup operation may be started periodically at regular time intervals. If the CPU 20 monitors an access command from the host and determines that the amount of data written to the nonvolatile memory 22 exceeds a certain value, the backup operation may be started. Further, if a command for specifying backup is given from the host, the backup operation may be started. Further, if data called “Pinned Set Data” (first data) is written to the nonvolatile memory 22, a backup operation of the data may be started. The reason is that “Pinned Set Data” is basically data that is not flushed to the disk 11 as will be described later.

  Next, the data read operation in this embodiment will be described with reference to the flowchart of FIG. Assume that a data read is requested from the host to the HDD by a read command. The CPU 20 determines whether the requested data exists in the nonvolatile memory 22 (step S11).

  If the requested data exists in the nonvolatile memory 22, that is, if the nonvolatile memory 22 is hit, the CPU 20 executes a read process for reading the data from the nonvolatile memory 22 (step S12). In this way, at the time of a hit, the data can be read at a higher speed than when the requested data is read from the disk 11. The read data is transferred to the host by the HDC 19.

  On the other hand, when the requested data does not exist in the nonvolatile memory 22 (step S11), the CPU 20 changes the requested data to the original area (data) on the disk 11 specified by the request from the host. Read from sector) (step S13). As is well known, this data read is executed by the CPU 20 controlling the motor driver 16, the read / write channel 17 and the HDC 19.

  Next, details of the read process (step S12) for reading data from the nonvolatile memory 22 will be described with reference to the flowchart of FIG. First, the CPU 20 functions as a first reading unit and reads data from the nonvolatile memory 22 (step S21). Next, the CPU 20 determines whether the read data is normal (step S22). For example, an error detection code is generally attached to data stored in the nonvolatile memory 22. Therefore, it is determined whether the data is normal based on the data and the error detection code.

  If the read data is normal (step S22), the CPU 20 causes the HDC 19 to transfer the data to the host and ends the read process. On the other hand, if the read data is not normal, the CPU 20 refers to the non-volatile memory management table 130 (see FIG. 13) to register the backup completion (mark indicating this) in association with the data. Is checked (step S23). As a result of this check, the CPU 20 determines whether the backed up data can be used based on whether the requested data has been backed up on the disk 11 (step S24).

  If the backup has been completed, the backup data can be used in place of the abnormal data read from the nonvolatile memory 22. In this case, the CPU 20 functions as a second reading unit, and reads backup data corresponding to the abnormal data read from the nonvolatile memory 22 from the backup area 111 of the disk 11 (step S25). The read backup data is transferred to the host by the HDC 19 instead of the data read from the nonvolatile memory 22.

  Then, the CPU 20 tests whether the area (corresponding area) of the nonvolatile memory 22 from which data has been read can be normally written / read (step S26). If the corresponding area of the nonvolatile memory 22 can be normally written / read, that is, if the nonvolatile memory 22 is normal (step S27), the process proceeds to step S28. In this step S28, the CPU 20 functions as a recovery means, and overwrites the read backup data in the corresponding area of the nonvolatile memory 22, thereby recovering the abnormal data in that area to normal data (step S28). . As a result, the CPU 20 ends the read process. On the other hand, if the nonvolatile memory 22 is not normal, the CPU 20 registers the area as a defect so as not to use the corresponding area of the nonvolatile memory 22 in the future (step S29), and ends the read process.

  On the other hand, if the requested data has not been backed up on the disk 11 and therefore backup data cannot be used in place of the read abnormal data (step S24), the CPU 20 reads the requested data (ie, read). It is determined whether the data to be present is data that exists only in the nonvolatile memory 22 (step S30). If the requested data exists also in the area on the disk 11 where the data is to be stored, the CPU 20 reads the data stored in the area on the disk 11 (step S31). Thereafter, the process proceeds to step S26. On the other hand, if the requested data does not exist in the area on the disk 11 where the data is to be stored, that is, the data exists only in the nonvolatile memory 22 (step S30), the CPU 20 If the read has failed, the HDC 19 notifies the host of an error (step S32), and the read process ends.

  Next, the data write operation in the present embodiment will be described with reference to the flowchart of FIG. Assume that a data write is requested from the host to the HDD by a write command. First, the CPU 20 has data in the nonvolatile memory 22 having the same logical block address LBA and size as the data to be written specified by the write command, that is, the data to be written hits the nonvolatile memory 22. Is determined (step S41).

  If the nonvolatile memory 22 is not hit, the CPU 20 determines whether there is a space necessary for writing data in the nonvolatile memory 22 (step S42). If there is a vacancy (empty area) in the nonvolatile memory 22, the CPU 20 proceeds to step S43. On the other hand, if the nonvolatile memory 22 is hit (step S41), the CPU 20 skips step S42 and proceeds to step S43.

  In step S43, the CPU 20 clears (cancels) the backup completion registration associated with the data of the area on the nonvolatile memory management table 130 when writing the data to the empty area or the hit area of the nonvolatile memory 22. ) After this clear process (step S43), the CPU 20 writes the requested data to the nonvolatile memory 22 (step S44), and the data write operation is terminated. The clear process (step S43) is performed in consideration that when the data in the nonvolatile memory 22 is updated, the backup content on the disk 11 corresponding to the data becomes invalid.

  On the other hand, if there is no free space (free space) in the nonvolatile memory 22 (step S42), the CPU 20 converts the data requested by the host into an area (data sector) on the disk 11 to which the data is to be originally written. (Step S45), and the data write operation is terminated. In the following description, for the sake of simplicity, an area (data sector) on the disk 11 to which data is originally written is expressed as an original area on the disk 11 or an original sector on the disk 11. When there is no free space in the non-volatile memory 22, for example, the data (old data) in the area of the non-volatile memory 22 having a size sufficient to write the data requested by the host is transferred to the original data on the disk 11. An empty area may be secured in the nonvolatile memory 22 by writing back (flashing) the area (sector). In addition, as will be described later, the operation for securing the free space in the nonvolatile memory 22 is performed in accordance with an explicit command from the host or independently of the execution of the read / write command, for example, periodically. , May be performed by the CPU 20 (control).

[First Modification]
Next, a first modification of the above embodiment will be described.
In the above embodiment, data stored only in the nonvolatile memory 22 is backed up (copied) to the backup area 111 of the disk 11. A feature of the first modification is that all data stored in the nonvolatile memory 22 is backed up.

  FIG. 6 is a flowchart showing the procedure of backup processing applied in the first modification. As is apparent from the flowchart of FIG. 6, in the first modification, when the frequency of access from the host to the HDD decreases (step S51), all the data stored in the nonvolatile memory 22 is stored on the disk 11. The backup area 111 is backed up (copied) (step S52).

  Then, as in the above embodiment, the CPU 20 registers the completion of data backup from the nonvolatile memory 22 to the backup area 111 in the nonvolatile memory management table 130 in association with the data (step S53). In the first modified example, extraction of data to be backed up from the nonvolatile memory 22 (step S2 in FIG. 2) is unnecessary, but the time required for data copy increases.

[Second Modification]
Next, a second modification of the above embodiment will be described.
The data backup operation, that is, the data copy operation involves data reading for reading data to be backed up (copied) from the nonvolatile memory 22 and data writing for writing the read data to the backup area 111 of the disk 11. . For this reason, the backup process takes time. In particular, in the first modified example, since all data stored in the nonvolatile memory 22 is backed up, it takes a long time to complete the backup (copy). Further, when a disk access is requested from the host during the backup processing period, execution of the requested access is waited for a long time. Therefore, a feature of the second modification is that not all data stored in the nonvolatile memory 22 is copied continuously, but is copied in a plurality of times. In this way, by performing the copy operation in a plurality of times and shortening the time required for one copy operation, it is possible to prevent the response to the host from being impaired.

  FIG. 7 is a flowchart showing the procedure of backup processing applied in the second modification. As apparent from the flowchart of FIG. 7, in the second modification example, the data stored in the nonvolatile memory 22 is divided into a plurality of copying operations and is backed up in the backup area 111 of the disk 11. The area of the non-volatile memory 22 to be subjected to the copy operation is limited to a relatively small part of the non-volatile memory 22. Therefore, in the second modification, when the frequency of access from the host to the HDD becomes low (step S61), only a part of the data stored in the nonvolatile memory 22 is copied to the backup area 111 on the disk 11. An operation is performed (step S62). Then, the CPU 20 registers the completion of data backup from the nonvolatile memory 22 to the backup area 111 in association with the data in the nonvolatile memory management table 130 (step S63).

  Next, the CPU 20 determines whether or not access from the host to the HDD is requested during one copy operation (step S64). If access is requested from the host (step S64), the CPU 20 immediately ends the backup operation in order to prioritize execution of the request. As a result, it is possible to prevent the host from waiting for a long time for the copy operation.

  On the other hand, if access from the host is not requested (step S64) and the backup of the nonvolatile memory 22 is not completed, the CPU 20 stores another part of the data stored in the nonvolatile memory 22. A copy operation for copying to the backup area 111 on the disk 11 is performed (step S62). As described above, the CPU 20 divides the data stored in the nonvolatile memory 22 into several pieces of data, and executes a copy operation in units of a small amount of data. Thereby, in the second modified example, it is possible to prevent the response to the host from being damaged due to the backup process.

[Third Modification]
Next, a third modification of the above embodiment will be described.
The capacity of the nonvolatile memory 22 is limited. For this reason, it is preferable that old data stored in the nonvolatile memory 22 is written to an original area on the disk 11 and the nonvolatile memory 22 in which the data is stored is released as a free area. This makes it possible to write other data to the released area. When the nonvolatile memory 22 is used in this way, old data stored in the nonvolatile memory 22 is always overwritten with other data. If it is necessary to temporarily write large data to the non-volatile memory 22, the old data on the non-volatile memory 22 is distributed and written to the original area on the disk 11.

  In FIG. 8, when the old data on the nonvolatile memory 22 is data A, B, and C, the data A, B, and C are written in a distributed manner in the original area on the disk 11 (arrangement). An example of the state is shown. In such a state, when data A, B, and C are needed again and an attempt is made to return them to the non-volatile memory 22, it takes a long time to read the data A, B, and C from the disk 11.

  Therefore, the feature of the third modification is that instead of writing the old data in the nonvolatile memory 22 to the original area on the disk 11, the data is stored in a specific area (second specific area) on the disk 11. Thus, the data is read from the disk 11 and returned to the nonvolatile memory 22 to speed up the operation. The writing of data to the second specific area on the disk 11 is performed, for example, in order to temporarily make the area in the nonvolatile memory 22 where the data is stored empty. In this respect, the operation of writing (moving) data to the second specific area on the disk 11 is the backup operation in the above embodiment, that is, the data in the nonvolatile memory 22 is transferred to the backup area (first specific area) on the disk 11. This is different from the backup operation of writing to (area) 111.

  The backup operation in the above embodiment is automatically performed. On the other hand, in the third modification, for example, when it becomes necessary to secure a free area in the nonvolatile memory 22 due to some request from the host, the old memory of the nonvolatile memory 22 is used as a “trigger”. An operation of writing (moving) data on the disk 1 is performed. The operation of writing the data of the nonvolatile memory 22 to the disk 11 (the second specific area on the upper side) applied in the third modification is a backup operation in a broad sense. Different. Therefore, in the second modification, in order to distinguish from the “backup operation” in the above embodiment, the operation of moving the data in the nonvolatile memory 22 to the disk 11 is referred to as the “extended area operation”. The “second specific area” on the disk 11 to which data is written is called an “extended area”. Details of the conditions that trigger the extended area operation will be described later.

  The backup area (111) and the expansion area can be shared by switching according to time. However, if the backup area and the expansion area are arranged separately and separately, the control is simpler and it is possible to operate more effectively.

  FIG. 9 shows an example of a state in which the above-described data A, B, and C are collectively written in the extension area (second specific area) 112 on the disk 11 in the third modification. In FIG. 9, the backup area 111 is omitted. In the example of FIG. 9, the extended area 112 is arranged on the outer periphery of the disk 11, but is not limited to this. In the example of FIG. 9, unlike in FIG. 8, data A, B, and C exist together in the extended area 112 of the disk 11. Therefore, when it becomes necessary to read the data A, B, and C from the disk 11 in order to return the data A, B, and C to the nonvolatile memory 22, the seek time can be shortened unlike FIG. That is, the data A, B, and C can be read from the disk 11 at high speed and returned to the nonvolatile memory 22.

  Next, a data write operation (extended region operation) from the nonvolatile memory 22 to the extended region 112, which is executed when the “trigger” as described above occurs, will be described with reference to the flowchart of FIG.

  First, the CPU 20 determines whether the data to be written (moved) from the nonvolatile memory 22 to the extended area 112 on the disk 11 is data that has already been written to the extended area 112 (step). S71). If the data to be written has already been written in the extended area 112, the CPU 20 ends the process.

  On the other hand, when the data to be written is not written to the extended area 112 on the disk 11, the CPU 20 needs a free space (empty space required for writing the data from the nonvolatile memory 22 to the extended area 112. It is determined whether or not (region) is within the expanded region 112 (step S72). If there is no necessary free area in the extended area 112, the CPU 20 copies the data in the extended area 112 (data in the area where the free area should be secured) to the original area (sector) on the disk 11. (Step S73). As a result, the area in the extended area 112 where the copied data is stored can be secured as a free area necessary for writing data to the extended area 112.

  When the necessary space is secured in the expansion area 112, the CPU 20 writes the data stored in the nonvolatile memory 22 to the expansion area 112 while rearranging the data in the order of the address of the data (logical block address LBA) ( Step S74). In the third modified example, the writing of data from the nonvolatile memory 22 to the expansion area 112 is data movement. That is, the CPU 20 functions as an extended area (specific area) operation means, and performs an operation including writing of the data to the extended area 112 to move the data stored in the nonvolatile memory 22 to the extended area 112. Do. As a result, the area of the nonvolatile memory 22 in which the data written (moved) in the extended area 112 is stored is released. That is, an empty area is secured in the nonvolatile memory 22.

  On the other hand, if there is an empty area in the extended area 112 (step S72), the CPU 20 skips step S73 and proceeds to step S74. In step S74, as described above, the CPU 20 writes the data in the nonvolatile memory 22 into the free area of the expansion area 112 while rearranging the data in the order of the addresses of the data (logical block address LBA).

  Here, the rearrangement in the address order is not necessarily required. However, at the time of data reading, if the target data is not present in the nonvolatile memory 22 but is present in the extended area 112 and there is no free space in the nonvolatile memory 22, the data in the extended area 112 is stored as described later. Directly lead. For this reason, considering the data read from the expansion area 112, rearranging is more advantageous in terms of processing speed.

  When the CPU 20 writes (moves) data from the nonvolatile memory 22 to the extension area 112 (step S74), the management information for managing the data indicates the state after the data is written to the extension area 112. Update (step S75). Unlike the above-described embodiment, this management information is held in a nonvolatile memory management table 150 (see FIG. 15) as will be described later.

  Next, the data read operation in the third modification will be described with reference to the flowchart of FIG. Assume that a data read is requested from the host to the HDD. The CPU 20 determines whether the requested data exists in the nonvolatile memory 22 (that is, hits) (step S81).

  When the requested data is hit in the non-volatile memory 22 (step S81), the CPU 20 executes a read process for reading the data from the non-volatile memory 22 (step S82), similarly to step S12 in the above embodiment. ), The data read operation is terminated. The read data is transferred to the host by the HDC 19.

  On the other hand, if the requested data does not exist in the nonvolatile memory 22 (step S81), the CPU 20 determines whether the data exists in the extended area 112 on the disk 11 (step S83). If the requested data exists in the extension area 112 on the disk 11, that is, if the extension area 112 is hit, the CPU 20 proceeds to step S84. Here, the state (specific state) in which the requested data does not exist in the nonvolatile memory 22 but exists in the extended area 112 means the following. That is, in this specific state, for example, since the data movement from the nonvolatile memory 22 to the extension area 112 is performed in step S74, the requested data is also moved to the extension area 112 and remains in the nonvolatile memory 22. Means not.

  In step S84, when the requested data is read from the extended area 112, the CPU 20 determines whether or not there is a space necessary for copying the read data to the nonvolatile memory 22. judge. If there is no necessary space in the non-volatile memory 22, the CPU 20 reads the requested data from the extended area 112 on the disk 11 (step S85) and ends the data read operation. Data read from the extension area 112 is transferred to the host by the HDC 19.

  On the other hand, if there is a necessary space in the nonvolatile memory 22 (step S84), the CPU 20 functions as a copy unit, and copies the data in the expansion area 112 to the space area in the nonvolatile memory 22 ( Step S86). Then, the CPU 20 functions as a read unit, reads the data copied to the nonvolatile memory 22 from the nonvolatile memory 22 (step S82), and ends the data read operation. Data read from the nonvolatile memory 22 is transferred to the host by the HDC 19.

  On the other hand, when the requested data does not exist in the extended area 112 on the disk 11 (step S83), the CPU 20 transfers the requested data to the original area on the disk 11 specified by the request from the host ( (Sector) is read (step S87), and the data read operation is terminated. The read data is transferred to the host by the HDC 19.

  Next, a data write operation including an operation of moving data from the nonvolatile memory 22 to the extended area 112 (extended area operation) applied in the third modification will be described with reference to the flowchart of FIG.

  Assume that a data write is requested from the host to the HDD by a write command. First, the CPU 20 has data in the nonvolatile memory 22 having the same logical block address LBA and size as the data to be written specified by the write command, that is, the data to be written hits the nonvolatile memory 22. Is determined (step S91).

  If the nonvolatile memory 22 is not hit, the CPU 20 determines whether data having the same logical block address LBA and size as the data to be written exists in the extended area 112 on the disk 11, that is, written. It is determined whether the data to be hit hits the extended area 112 (step S92). If the extension area 112 is not hit, that is, if neither the nonvolatile memory 22 nor the extension area 112 is hit, the CPU 20 has a free space necessary for writing data in the nonvolatile memory 22. It is determined whether or not there is (step S93).

  If there is a necessary space (empty area) in the nonvolatile memory 22 (step S93), the CPU 20 performs the same processing as in the above embodiment, that is, processing corresponding to steps S43 and S44 in FIG. 5 (step S94). And S95). Even when the nonvolatile memory 22 is hit (step S91), the CPU 20 executes processing (steps S94 and S95) corresponding to steps S43 and S44 of FIG.

  On the other hand, if the extension area 112 is hit, the CPU 20 performs an operation of writing the requested data (new data) to the extension area 112, that is, an operation of updating old data in the extension area 112 to the new data (step S96). ).

  On the other hand, if there is no free space in the non-volatile memory 22 (required space for writing the requested data to the non-volatile memory 22) (step S93), the CPU 20 determines whether there is a free space in the expansion area 112 (step S93). Step S97). The “space” in the extension area 112 is to move data of a size corresponding to the necessary space from the nonvolatile memory 22 to the extension area 112 in order to secure the necessary space in the nonvolatile memory 22. It is necessary space.

  If there is no free space in the extended area 112 (step S97), the CPU 20 stores some data (old data) in the extended area 112 necessary for securing free space in the extended area 112 on the disk 11. Is copied to the original area (sector) (step S98). As a result, a space necessary for moving (writing) data from the nonvolatile memory 22 to the extended area 112 is secured in the extended area 112.

  When the necessary space (free space) is secured in the expansion area 112, the CPU 20 rearranges a part of the data (old data) in the nonvolatile memory 22 in the order of the address (logical block address LBA) of the data. It moves to a free area of the extended area 112 (step S99). By this data movement, the area of the nonvolatile memory 22 in which the data moved to the expansion area 112 is stored is released. That is, an empty area is secured in the nonvolatile memory 22.

  On the other hand, if there is free space (free space) in the extended area 112 (step S97), the CPU 20 skips step S98 and proceeds to step S99 to transfer some old data in the nonvolatile memory 22 to the extended area 112. Move to free space.

  As described above, in the third modified example, the data to be transferred from the nonvolatile memory 22 to the extension area 112 (step S99) is that the data to be written requested by the write command from the host is stored in the nonvolatile memory 22 and the extension memory. A case where none of the areas 112 is hit is executed as a “trigger”. However, data movement (data storage) to the extended area 112 may be executed with another condition as a “trigger” as described later.

  Here, the rearrangement of the data in the address order, which is performed when the data is moved from the nonvolatile memory 22 to the expansion area 112, is not necessarily required. However, as described above, at the time of data reading, if the target data does not exist in the nonvolatile memory 22 but exists in the extended area 112 (step S81 in FIG. 11) and there is no free space in the nonvolatile memory 22 (FIG. 11 step S82), the data in the extension area 112 is directly read (step S85 in FIG. 11). For this reason, considering the data read from the expansion area 112, rearranging is more advantageous in terms of processing speed.

  When the CPU 20 moves the data from the nonvolatile memory 22 to the expansion area 112 (step S99), the management information held in the nonvolatile memory management table 150 (see FIG. 15) for managing the data is stored. Update to represent the state after movement (step S100). Next, the CPU 20 proceeds to step S95, and writes the requested data (new data) to the reserved free space in the nonvolatile memory 22.

[Management of data stored in non-volatile memory 22]
Next, details of management of data stored in the nonvolatile memory 22 applied in the above embodiment will be described. Data stored in the nonvolatile memory 22 is classified into two types, first data and second data.

  The first data is called “Pinned Set Data”. The reading / writing of the first data is performed on a previously reserved area secured in the nonvolatile memory 22. The first data is managed using a logical block address LBA indicating an original area (sector) on the disk 11. However, all accesses to the logical block address LBA corresponding to the first data are replaced with accesses to the nonvolatile memory 22.

  In the present embodiment, for the first data, a command for specifying a corresponding logical block address LBA and a data size and reserving an area for the first data in the nonvolatile memory 22, the area (reserved area) ), A command for reading data from the disk 11 as first data, a command for releasing the reserved area and writing the data of the reserved area to the disk 11 are prepared. For the first data, no disk access occurs without these commands being given from the host.

  On the other hand, the second data is referred to as “Unpinned Data”. The second data is data that is handled as if the nonvolatile memory 22 was replaced with a conventional cache memory. The nonvolatile memory 22 is preferentially accessed over the disk 11 like the cache memory. However, a reserved area for the second data is not secured in the nonvolatile memory 22. For this reason, there are cases where there is no data to be read or no free area to be written in the nonvolatile memory 22 for the second data. In such a case, disk access occurs.

  In the present embodiment, in order to manage the nonvolatile memory 22 as a memory for storing the first and second data described above and to manage the backup operation using the backup area 111 on the disk 11, the nonvolatile memory shown in FIG. The memory management table 130 is used. The management table 130 includes a cache table 131 and a memory link table 132.

  Information (management information) of each entry in the cache table 131 relates to stored data in the nonvolatile memory 22 managed by the information, and includes an identification mark, a backed up mark, a disk flushed mark, a start LBA (logical block address), data The size and the start address of the nonvolatile memory 22 are included. The identification mark is flag information indicating whether corresponding data is first data (Pinned Set Data) or second data (Unpinned Data). In FIG. 13, the identification mark indicating the first data is represented by the symbol P, and the identification mark indicating the second data is represented by the symbol U. The start address indicates the address of the start position of the area in the nonvolatile memory 22 where the corresponding data is stored. The backed up mark and the disk flashed mark will be described later.

  Here, if the start address (the start address of the nonvolatile memory 22) is replaced with the start address of the cache memory, information obtained by removing the backup mark and the disk flush mark from the entry information in the cache table 131, that is, an identification mark, The start LBA, the data size, and the start address are the same as the entry information of the management table (cache table) that manages the cache memory that has been conventionally known. The feature of the entry information of the cache table 131 shown in FIG. 13 is that a backup completed mark and a disk flush completed mark are added to the entry information (management information) of the cache table known conventionally. .

  The memory link table 132 has an entry prepared for each erasable unit area of the nonvolatile memory 22. The information (management information) of this entry relates to the erasable unit area of the nonvolatile memory 22 managed by the information, and the start address of the unit area and the address of the next unit area linked to the unit area ( Next address). It should be noted that the entry information of the memory link table 132 includes the last address instead of the next address when there is no next unit area linked to the unit area managed by the information, that is, in the case of the last unit area. In the case where the unit area managed by the information includes a mark (END) to be used and is not used (empty), an unused mark (UNUSED) is included instead of the next address.

  As described above, the backup area 111 on the disk 11 has the same capacity as the nonvolatile memory 22. Therefore, the backup destination address is assigned in accordance with the address of the nonvolatile memory 22 in order from the head logical block address LBA of the backup area 111. As a result, the address of the nonvolatile memory 22 and the backup destination address are associated with 1: 1, that is, the data is stored in the backup area 111 with the same image as the nonvolatile memory 22. Can be used in common. However, the backup destination address is normally designated by a logical block address LBA of 512 bytes per sector. For this reason, it is necessary to divide the address (byte address) of the nonvolatile memory 22 by 512 and convert it to an LBA represented by the integer part of the result.

  For example, when a command for specifying data read (read command) is given from the host to the HDD in FIG. 1, the CPU 20 in the HDD has the data requested by the command in the nonvolatile memory 22 as described above. Is determined (step S11 in FIG. 3). This determination is performed as follows using the cache table 131 of the nonvolatile memory management table 130.

  First, the read command from the host designates the start logical block address LBA indicating the start position of the area in the logical space of the data to be read and the data size. Therefore, the CPU 20 determines whether or not the data having the start logical block address LBA and the data size exists in the nonvolatile memory 22 (that is, whether the nonvolatile memory 22 is hit) by searching the cache table 131.

  If there is a hit, the CPU 20 follows the link of the memory link table 132 from the start address (start address of the nonvolatile memory 22) registered in the cache table 131 and searches for the corresponding address in the nonvolatile memory 22. Thus, data is read from the address (step S12 in FIG. 3). On the other hand, if there is no hit, the CPU 20 accesses the disk 11 (step S13 in FIG. 3).

  If there is no free space in the nonvolatile memory 22 necessary for writing the data requested by the write command from the host (step S41 in FIG. 5), as described above, the data in the nonvolatile memory 22 (For example, the second data) may be written back (flashed) to an original area (sector) on the disk 11 and the area may be released (as a free area). In this case, the CPU 20 may rewrite the nonvolatile memory management table 130 (the cache table 131 and the memory link table 132) by writing the data requested by the host to an open area in the nonvolatile memory 22. However, it is not permitted to open the area (reserved area) where the first data is stored in the nonvolatile memory 22 without permission.

  Here, the backed up mark and the disk flashed mark will be described. In the example of the cache table 131 in FIG. 13, the backed-up mark and the disk flushed mark included in the entry information of the cache table 131 are both indicated by a symbol ◯ or x.

  The backed up mark indicates that the data (corresponding data) in the nonvolatile memory 22 specified by the corresponding entry information is backed up in the backup area 111 of the disk 11 (backed up) when the symbol ○ X indicates that the data has not been backed up. A disk flashed mark indicates that the corresponding data is written (flashed) to the original area (sector) on the disk 11 (disk flashed) when the symbol is ○, and when the symbol is x , Indicating that the data has not been flushed to the disk 11.

  FIG. 14 shows all combinations of backed up marks and disk flushed marks, and state transitions between states corresponding to each combination. As shown in FIG. 14A, there are four combinations of “x, x”, “x, o”, “o, x”, and “o, x” as shown in FIG. .

  “X, x” indicates a state (state (1)) in which the corresponding data is “not backed up and not flushed to the original area (sector) on the disk 11”. This state (1) is a state in which the corresponding data is stored only in the nonvolatile memory 22 and also indicates that the state should be preferentially backed up. By searching the entry information indicating the state (1) with reference to the cache table 131, data stored only in the nonvolatile memory 22 is extracted (step S2 in FIG. 2), and the data is stored on the disk 11 Backup can be made in the backup area 111 (step S3 in FIG. 2). After the data is backed up, a backed up mark “◯” is attached on the cache table 131 in association with the data (step S4 in FIG. 2). Thereby, as shown in FIG.14 (b), it changes from a state (1) to a state (3).

  “X, ○” indicates a state (state (2)) in which the corresponding data is “not backed up, but is flushed to the original area (sector) on the disk 11”. State (2) is basically a state where backup is not required. However, when the entire nonvolatile memory 22 is backed up as in the first modification, the data corresponding to the state (2) is also backed up (step S52 in FIG. 6). After backup, a backed up mark “◯” is attached (step S53 in FIG. 6), and the state (2) is changed to the state (4) as shown in FIG. 14 (b). The data corresponding to the state (2) is already flushed to the original area on the disk 11, but the data is held in the non-volatile memory 22 by reading the data next (that is, This is because, when a read of data having the same LBA and the same size as that data is requested, reading from the nonvolatile memory 22 instead of the disk 11 makes the access faster.

  “◯, x” indicates a state (state (3)) in which the corresponding data is “backed up but not flushed to the original area on the disk 11”. State (3) is a state that does not need to be flushed if the corresponding data is the first data. The state (3) is basically a state where it is not necessary to flush even when the corresponding data is the second data. However, in the case of the state (3), the CPU 20 may positively flush the corresponding data (second data). If the flash is aimed at the time when the access frequency to the HDD becomes low, the performance degradation caused by the flash can be suppressed. Whether the second data in the state (3) is actively flushed or kept in the nonvolatile memory 22 as much as possible may be properly used as necessary.

  “◯, ○” indicates the safest state (state (4)) in which the corresponding data is “backed up and flushed to the original area on the disk 11”. However, from the viewpoint of data safety, the state (2) or the state (3) is sufficient.

  When writing of data having the same LBA and the same size as the data of the nonvolatile memory 22 in any of the states (2) to (4) is requested, the requested data is written to the nonvolatile memory 22 (Steps S41 to S44 in FIG. 5). In this case, as shown in FIG. 14B, each state transits to a state (1) “not backed up and not flushed to the original area (sector) on the disk 11”. On the other hand, when writing of data having the same LBA and the same size as the data of the nonvolatile memory 22 in the state (1) is requested, the requested data is written to the nonvolatile memory 22 (step S41 in FIG. 5). To S44), no state transition occurs as shown in FIG. In the case of a read operation, no state transition occurs in any of the states (1) to (4).

  When the second data stored in the non-volatile memory 22 is flushed to the disk 11 due to a lack of free space in the non-volatile memory 22, the data in the non-volatile memory 22 in which the data is stored is stored. This area is freed. For this reason, the information in the nonvolatile memory management table 130 (the cache table 131 and the memory link table 132 included therein) managing this data is deleted. Here, if the corresponding information in the cache table 131 is already marked with a disk flush mark “◯”, data write (flash) to the disk 11 is unnecessary. In this case, the corresponding information in the nonvolatile memory management table 130 (the cache table 131 and the memory link table 132 included therein) may be deleted.

  Next, use of the backed up data will be described in association with the states (1) to (4). First, it is assumed that certain management information is registered in the nonvolatile memory management table 130 (the cache table 131). This indicates that the data specified (managed) by the management information exists in the nonvolatile memory 22 in any of the states (1) to (4). Therefore, in the data read operation, the read process (step S12 in FIG. 3) for reading the data designated by the management information from the nonvolatile memory 22 is normally performed regardless of any of the states (1) to (4). It ’s fine.

  When data is read from the nonvolatile memory 22 in the read process (step S21 in FIG. 4), it is determined whether the data is normal (step S22 in FIG. 4). The determination of the read data is performed by checking the error detection code added to the data. If the data read from the nonvolatile memory 22 is not normal, the following operation is performed.

  First, if the read data is in the state (1), since the data exists only in the nonvolatile memory 22 (step S30 in FIG. 4), a read error occurs (step S32 in FIG. 4). Next, in the state (2), although the backup data cannot be used (step S24 in FIG. 4), the data corresponding to the original area (sector) on the disk 11 exists (step S30 in FIG. 4). What is necessary is just to read (step S31 of FIG. 4).

  In the state (3), since the data in the nonvolatile memory 22 is backed up in the backup area 111 on the disk 11 (step S24 in FIG. 4), the backed up data may be read (step S25 in FIG. 4). ). Next, in the state (4), the data in the nonvolatile memory 22 is backed up to the backup area 111 on the disk 11 and flushed to the original area (sector) on the disk 11. In this case, either the backup area 111 or the original area (sector) data may be read. However, it is more advantageous in terms of access speed to read data (backup data) in the backup area 111 (step S25 in FIG. 4). This is because the data in the backup area 111 can be expected to be arranged in a continuous area.

  As described above, even if the data read from the nonvolatile memory 22 is not normal, if the data is in any of the states (2) to (4), the data is read from the disk 11 (the original sector or the backup area 111). It is possible to read normal data. The data read from the disk 11 may be used to try to restore the corresponding data on the nonvolatile memory 22. The reason is that, in the case of a temporary error (soft error) in the nonvolatile memory 22, the data read from the disk 11 may be rewritten to the nonvolatile memory 22 so that it can be used without any problem thereafter. It is.

  Therefore, the CPU 20 writes the test data to the area of the nonvolatile memory 22 where the error data is stored, and performs a write / read test for reading the written data (step S26 in FIG. 4). If no error has occurred in the write / read test (step S27 in FIG. 4), the CPU 20 writes the normal data read from the disk 11 into the area of the nonvolatile memory 22 (step S28 in FIG. 4). ).

  In the present embodiment, the non-volatile memory management table 130 is stored in a specific area in the non-volatile memory 22 or another non-volatile memory provided independently from the non-volatile memory 22 in consideration of data consistency. The In this case, a situation may occur in which the nonvolatile memory management table 130 itself cannot be read correctly due to an abnormality in the nonvolatile memory.

  Therefore, in this embodiment, the nonvolatile memory management table 130 itself is also backed up to the disk 11. Further, in this embodiment, in order to be able to check the abnormality of the nonvolatile memory management table 130 itself, as shown in FIG. 14, the cache table 131 and the memory link table 132 constituting the nonvolatile memory management table 130 are Error detection codes (table error detection codes) 131a and 132a are respectively added.

  In an operation that needs to use the nonvolatile memory management table 130, the CPU 20 determines whether the cache table 131 and the memory link table 132 constituting the management table 130 are normal by using error detection codes 131a and 132a, respectively. Judge by collating. It is also possible to copy the nonvolatile memory management table backed up to the disk 11 to the RAM during operation without using the nonvolatile memory.

  The nonvolatile memory management table 130 (cache table 131 and memory link table 132) having the data structure shown in FIG. 13 is not limited to the above embodiment, and can be applied to the first and second modifications.

  Next, details of management of data stored in the nonvolatile memory 22 applied in the third modification will be described. First, the features of the third modification will be summarized.

  In the above embodiment (and the first and second modifications), the data backed up in the backup area 111 on the disk 11 is not used as long as no abnormality occurs in the data in the nonvolatile memory 22. However, it can be expected that the data backed up in the backup area 111 on the disk 11 is arranged in a continuous area on the disk 11 as described above. For this reason, accessing the backup area 111 on the disk 11 is faster than accessing the original sector on the disk 11. Further, the capacity of the non-volatile memory 22 is usually very small compared to the disk 11. In consideration of this point, in the third modified example, an extended area 112 as a backup area (second backup area) corresponding to the backup area (first backup area) 111 is provided in a specific area on the disk 11. Assigned. The expansion area 112 is used as an area for expanding the substantial capacity of the nonvolatile memory 22.

  In the third modified example, unlike the above embodiment, not only to manage the non-volatile memory 22 and the backup operation, but also to manage the operation using the extended area 112 on the disk 11 (extended area operation) A nonvolatile memory management table 150 shown in FIG. 15 is used. The management table 150 includes a cache table 151, a memory link table 152, and an extended area link table 153.

  The feature of the entry information (management information) of the cache table 151 is that an extended area use mark is added to the entry information (management information) of the cache table 131 including the backed up mark and the disk flushed mark shown in FIG. Is a point. In the example of the cache table 151 in FIG. 15, the extended area use mark in the entry information of the table 151 is indicated by a symbol “O” or “X” as in the case of the backed-up mark and the disk flushed mark. The extension area use mark indicates that the data of the nonvolatile memory 22 (corresponding data) specified by the corresponding entry information exists in the extension area 112 of the disk 11 when the symbol is ◯, It indicates that the data extension area 112 does not exist.

  The extended area link table 153 is an extended area operation link table. In the backup operation, data is stored in the backup area 111 on the disk 11 in the same image as the nonvolatile memory 22 as described above. On the other hand, in the extended area operation, the data stored in the nonvolatile memory 22 is written to the extended area 112 on the disk 11 while being rearranged in the order of the logical block addresses LBA. That is, in the extended area operation, data is stored in the extended area 112 on the disk 11 as an image different from that of the nonvolatile memory 22. Therefore, in the third modification, an extended area link table 153 is provided separately from the memory link table 152.

  FIG. 16 shows all possible combinations of the extended area use mark, the backed up mark, and the disk flushed mark, and state transitions between states corresponding to each combination. Since the data with the extended area use mark “◯” does not exist on the nonvolatile memory 22, the backed up mark “◯” is not attached to the data. For this reason, the combination of the extended area use mark, the backed up mark, and the disk flashed mark is “×, ×, ×”, “×, ×, ○”, “×, ○” as shown in FIG. , X ”,“ x, o, o ”,“ o, x, x ”and“ o, x, o ”.

  “X, x, x” indicates that the corresponding data is “not in the extended area 112, not backed up, and not flushed to the original area (sector) on the disk 11” (state (1)). Indicates.

  “X, X, ○” indicates that the corresponding data is “not in the extended area 112 and not backed up, but is flushed to the original area (sector) on the disk 11” (state (2)). Indicates.

  “X, o, x” indicates a state (state (3)) in which the corresponding data is “not in the extended area 112 but backed up and not flushed to the original area on the disk 11”.

  “X, ○, ○” indicates a state (state (4)) in which the corresponding data is “not in the extended area 112 but backed up and flushed to the original area on the disk 11”.

  “O, X, X” indicates that the corresponding data is “in the extended area 112 but not backed up and not flushed to the original area (sector) on the disk 11” (state (5) ).

  “O, X, O” indicates a state (state (6)) in which the corresponding data is “in the extended area 112, not backed up, and flushed to the original area (sector) on the disk 11”. Show.

  The state transition when the read / write operation, the backup operation, and the flash operation are performed in the states (1) to (4) is the same as that in the embodiment (FIG. 14) as shown in FIG. In the case of a read operation, no state transition occurs in any of the states (1) to (6). Further, the state (5) and the state (6) are the same except for whether or not the corresponding data is flushed to the original area on the disk 11. Therefore, when the flash operation is performed in the state (5), the state (5) transitions to the state (6) as shown in FIG.

  Next, state transition in the case of an extended area operation for writing (moving) data from the nonvolatile memory 22 to the extended area 112 on the disk 11 will be described. Here, in the extended area operation, in addition to the operation of writing (moving) data from the nonvolatile memory 22 to the extended area 112 on the disk 11, the data in the extended area 112 is transferred to the nonvolatile memory 22 as described later. There is a write back (return) operation.

  First, in the states (5) and (6), the data in the nonvolatile memory 22 already exists in the extended area 112. Therefore, in the states (5) and (6), it is not necessary to move the data in the nonvolatile memory 22 to the extension area 112.

  Next, the state (5) and the state (1) are the same except for whether or not corresponding data exists in the extension area 112. Therefore, when the data in the nonvolatile memory 22 is moved to the extended area 112 in the state (1), the state transitions to the state (5) as shown in FIG. Conversely, when the data in the extended area 112 is restored to the nonvolatile memory 22 in the state (5), the state transitions to the state (1) as shown in FIG. Next, in the state (3), when the data in the non-volatile memory 22 is moved to the extended area 112, the data does not exist in the non-volatile memory 22, so the backed up mark “◯” is removed (O). To x). As a result, the state (3) transitions to the state (5) as shown in FIG.

  Similarly, when the data in the nonvolatile memory 22 is moved to the extended area 112 in the state (2), the state transitions to the state (6) as shown in FIG. Conversely, when the data in the extended area 112 is restored to the nonvolatile memory 22 in the state (6), the state transitions to the state (2) as shown in FIG. When the data in the nonvolatile memory 22 is moved to the expansion area 112 in the state (4), the state transitions to the state (6) as shown in FIG.

  Here, the timing at which the extended region operation is started will be described. In the third modified example, the extended area operation is started with a request from the host as a “trigger”. This trigger is caused by an explicit command from the host, such as a command (Flush NV Cache Command) for instructing to secure a designated free area in the nonvolatile memory 22. When a free area having the capacity requested by this command cannot be secured in the nonvolatile memory 22, the extended area operation is started, and some old second data stored in the nonvolatile memory 22 is stored on the disk 11. Moved to the extended area. As a result, an empty area having the requested capacity is secured in the nonvolatile memory 22.

  In addition, when a write command for writing the second data given from the host is executed, there is not enough free space in the nonvolatile memory 22 to write the second data requested by the command. When the HDD side (CPU 20) determines that it is being performed, the extended area operation may be started using this as a trigger. In this case, the write command from the host can be regarded as a command that implicitly instructs a data write operation (extended area operation) from the nonvolatile memory 22 to the extended area of the disk 11. By this extended area operation, some old second data stored in the nonvolatile memory 22 is moved to the extended area 112 on the disk 11 (step S99 in FIG. 12). Then, the area of the nonvolatile memory 22 in which the data moved to the expansion area 112 is stored is secured as a free area. In order to reflect this data movement state, the nonvolatile memory management table 150 (the cache table 151, the memory link table 152, and the extended area link table 153 is updated (step S100 in FIG. 12). The data requested by the write command is written into the free space secured in (Step S95 in FIG. 12).

  In addition, the first command for temporarily saving (moving) the specific first data stored in the reserved area of the nonvolatile memory 22 to the extended area 112 on the disk 11 and the first command for saving to the extended area 112 A second command for returning the first data to the nonvolatile memory 22 may be newly provided. The CPU 20 saves / restores the first data when the first / second command is given from the host to the HDD.

  Incidentally, the data stored in the nonvolatile memory 22 may include data (random data) indicated by the cache table 151 that the LBA is discontinuous and the size is relatively small. If such random data is flushed from the non-volatile memory 22 to the original sector of the disk 11, the access positions are randomly accessed during re-access, which is disadvantageous in terms of performance. Therefore, for example, when the extended area operation is executed in order to secure an empty area in the nonvolatile memory 22 as described above, such random data is preferentially written (moved) in the extended area 112. good. In this way, although it is originally random data, it is possible to access it collectively at the time of re-access to suppress the performance degradation.

  When reading data, the CPU 20 refers to the cache table 151. Thus, the CPU 20 can determine whether the target data exists in the nonvolatile memory 22 (step S81 in FIG. 11) or exists in the extended area 112 on the disk 11 (step S83 in FIG. 11). In the third modified example, when data existing in the extension area 112 is read, if there is an empty area having a size equal to or larger than the size in the nonvolatile memory 22 (step S84 in FIG. 11), the data is stored. The data is written back to the empty area of the nonvolatile memory 22 (step S86 in FIG. 11). This increases the access speed when the same data is read next time.

  When moving (writing) data from the non-volatile memory 22 to the extended area 112 on the disk 11, if a sufficient free area cannot be secured in the extended area 112 (step S72 in FIG. 10), the extended area is already stored. The old data existing in 112 is copied to the original area (sector) on the disk 11 (step S73 in FIG. 10). Here, the data write (step S73 in FIG. 10) to the original area (sector) on the disk 11 and the data write (step S74 in FIG. 10) to the extended area 112 on the disk 11 are performed continuously. HDD performance is degraded.

  Therefore, it is preferable to reserve a free area in the extended area 112 as much as possible. For this purpose, for example, if the data in the extended area 112 can be written back to the nonvolatile memory 22, the data is written back (step S86 in FIG. 11), and the area in the extended area 112 where the data is stored may be quickly released. Further, the above-described command for instructing the free area or the command for saving the first data to the expansion area 112 may be given from the host to the HDD so that the flash operation is not concentrated. .

  In the third modification, the operation of moving the data in the nonvolatile memory 22 to the extension area 112 is performed while rearranging the data in the LBA order of the data on the nonvolatile memory 22 (step S74 in FIG. 10). In this case, since the data is arranged in the extension area 112 in the LBA order, subsequent reads can be performed at high speed.

  Here, the management information of the cache table 151 shown in FIG. 15 for managing the data to be moved stored in the nonvolatile memory 22 is denoted by reference numeral 151a. It is assumed that the data managed by the management information 151a is not arranged in the LBA order (access order) in the nonvolatile memory 22, as is clear from the contents of the memory link table 152 shown in FIG. Since the nonvolatile memory 22 has high random accessibility, there is no major problem even if the data arrangement in the nonvolatile memory 22 is not changed. However, when data is moved from the non-volatile memory 22 to the extended area 112 on the disk 11, it is advantageous in terms of access speed that the data is arranged on the disk 11 in the order of access.

  Therefore, in the third modification, the CPU 20 moves the data managed by the management information 151 a from the nonvolatile memory 22 to the extended area 112 on the disk 11. Write to. This access order is the order of addresses linked by the memory link table 152 of FIG. 15 starting from the address specified by the start address in the management information 151a. Here, as is apparent from FIG. 15, the corresponding data is written to the extension area 112 in the order of 000A0000h → 0800000h → 00100000h. Thereby, the corresponding data in the above order is continuously arranged in the extension area 112.

  In order to compare the management information for managing the data after the data is moved from the nonvolatile memory 22 to the extension area 112 in the above order, the management information 151a is stored. An entry different from the existing entry is used for convenience and is denoted by reference numeral 151b. In the third modification, the contents of the extended area link table 153 for managing the data moved to the extended area 112 also reflect the data arrangement order in the extended area 112 as shown in FIG. Become.

  However, if the purpose is only to increase the speed at the time of data reading, the operation of moving the data in the nonvolatile memory 22 to the expansion area 112 is performed separately one by one regardless of the address order (such as 0800000h, 000A0000h, 00100000h). You may go to In short, it is only necessary that the arrangement of data in the expansion area 112 is continuous. That is, it is only necessary that the extended area link table 153 indicates that the data is continuously arranged in the extended area 112. Details of “continuous arrangement” will be described later.

  As described above, in the third modified example, the data in the nonvolatile memory 22 is written and continuously arranged in a limited specific area on the disk 11 called the expansion area 112. For this reason, the extended area 112 on the disk 11 has an aspect similar to that of the nonvolatile memory 22 in terms of access speed due to continuous arrangement of data, in addition to the aspect of “backup (in a broad sense)”. The non-volatile memory 22 has a high access speed but a limited capacity. On the other hand, although the disk 11 has a sufficient capacity, if the data is distributed and arranged on the disk 11, the access speed to the data becomes slow. According to the third modified example, it is possible to perform an operation that complements both of these, and the advantages of using the nonvolatile memory 22 can be further utilized. The same applies to the backup area 111 on the disk 11.

  Hereinafter, the extension area 112 and an example of data arrangement in the extension area 112 will be described with reference to FIG. In the example of FIG. 17, as shown in FIGS. 17A to 17C, a plurality of tracks having consecutive cylinder numbers on the disk 11 are allocated as the extension area 112. In the figure, the shaded area is an area where data is actually arranged (recorded) in the extended area 112. FIG. 17A shows an example in which data is arranged completely continuously in the extended area 112. The examples of FIGS. 17B and 17C show an example in which data is not arranged completely in the extended area 112 but is continuously arranged. The continuous arrangement refers to an arrangement in which the effect on the access time including the seek time is equivalent to that of the completely continuous arrangement even if there are gaps of several sectors or several cylinders between the data. Specifically, an arrangement in which an access time including a seek time does not exceed a certain time or a certain ratio (that is, a threshold value) with respect to an access time in the case of a completely continuous arrangement is called a continuous arrangement. A continuous arrangement including such an arrangement and a completely continuous arrangement is defined.

  On the other hand, FIGS. 17D and 17E show an example in which data is not arranged continuously in the extended area 112 (an example of non-continuous arrangement). In this arrangement, the access time including the seek time exceeds the threshold with respect to the access time in the case of a completely continuous arrangement.

  Now, even if a plurality of tracks having consecutive cylinder numbers are assigned as the extension area 112, the data itself stored in the extension area 112 is divided as the write / read is repeated, and FIG. As in the example of c), it is predicted that a gap will be generated between the data. However, this segmentation (gap) is limited to the range of the expansion region 112 and does not extend over almost the entire surface of the disk 11 as shown in FIGS. 17 (d) and 17 (e). For this reason, the effect on the access time is almost the same as when there is no gap between data.

  The description of the continuous arrangement / non-continuous arrangement in the extended area 112 is the same in the backup area 111. If necessary, replace the extended area 112 in this description with the backup area 111.

  In the above embodiment and its modifications (first to third modifications), the case where the present invention is applied to a magnetic disk device (HDD) has been described. However, the present invention can be applied to a disk storage device other than the magnetic disk device, such as a magneto-optical disk device, as long as the disk storage device (hybrid disk storage device) includes a nonvolatile memory.

  In addition, this invention is not limited to the said embodiment or its modification example as it is, A component can be deform | transformed and embodied in the range which does not deviate from the summary in an implementation stage. In addition, various inventions can be formed by appropriately combining a plurality of constituent elements disclosed in the embodiment or the modification thereof. For example, you may delete a some component from all the components shown by embodiment or its modification.

FIG. 1 is a block diagram showing a configuration of a hybrid magnetic disk apparatus according to an embodiment of the present invention. 6 is a flowchart showing the procedure of backup processing applied in the embodiment. 6 is a flowchart showing a procedure of a data read operation applied in the embodiment. 6 is an exemplary flowchart showing the procedure of a read process for reading data from the nonvolatile memory applied in the embodiment. 6 is a flowchart showing a procedure of a data write operation applied in the embodiment. 9 is a flowchart showing the procedure of backup processing applied in the first modification of the embodiment. 9 is a flowchart showing the procedure of a backup process applied in the second modification of the embodiment. The figure which shows the state in which the data of the non-volatile memory are distributed and written in the original area | region on a disk in order to demonstrate the 3rd modification of the embodiment. The figure which shows the state in which the data of the non-volatile memory were collectively written in the specific area | region on the disk in order to demonstrate the 3rd modification of the embodiment. The flowchart which shows the procedure of the data write-in operation | movement from the non-volatile memory applied to the said 3rd modification to an expansion area. 14 is a flowchart showing a procedure of a data read operation applied in the third modified example. 14 is a flowchart showing a procedure of a data write operation applied in the third modified example. The figure which shows the example of a data structure of the non-volatile memory management table applied in the embodiment. FIG. 14 is a diagram showing all combinations of backed up marks and disk flushed marks indicated by a cache table included in the nonvolatile memory management table shown in FIG. 13 and state transitions between states corresponding to the combinations. The figure which shows the data structure example of the non-volatile memory management table applied in the said 3rd modification. FIG. 15 shows all combinations of the extended area use mark, the backed up mark, and the disk flushed mark indicated by the cache table included in the nonvolatile memory management table shown in FIG. 15, and state transitions between states corresponding to each combination. Figure. The figure which shows the example of the continuous arrangement | positioning in the expansion area | region 112 / the arrangement | positioning which is not continuous.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 11 ... Disk, 12 ... Head, 19 ... HDC (disk controller), 20 ... CPU, 21 ... ROM, 22 ... Nonvolatile memory, 111 ... Backup area (specific area), 112 ... Expansion area (specific area), 130, 150: nonvolatile memory management table, 131, 151 ... cache table, 132, 152 ... memory link table, 1153 ... extended area link table.

Claims (11)

Rewritable nonvolatile memory,
A disk on which data is read / written by a head, and a specific area for storing the data stored in the nonvolatile memory is secured in advance, which is different from the area where the data is to be written. When,
Data write means for writing data to be written to the disk requested by the host to the nonvolatile memory in preference to the disk;
A hybrid disk storage device comprising: specific area operating means for writing data stored in the nonvolatile memory to the specific area of the disk.   2. The hybrid disk storage device according to claim 1, wherein the specific area operation means continuously arranges data stored in the nonvolatile memory in the specific area of the disk.   The specific area operating means rearranges the data in the order of addresses at which the data should be written to the disk when writing the data stored in the nonvolatile memory to the specific area of the disk. 3. The hybrid disk storage device according to claim 2, wherein writing is performed in a specific area. The specific area of the disk is used as a backup area for backing up data stored in the nonvolatile memory,
The specific area operating means is for specifying the specific area for copying data including at least data not stored in the disk among data stored in the nonvolatile memory to the specific area of the disk as backup target data. The hybrid disk storage device according to claim 1, wherein the backup target data is backed up to the specific area of the disk by an operation including writing of the backup target data to the area. A first read unit that reads target data from the nonvolatile memory when data to be read from the disk requested by the host is stored in the nonvolatile memory;
When the data read from the nonvolatile memory is not normal and the read data is backed up in the specific area of the disk, the data is backed up in the specific area instead of the abnormal data The hybrid disk storage device according to claim 4, further comprising: second read means for reading data.   6. The data processing apparatus according to claim 5, further comprising recovery means for recovering normal data by overwriting the abnormal data in the nonvolatile memory with data read from the specific area instead of the data. Hybrid disk storage device. The specific area of the disk is used as an expansion area for expanding the capacity of the nonvolatile memory;
The specific area operating means is free in the non-volatile memory by an operation including writing the data to the specific area for moving the data stored in the non-volatile memory to the specific area of the disk. The hybrid disk storage device according to claim 1, wherein an area is secured. If the data requested from the disk to be read from the disk does not exist in the nonvolatile memory but exists in the extension area of the disk, the data existing in the extension area is copied to the nonvolatile memory. Copy means;
The hybrid disk storage device according to claim 7, further comprising: a reading unit that reads the data copied to the nonvolatile memory as target data. A disk writing method applied to a hybrid disk storage device including a rewritable nonvolatile memory and a disk from which data is read / written by a head,
Writing data to be written to the disk requested by the host to the non-volatile memory in preference to the disk;
A specific area operation step of writing data stored in the non-volatile memory in a specific area reserved in advance on the disk, different from the area on the disk to which the data is to be written. A disc writing method characterized by the above. The specific area of the disk is used as a backup area for backing up data stored in the nonvolatile memory,
In the specific area operation step, the specific area for copying data including at least data not stored in the disk among data stored in the nonvolatile memory to the specific area of the disk as backup target data The disk write method according to claim 9, wherein the backup target data is backed up to the specific area of the disk by an operation including writing of the backup target data to the area. The specific area of the disk is used as an expansion area for expanding the capacity of the nonvolatile memory;
In the specific area operation step, there is a free space in the non-volatile memory by an operation including writing of the data to the specific area for moving the data stored in the non-volatile memory to the specific area of the disk. The disk writing method according to claim 9, wherein an area is secured.

Images