JP2006251844A - Data storage device, and its write cache control method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To effectively determine the execution sequence of commands for transmitting a command completion notice after the completion of writing to a medium. <P>SOLUTION: When an FUA (Forced Unit Access) command is received in a state where a plurality of write commands WR1-WR5 are cached, an HDD 1 concerning one embodiment of the present invention determines the execution sequence where the FUA command is given a priority to the other commands. The command completion notice is transmitted after the completion of writing to a magnetic disk concerning the FUA command, so that a host latency period is shortened by the method. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a data storage device and write cache control, and more particularly to execution order control of write commands registered in a write cache of a data storage device.

  As data storage devices, devices using various types of media such as optical disks, magnetic tapes, and semiconductor memories are known. Among them, hard disk drives (hereinafter referred to as HDDs) are widely used as computer storage devices, and are one of the storage devices indispensable in current computer systems. In addition to the computer, the use of the HDD such as a removable memory used in a moving image recording / reproducing apparatus, a car navigation system, a digital camera, and the like is expanding due to its excellent characteristics.

  A magnetic disk used in an HDD has a plurality of tracks formed concentrically, and each track is divided into a plurality of sectors (servo sectors). Servo data and user data are recorded in each servo sector. In the servo sector, user data of a plurality of sectors (data sectors) is recorded. When the head element unit moves according to the servo data, data can be written to or read from the desired address. In the data reading process, a signal read from the magnetic disk by the head element unit is subjected to predetermined signal processing such as waveform shaping and decoding processing by a signal processing circuit and transferred to the host. Transfer data from the host is similarly processed by the signal processing circuit and then written to the magnetic disk by the head element unit.

  Write cache is known as one of methods for improving the performance of HDDs. In this method, the HDD caches the write command from the host and stores the write data in the write buffer. When storage of write data in the write buffer is completed, a command completion notification is sent to the host. In parallel with communication with the host, the HDD writes the write data stored in the write buffer to the magnetic disk. Upon receiving the command completion notification from the HDD, the host can issue the next command to the HDD.

  On the other hand, the write order of write data to the disk is optimized by using an algorithm such as RPO (Rotational Positioning Optimization) so that the processing in the HDD is made efficient and all the write data can be written to the magnetic disk in the shortest time. A technique is known (see, for example, Patent Document 1). The RPO, for example, when a command waiting for execution is executed, seek time from the start of seeking to the target track on the magnetic disk to reach it, and after reaching the target sector until the start of access to the target sector Expect disk rotation latency. The execution waiting command that minimizes the sum of the seek time and the rotation waiting time is selected as the next command to be executed.

JP 2003-122631 A

  SCSI and ATA standards are known as interface specifications between the host and the HDD, but FUA (Forced Unit Access) series commands are newly supported in the next standard, ATA standard 7. The HDD that has acquired the FUA command from the host must send a command completion notification to the host after writing the write data to the magnetic disk even if the write cache is in the ON state.

  When the write cache is ON, the cached write data may not be written to the magnetic disk for some reason such as power down. By issuing the FUA command, the host can confirm that the write data has been reliably written to the magnetic disk, thus preventing the critical data from being lost without being written to the magnetic disk. it can.

  On the other hand, for less important data, the host issues a normal command (non-FUA command) to instruct writing. This makes it possible to improve the performance of the host and HDD by using the write cache function. Thus, it can be said that the FUA command is a technique for achieving both safety of data writing and performance.

  However, when the write cache is ON, there are frequent cases where the HDD already caches a plurality of write commands at the timing when the FUA command is acquired. For example, if five write commands are already cached, a newly acquired FUA command may be executed after the completion of the five write commands. The HDD does not send a notification of completion of the FUA command to the host until the corresponding data is written to the magnetic disk. Therefore, the host must wait for completion of execution of five other write commands in addition to completion of the FUA command in order to issue the next command.

  Therefore, a new method for determining the execution order of write commands including FUA commands is required. In addition, in determining the execution order of FUA commands, it is also important to consider HDD throughput in addition to reducing the waiting time of the host.

  The present invention has been made in the background as described above, and improves the write command execution order determination method in the data storage device, and improves the safety of data and the data processing efficiency of the host and the data storage device. It aims to plan.

  The data storage device according to the first aspect of the present invention includes a write buffer for temporarily storing write data to be written to a medium, and information relating to a write command corresponding to each of the write data to be written to the medium. And a cache manager that determines the execution order of the write commands registered in the table, and the cache manager sends a command completion notification to the host after writing to the medium is completed. When the media write forced command is acquired, the execution order is determined at a higher priority than when the non-media write forced command is acquired. When the media write forced command is acquired, the execution order is determined at a higher priority than when the non-media write forced command is acquired, so the transmission timing of the command completion notification to the host can be advanced, and host processing Efficiency can be improved.

  According to a second aspect of the present invention, in the first aspect, the cache manager is cached and re-executes reordering for a write command executed after the media write forcing command. As a result, command execution after the media write forced command can be made more efficient.

  According to a third aspect of the present invention, in the first aspect, when the cache manager acquires a media write forced command from the host, the cache manager includes the media write forced command or the media write forced command and associates the media write forced command with each other. Set the specified write command to the first execution order. By using the first execution order, the completion notification of the media write forced command can be returned quickly.

  According to a fourth aspect of the present invention, in the third aspect, the series of write commands associated with each other has an address within a predetermined range, and the corresponding write data is written once. Is written on the medium. As a result, data can be efficiently written to the medium.

  According to a fifth aspect of the present invention, in the first aspect, the write buffer is a ring buffer type buffer, and the cache manager is the oldest cached write command. The execution order of the media write forced command is set after the write command. As a result, when a ring buffer is used, it is possible to achieve both buffer utilization efficiency and suppression of a decrease in processing efficiency due to a delay in notification of completion of the media write forced command.

  According to a sixth aspect of the present invention, in the fifth aspect, the cache manager sets the oldest write command to the first execution order among the cached write commands, and the media The execution order of the write forced command is set next to the oldest write command. As a result, the buffer utilization efficiency is improved and an early reply of the completion notification of the media write forced command is realized.

  A data storage device according to a seventh aspect of the present invention includes a write buffer for temporarily storing write data to be written to a medium, and a write corresponding to each of the write data stored in the write buffer. A table for registering information about commands and a cache manager for determining the execution order of write commands registered in the table, the cache manager notifying completion of commands after writing to the medium is completed When the media write forcing command to send to the host is acquired, the media write forcing is selected from a plurality of determination methods by referring to the write command already registered in the table in order to determine the execution order. A method for determining the execution order of commands is selected. When the media write forced command is acquired, an appropriate execution order determination method in the data storage device can be selected.

  According to an eighth aspect of the present invention, in the seventh aspect, the cache manager selects a method for determining the execution order of the media write forced command based on an empty area of the write buffer. Even if priority is given to the execution of the media write forced command to prevent delays in notification of completion of the media write forced command, if there is no free space in the buffer, the next write command will be waited and processing efficiency will be reduced. . Based on the buffer free space, it is possible to efficiently execute the cached write command to determine whether to give priority to area reservation or to give priority to the processing of the media write forced command. Achieving both reduction in processing efficiency due to delay in notification of compulsory command completion.

  According to a ninth aspect of the present invention, in the seventh aspect, the cache manager selects an execution order determination method for the media write forced command based on the number of cached write commands. Based on the number of cached write commands, it is possible to efficiently execute cached write commands to prioritize area allocation or media write forced command processing. The buffer utilization efficiency and the reduction in processing efficiency due to the delay in the completion notification of the media write forced command are both reduced.

  According to a tenth aspect of the present invention, in the seventh aspect, the cache manager is configured to determine the execution order of the media write forced command based on the write data amount corresponding to the cached write command. Select. Based on the number of sectors of the write command being cached, that is, the amount of data, whether the cached write command is executed efficiently to give priority to area allocation or media write forced command processing By making the determination, both buffer utilization efficiency and suppression of a decrease in processing efficiency due to a delay in notification of completion of the media write forced command are achieved.

  According to an eleventh aspect of the present invention, in the seventh aspect, the cache manager uses the execution order determination method of the media write forced command based on an expected execution time of a cached write command. select. In the writing to the media, the efficient order is selected, so that the processing of the media writing forced command is postponed, and it is possible to prevent the completion notification from being delayed beyond the allowable range.

  According to a twelfth aspect of the present invention, in the seventh aspect, the plurality of determination methods determine the execution order of the media write forced command with a higher priority than when the non-media write forced command is acquired. And a second determination method for determining the execution order without distinguishing between the media write forced command and the non-media write forced command. Thereby, the return timing of the command completion notification of the media write forced command can be made appropriate for the data storage device.

  In a thirteenth aspect of the present invention based on the twelfth aspect, the cache manager selects the first determination method on condition that the free area of the write buffer is equal to or greater than a reference value. This improves both the write efficiency of cached write commands and the suppression of degradation in processing efficiency due to the delay in notification of completion of media write forced commands.

  In a fourteenth aspect of the present invention, in the twelfth aspect, the cache manager selects the first determination method on condition that the number of cached write commands is smaller than a reference value. . As a result, it is possible to achieve both the improvement of the write efficiency of the cached write command and the suppression of the decrease in the processing efficiency due to the delay of the completion notification of the media write forced command.

  According to a fifteenth aspect of the present invention, in the twelfth aspect, the cache manager is configured on the condition that a write data amount corresponding to a write command already cached is smaller than a reference value. Select one determination method. As a result, it is possible to achieve both the improvement of the write efficiency of the cached write command and the suppression of the decrease in the processing efficiency due to the delay of the completion notification of the media write forced command.

  According to a sixteenth aspect of the present invention, in the twelfth aspect, the cache manager has a write data amount corresponding to the already cached write command and the media write forcing command smaller than a reference value. On the condition that the first determination method is selected. As a result, it is possible to achieve both the improvement of the write efficiency of the cached write command and the suppression of the decrease in the processing efficiency due to the delay of the completion notification of the media write forced command.

  According to a seventeenth aspect of the present invention, in the twelfth aspect, the cache manager is configured on the condition that a write command execution time expected until execution of a media write forced command is smaller than a reference value. A second determination method is selected. As a result, it is possible to achieve both the improvement of the write efficiency of the cached write command and the suppression of the decrease in the processing efficiency due to the delay of the completion notification of the media write forced command.

  According to an eighteenth aspect of the present invention, there is provided a write cache control method in a data storage device, wherein information relating to an acquired write command is registered, and write data to be written to a medium is written corresponding to the write command. -When determining the execution order of multiple write commands stored in the buffer and receiving a media write forced command that sends a command completion notification after writing to the media is completed, obtain a non-media write forced command The execution order is determined at a higher priority than the above case.

  According to a nineteenth aspect of the present invention, there is provided a write cache control method in a data storage device, wherein information relating to an acquired write command is registered, and write data to be written to a medium is written corresponding to the write command. When a media write forcing command is stored that is stored in a buffer and a command completion notification is transmitted after completion of writing to the medium, a write command that has already been registered to determine the execution order of write commands to be executed With reference to the command, an execution order determination method is selected from a plurality of determination methods.

  According to the present invention, the write command execution order determination method can be improved, and data safety and data processing efficiency of the host and the data storage device can be improved.

  Hereinafter, embodiments to which the present invention can be applied will be described. For clarity of explanation, the following description and drawings are omitted and simplified as appropriate. Moreover, in each drawing, the same code | symbol is attached | subjected to the same element and the duplication description is abbreviate | omitted as needed for clarification of description.

  Specific embodiments to which the present invention is applied will be described in detail with reference to the drawings. This embodiment is characterized by a method for determining the execution order of write commands (write order to media) when the data storage device is in the write cache ON state. In particular, the command completion notification to the host is sent to the media. It is characterized by a technique for determining the execution order of write commands performed after writing is completed. For easy understanding of the present invention, an outline of the overall configuration of a hard disk drive (HDD) as an example of a data storage device will be described first.

  FIG. 1 is a block diagram schematically showing the configuration of the HDD 1 according to the present embodiment. As shown in FIG. 1, an HDD 1 includes a magnetic disk 11 as an example of a medium (recording medium), a head element unit 12, an arm electronic circuit (arm electronics: AE) 13, a spindle motor in a sealed enclosure 10. (SPM) 14, voice coil motor (VCM) 15, and actuator 16.

  The HDD 1 includes a circuit board 20 that is fixed to the outside of the enclosure 10. On the circuit board 20, a read / write channel (R / W channel) 21, a motor driver unit 22, a hard disk controller (HDC) and an MPU integrated circuit (hereinafter referred to as HDC / MPU) 23, and a memory Each IC such as a RAM 24 as an example is provided. Each circuit configuration can be integrated into one IC, or can be divided into a plurality of ICs.

  Write data from the external host 51 is received by the HDC / MPU 23 and written to the magnetic disk 11 by the head element unit 12 via the R / W channel 21 and the AE 13. Read data stored in the magnetic disk 11 is read by the head element unit 12, and the read data is output from the HDC / MPU 23 to the external host 51 via the AE 13 and the R / W channel 21. The

  Next, each component of the HDD 1 will be described. The magnetic disk 11 is fixed to the SPM 14. The SPM 14 rotates the magnetic disk 11 at a predetermined speed. The motor driver unit 22 drives the SPM 14 according to control data from the HDC / MPU 23. The magnetic disk 11 of this example has recording surfaces for recording data on both sides, and a head element unit 12 corresponding to each recording surface is provided.

  Each head element unit 12 is fixed to a slider (not shown). The slider is fixed to the actuator 16. The actuator 16 is connected to the VCM 15, and swings about the rotation axis to move the head element unit 12 (and the slider) in the radial direction on the magnetic disk 11. The motor driver unit 22 drives the VCM 15 according to control data from the HDC / MPU 23.

  The head element unit 12 typically includes a recording head that converts an electric signal into a magnetic field according to recording data on the magnetic disk 11 and a reproducing head that converts a magnetic field from the magnetic disk 11 into an electric signal. Yes. One or more magnetic disks 11 may be provided, and the recording surface can be formed on one side or both sides of the magnetic disk 11.

  Subsequently, each circuit unit will be described. The AE 13 selects one head element unit 12 to which data access is performed from the plurality of head element units 12, and amplifies a reproduction signal reproduced by the selected head element unit 12 with a certain gain (preamplifier). To the R / W channel 21. Also, the recording signal from the R / W channel 21 is sent to the selected head element unit 12.

  The R / W channel 21 performs a write process on the data transferred from the host 51. In the write process, the R / W channel 21 code-modulates the write data supplied from the HDC / MPU 23, converts the code-modulated write data into a write signal, and supplies it to the AE 13. When data is supplied to the host 51, read processing is performed. In the read process, the R / W channel 21 amplifies the read signal supplied from the AE 13 to have a constant amplitude, extracts data from the acquired read signal, and performs a decoding process. Data to be read out includes user data and servo data. The decoded read data is supplied to the HDC / MPU 23.

  In the HDC / MPU 23, the MPU operates according to the microcode loaded in the RAM 24. As the HDD 1 starts up, the RAM 24 is loaded with data necessary for control and data processing from the magnetic disk 11 or ROM (not shown), in addition to the microcode operating on the MPU. The HDC / MPU 23 performs overall control of the HDD 1 in addition to necessary processing relating to data processing such as command execution order management, head element unit 12 positioning control, interface control, and defect management.

  The HDC / MPU 23 transmits the read data from the magnetic disk 11 acquired from the R / W channel 21 to the host 51. Read data from the magnetic disk 11 is temporarily stored in a read buffer in the RAM 24 and then transferred to the host 51 via the HDC / MPU 23. The write data from the host 51 is temporarily stored in the write buffer in the RAM 24 via the HDC / MPU 23 and then transferred to the magnetic disk 11 via the HDC / MPU 23 at a predetermined timing.

  Subsequently, the write cache process of the HDD 1 according to the present embodiment will be described. The HDD 1 of this embodiment is in a state in which the write cache is on, and can send a corresponding write command completion notification to the host 51 at the timing when the write data is stored in the write buffer. The host 51 can issue a new command to the HDD 1 when it receives the command completion notification from the HDD 1.

  In addition, the HDD 1 and the host 51 of the present embodiment support a write command for issuing a command completion notification from the HDD 1 to the host 51 after the writing to the recording disk 11 is completed. One of the most typical examples of such a write command is a FUA (Forced Unit Access) series command standardized by ATA Standard 7. The HDD 1 that has acquired the FUA command from the host 51 transmits a command completion notification to the host 51 after writing the write data to the magnetic disk 11 even if the write cache is in an ON state.

  In ATA Standard 7, FUA commands include WRITE DMA EXTEND FUA and WRITE DMA MULTIPLE EXTEND FUA. In the following description, the present embodiment will be described using the FUA command as an example. In the following, a write command not particularly specified is a command other than FUA (non-FUA command), and a command completion notification is sent before the write data is stored in the write buffer and then written to the magnetic disk 11. It is sent to the host 51.

  In the following, determination of command execution order when the HDD 1 obtains a FUA command from the host 51 at the timing when an unexecuted write command (data writing to the magnetic disk 11 is incomplete) is queued will be described. . When only non-FUA commands are cached, the HDD 1 uses an algorithm such as RPO (Rotational Positioning Optimization) to optimize the order of writing write data to the magnetic disk 11 in order to perform efficient internal processing. Turn into.

  When the FUA command is issued, the HDD 1 cannot return a command completion notification until the execution of the command is completed, that is, until the writing of write data to the magnetic disk 11 is completed. The interface between is busy. For this reason, when the FUA command is acquired from the host 51, it is preferable to re-determine (reorder) the execution order of the write command at a higher priority than when the acquired write command is a non-FUA command.

  In one preferred example, when receiving an FUA command from the host 51, the HDD 1 executes the FUA command with the highest priority over other cached write commands. That is, the FUA command newly issued first is executed before all the cached write commands.

  FIG. 2 shows an example in which FUA commands are executed in the first order. As shown in FIG. 2A, the HDD 1 acquires five write commands WR from the host 51 in the order of WR1-WR2-WR3-WR4-WR5. Thereafter, as shown in FIG. 2B, reordering of the write command is executed. Since each write command is a non-FUA command, reordering is executed according to a conventional algorithm such as RPO. In the example of FIG. 2B, the execution order is set in the order of WR2-WR3-WR1-WR5-WR4.

  As shown in FIG. 2C, the HDD 1 receives a FUA command from the host 51 before executing these write commands RW. Then, as shown in FIG. 2D, the HDD 1 rearranges the command execution order with the FUA command as the highest priority. That is, the execution order of the newly acquired FUA command is first. Thus, setting the execution order of the FUA commands to the first order among the unexecuted write commands can suppress a reduction in processing efficiency of the host 51 due to a delay in replying the command completion notification.

  Here, preferably, the FUA command is set at the head of the execution order, and the reordering of the execution order is re-executed for another write command WR to be executed thereafter. In the example of FIG. 2E, the execution order of the write command WR2 is changed from the second to the fifth in accordance with the execution order of the FUA command being set to the first (first).

  Considering the positional relationship between the head element unit 12 and the magnetic disk 11 after executing the FUA command, the original item order (WR2-WR3-WR1-WR5-WR4) is not always the most efficient execution order. . Therefore, the reordering process is executed again with the last LBA (Logical Block Address) of the FUA command as a reference, and the execution order is rearranged. As a result, more efficient write command execution processing can be realized.

  3 and 4 show another preferable example in which the execution order of the acquired FUA commands is determined at a higher priority than when a non-FUA command is acquired. In this example, when the FUA command is included in a series of commands associated with each other, the execution order is set with the highest priority by using the series of commands as a set. Specifically, the HDD 1 supports a write operation called sequential write.

  In the sequential write, if the LBA (write data write LBA) in the cached write command is within a predetermined range, the write data corresponding to the write command is stored once. Writing to the magnetic disk 11 by a writing process. By executing writing of adjacent write data to the magnetic disk 11 as a single write process, the overhead of the write process can be reduced and the write time can be shortened.

  First, with reference to FIG. 3, the relationship of write commands for executing sequential write will be described. Sequential write is executed when the difference between the last LB of the first write command and the first LAB of the second write command is less than or equal to the reference value between the two write commands. For example, as shown in FIG. 3A, when the LBAs of two write commands are continuous, the HDD 1 executes sequential write of these two write commands. In the example of FIG. 3A, the first LBA of the first write command WR1 is 200h, and the final LBA is 27Fh. On the other hand, the first LBA of the second write command WR2 is 280h, and the final LBA is 2FFh. In this case, the HDD 1 writes write data to the magnetic disk 11 in a single write process in sectors from LBA 200h to 2FFh.

  Alternatively, as shown in FIG. 3B, when the LBAs of two write commands overlap, the HDD 1 executes sequential write. That is, the first LBA of the second write command WR2 is included in the LBA area of the first write command WR1 (area between the first LBA and the last LBA). In the example of FIG. 3B, the first LBA of the first write command WR1 is 200h, and the final LBA is 2BFh. On the other hand, the first LBA of the second write command WR2 is 280h, and the final LBA is 2FFh. In this case, the HDD 1 executes the Discard process for erasing old data (the write data of the first write command) in the address portion (280h to 2BFh) where the LAB overlaps, and then writes the write data from the LBA 200h to 2FFh. Data is written to the magnetic disk 11 by a single writing process.

  Alternatively, as shown in FIG. 3C, even when the LBAs of two write commands are not continuous or do not overlap, if the interval is equal to or smaller than a predetermined reference value, The HDD 1 performs sequential write. In the example of FIG. 3C, the first LBA of the first write command WR1 is 200h, and the final LBA is 27Fh. On the other hand, the first LBA of the second write command WR2 is 290h, and the final LBA is 2FFh. For example, if the reference value is 40h, these two write commands are written in batches with sequential write. In this case, the HDD 1 skips from LBA 280h to 28Fh without writing data to the magnetic disk 11.

  FIG. 4 shows an execution order determination process when the FUA command is received from the host 51 in the HDD 1 that supports sequential write. In this example, when the acquired FUA command and the write command for executing the sequential write are queued, the HDD 1 determines the execution order with the sequential write as the highest priority.

  Specifically, as shown in FIG. 4A, the HDD 1 acquires five write commands WR from the host 51 in the order of WR1-WR2-WR3-WR4-WR5. Subsequently, as shown in FIG. 4B, the reordering of the write command is executed, and the execution order is set in the order of WR2-WR3-WR1-WR5-WR4. As shown in FIG. 4C, the HDD 1 receives a FUA command from the host 51 before executing these write commands WR. Then, as shown in FIG. 4D, the HDD 1 searches for a write command that has a relationship between the FUA command and the sequential write. In this example, the write command WR3 is related to the FUA command and the sequential write, and therefore these are associated.

  Further, as shown in FIG. 4 (e), the FUA command and the command group having the sequential write relationship are given the highest priority, and the execution order is set first. Among command groups, the order is determined according to the sequential write order. In this example, the write data of the write command WR3 is written before the FUA command. As described above, when a FUA command is received, a command group having a sequential write relationship with the FUA command is given priority, thereby simultaneously suppressing the command completion notification delay for the host 51 and increasing the processing efficiency in the HDD 1. it can. As described with reference to FIG. 2, after determining the execution order of command groups associated with each other with the highest priority, the execution order of subsequent write commands is reset based on these final LBAs. Is preferred.

  FIG. 5 shows another preferred example in which the execution order of the acquired FUA commands is determined at a higher priority than when a non-FUA command is acquired. In this example, a ring buffer system is used as a write buffer for temporarily storing write data. In the ring buffer method, write data from the host is stored in time series from the previous address to the subsequent address. When the write buffer is used up to its end position, it returns to the tip position of the write buffer. At this time, if the write data stored at the tip position has not yet been written to the magnetic disk 11, the write buffer cannot store new write data, so the next write data is stored in the host. Cannot receive from.

  For this reason, in the ring buffer method, it is preferable that the oldest write data among the uncompleted data written to the magnetic disk 11 is written to the magnetic disk 11 earlier. That is, it is preferable to preferentially execute the oldest unexecuted command. In this example, the HDD 1 executes the FUA commands in an order after the oldest unexecuted write command.

  This will be specifically described. As shown in FIG. 5A, it is assumed that the HDD 1 receives write commands from the host in the order of WR1-WR2-WR3-WR4-WR5. FIG. 5B shows the execution order of each command when reordering is performed according to the RPO. In this example, WR2-WR3-WR1-WR5-WR4 are set in this order. Here, as shown in FIG. 5C, the HDD 1 receives the FUA command from the host 51. According to the example of FIG. 2, the FUA command is set in the first execution order as shown in FIG.

  However, as described above, in the ring buffer type write buffer, execution of the oldest command with priority leads to securing a free space in the buffer. Therefore, in this example, as shown in FIG. 5E, the execution order of the write command WR1 first obtained from the host 51 is first set. Further, the FUA command is set to the next execution order of the write command WR1. As a result, it is possible to suppress an increase in the waiting time of the host 51 due to a delay in the execution of the FUA command while securing a free area in the ring buffer method. Note that, from the viewpoint of early execution of the FUA command, it is preferable to set the FUA command execution order second, but the FUA command is not limited to this as long as the FUA command is prioritized.

  As described above, it is preferable to preferentially execute the FUA command from the viewpoint of shortening the waiting time of the host 51. However, in processing control of the HDD 1, it may be preferable to prioritize other characteristics such as the throughput of the HDD 1. Therefore, it is preferable that the HDD 1 includes a plurality of different execution order determination methods, and when an FUA command is received, an optimal determination method is selected from the plurality of determination methods according to a predetermined criterion. As described with reference to FIG. 2-4, the execution order determination method has a higher priority than the non-FUA command when a FUA command is received, such as a determination method that determines the execution order with the FUA command as the highest priority. In addition to the method for determining the execution order, a method for determining the execution order without distinguishing FUA commands from non-FUA commands, such as RPO that does not consider FUA commands, is included.

  For example, the HDD 1 acquires five write commands WR from the host 51 in the order of WR1-WR2-WR3-WR4-WR5 as shown in FIG. Subsequently, as shown in FIG. 6B, reordering such as RPO is executed, and the execution order is set in the order of WR2-WR3-WR1-WR5-WR4. Further, as shown in FIG. 6C, the HDD 1 receives the FUA command from the host 51 before executing these write commands WR.

  Here, the HDD 1 selects an execution order determination method according to a predetermined condition. For example, as shown in FIG. 6D, the HDD 1 determines the execution order of the write commands according to the RPO regardless of the FUA command. In the example of FIG. 6D, the FUA command is set at the end of the six write commands. Alternatively, as shown in FIG. 6E, reordering is executed with priority given to the FUA command. In the example of FIG. 6E, the newly acquired FUA command is set in the first execution order.

  In the following, an example will be described that includes two methods: a FUA priority determination method that prioritizes FUA and a reordering method (through throughput priority priority) of HDD 1 regardless of the FUA command (priority in write efficiency). Some criteria for the HDD 1 to select the execution order determination method will be described.

  In one preferred example, the HDD 1 selects an execution order determination method based on the free capacity of the write buffer. Specifically, when the free capacity of the write buffer is large, the HDD 1 selects the FUA priority determination method. On the other hand, if the number is low, the throughput priority determination method is selected.

  This will be specifically described with reference to the flowchart of FIG. The HDD 1 reorders a plurality of unexecuted write commands not including the FUA command according to the throughput priority determination method (S11). Thereafter, when the HDD 1 acquires the FUA command (S12), the HDD 1 calculates the free capacity of the write buffer and determines whether the free capacity is less than a predetermined reference value (S13). If it is less than the reference value, the HDD 1 gives priority to the throughput of the HDD 1 without giving priority to the FUA command, and selects a throughput priority determination method (S14).

  On the other hand, when the free capacity of the write buffer is equal to or larger than the reference value, the HDD 1 prioritizes the FUA command and determines the command execution order (S15). That is, the HDD 1 selects the FUA priority determination method and determines the execution order of the write commands. As described above, when the FUA command is acquired, by changing the execution order determination method according to the free capacity of the write buffer, when the write buffer is sufficiently free, the interface with the host 51 is made ready and the subsequent Receive commands. On the other hand, when there is no room in the write buffer, writing to the magnetic disk 11 is prioritized so that the write buffer can be processed quickly.

  In another preferred example, the HDD 1 selects an execution order determination method based on the number of write commands already cached (five in the example of FIG. 6). Specifically, when the number of cached write commands is large, the HDD 1 selects a throughput priority determination method. On the other hand, if the number is small, the FUA priority determination method is selected.

  This will be specifically described with reference to the flowchart of FIG. The HDD 1 reorders a plurality of unexecuted write commands not including the FUA command according to the throughput priority determination method (S21). Thereafter, when the HDD 1 obtains a FUA command (S22), it calculates the number of cached write commands and determines whether the number is less than a predetermined reference value (S23). In the example of FIG. 6, the number of write commands is five. If the number of write commands is equal to or greater than the reference value, the HDD 1 selects the throughput priority determination method by giving priority to the throughput of the HDD 1 without giving priority to the FUA command (S24).

  On the other hand, if the number of cached write commands is less than the reference value, the HDD 1 prioritizes the FUA command and determines the command execution order (S25). That is, the HDD 1 selects the FUA priority determination method and determines the execution order of the write commands. In this way, when the FUA command is acquired, when the number of write commands cached in the write buffer is small by changing the execution order determination method according to the number of write commands, the interface is ready and the subsequent command is changed. When there are many cached write commands, the writing to the magnetic disk 11 is prioritized and processing is performed so that the write buffer is quickly cleared.

  In addition to the number of cached write commands, selecting the execution order determination method based on the number of write data sectors corresponding to the cached write command is another preferred example. When the number of sectors is large, the HDD 1 selects a throughput priority determination method. On the other hand, if the number is small, the FUA priority determination method is selected. When the HDD 1 obtains the FUA command, the HDD 1 sums up the number of data sectors (write data amount) of the cached write command and determines whether it is equal to or greater than the reference value. If it is equal to or greater than the reference value, the HDD 1 selects the throughput priority determination method, and if it is less than the reference value, the HDD 1 selects the FUA priority determination method. The calculation of the number of sectors can include newly acquired FUA command write data. In this way, by changing the execution order determination method according to the number of data sectors to be written, when the write data cached in the write buffer is small, the interface is ready and the subsequent command is received soon. When there is a large amount of write data being processed, the writing to the magnetic disk 11 is prioritized and processing is performed so that the write buffer becomes free early.

  Alternatively, selecting the execution order determination method based on the execution time of the cached write command is another preferable example. If the execution time is shorter than the reference value, the HDD 1 selects the throughput priority determination method. On the other hand, if the execution time is longer than the reference value, the FUA priority determination method is selected. When the HDD 1 acquires the FUA command, the HDD 1 calculates the execution time of each write command to be executed before the FUA is executed after reordering with priority on throughput. The execution time includes a seek time, a rotation waiting time, and a writing time. These execution times are totaled and it is determined whether the value is equal to or greater than a reference value. If it is equal to or greater than the reference value, the HDD 1 selects the FUA priority determination method, and if it is less than the reference value, the HDD 1 selects the throughput priority determination method.

In calculating the execution time, it is preferable to multiply the seek time and the write time by a coefficient that takes into account the retry process due to an error and variations at the time of execution. In this way, it is possible to prevent the FUA command completion notification from becoming too late by changing the execution order determination method according to the execution time of writing. In other words, if an efficient order is selected for writing to the media and the processing of the FUA command is postponed, the FUA command may time out in some cases. By predicting the execution time and notifying the completion of the FUA command within the specified time, it is possible to prevent a timeout of the FUA command. Note that when the FUA command is acquired, the execution time reference value may be determined using only the execution time of the cached write command without including the FUA command.

  In this embodiment, the HDC / MPU 23 performs the above-described write cache control. The details of the write cache control process of the HDC / MPU 23 will be described. FIG. 9 is a block diagram showing a configuration relating to write cache control in the HDD 1. The HDC / MPU 23 includes a host interface 231, a drive interface 232, and a memory manager 233 as hardware configurations. With the micro code running on the MPU, the MPU can function as a host interface manager 234, a command execution manager 235, and a cache manager 236. The memory RAM 24 temporarily stores commands and data, and functions as a write buffer 241 and a write cache table 247.

  The host interface 231 executes actual data transmission processing with the host 51 and functions as a data transmission unit. The drive interface 232 performs actual data input / output processing with the magnetic disk 11 (or with the read / write channel 21). The memory manager 233 controls data storage in the RAM (memory) 24, and performs mediation processing of commands and user data (write / read data) between the RAM 24 and other functional blocks in the HDC / MPU 23.

  The host interface manager 234 functions as a data transmission control unit that manages the host interface 231, and sends and receives predetermined notifications and commands to and from the host interface 231. The cache manager 236 performs rescheduling of the queued write commands and determines an appropriate command execution order. The command execution manager 235 controls the execution of commands according to the order determined by the cache manager 236. Further, by controlling the drive interface 232, data writing to and data reading from the magnetic disk 11 are controlled.

  FIG. 10 shows an example of the write cache table 247. In FIG. 10, the command number is an identifier of each command in the write cache table 247. The command type indicates whether each write command is a FUA command. The data length is the data length of the write data corresponding to the write command, and the buffer start position is the write data storage start position in the write buffer. The order of writing to the magnetic disk is the order in which each write data is written to the magnetic disk 11, and in the example shown in FIG.

  The operation of the write command reordering process will be described. When the host command 231 receives a write command and write data from the host 51, the memory manager 233 stores the write data in the write buffer 241. The cache manager 236 records the write command in the write cache table 247. That is, the host interface 231 notifies the cache manager 236 of the write command reception, the command type, the LBA (Logical Block Address), and the data length via the host interface manager 234.

  When the cache manager 236 receives the write command, the cache manager 236 records data relating to the write command in each entry of an empty record in the write cache table 247. At the timing of registration in the write cache table 247, the execution order of the write commands can be, for example, the last order.

  If the received write command is a non-FUA command, the cache manager 236 requests the host interface manager 234 to return a command completion notification to the host 51. The host interface manager 234 returns a command completion notification to the host 51 via the host interface 231. Further, the cache manager 236 executes reordering of the execution order of each write command, that is, the order of writing the write data to the magnetic disk 11. At this time, the cache manager 236 determines the execution order according to the execution order determination method described above.

  On the other hand, if the received write command is a FUA command, the cache manager 236 determines the execution order of each write command without returning a command completion notification to the host 51. At this time, as described above, the cache manager 236 executes the reordering of the write command according to the execution order determination method depending on the predetermined condition.

  Thereafter, the cache manager 236 notifies the command execution manager 235 of a write command to be executed. The command execution manager 235 instructs the drive interface 232 to write the write data of the execution write command to the magnetic disk 11. The drive interface 232 acquires the write data from the write buffer 241 via the memory manager 233 and transfers the write data for writing to the magnetic disk 11.

  When the writing to the magnetic disk 11 is completed, the drive interface 232 notifies the command execution manager 235 that the writing has been completed. When the execution of the FUA command is completed, the command execution manager 235 requests the host interface manager 234 to send a completion notification of the FUA command to the host 51. In response to the request, the host interface manager 234 returns a FUA command completion notification to the host 51.

  In addition, said description demonstrates embodiment of this invention and this invention is not limited to said embodiment. A person skilled in the art can easily change, add, and convert each element of the above-described embodiment within the scope of the present invention. For example, the write buffer control of this embodiment can be applied to a data storage device that uses media other than magnetic disks.

1 is a block diagram schematically showing an overall configuration of an HDD according to an embodiment. FIG. 10 is a diagram illustrating a method for reordering a FUA command with priority when an FUA command is received in the HDD according to the embodiment. It is a figure explaining the technique of the sequential write in HDD concerning this embodiment. FIG. 6 is a diagram illustrating a method of reordering using sequential write when an FUA command is received in the HDD according to the present embodiment. In this embodiment, it is a figure which shows the reordering method in case HDD which uses a ring buffer system receives a FUA command. FIG. 10 is a diagram illustrating an example of selecting one from a plurality of reordering methods when an FUA command is received in the HDD according to the present embodiment. 6 is a flowchart showing an example of selecting one from a plurality of reordering methods when an FUA command is received in the HDD according to the embodiment. 10 is a flowchart showing another example of selecting one from a plurality of reordering methods when an FUA command is received in the HDD according to the embodiment. 3 is a block diagram showing a logical configuration related to write cache control in the HDD according to the present embodiment. FIG. 4 is a diagram illustrating an example of a write cache table for write cache control in the HDD according to the present embodiment. FIG.

Explanation of symbols

10 enclosure, 11 magnetic disk, 12 head element,
13 Arm Electronics, 14 Spindle Motor,
15 voice coil motor, 16 actuator, 20 circuit board 21 read / write channel, 22 motor driver unit, 51 host 231 host interface, 232 drive interface,
233 Memory manager, 234 Host interface manager 235 Command execution manager, 236 Cache manager

Claims (19)

A write buffer for temporarily storing write data to be written to the medium;
A table for registering information on a write command corresponding to each write data to be written to the medium;
A cache manager that determines the execution order of write commands registered in the table,
When the cache manager obtains a media write forcing command that sends a command completion notification to the host after writing to the media is completed, it determines the execution order at a higher priority than when it obtains a non-media write forcing command. To
Data storage device.   The data storage device according to claim 1, wherein the cache manager is cached and re-executes reordering for a write command executed after the media write forcing command.   When the cache manager acquires a media write forced command from the host, the cache manager sets the media write forced command or a series of write commands including the media write forced command and associated with each other in the first execution order. The data storage device according to claim 1.   4. The data storage device according to claim 3, wherein the series of write commands associated with each other has an address in a predetermined range, and corresponding write data is written to the medium in one write process. . The write buffer is a ring buffer type buffer,
The cache manager sets the execution order of the media write forced command after the oldest write command among the cached write commands.
The data storage device according to claim 1.   The cache manager sets the oldest write command among the cached write commands to the first execution order, and sets the execution order of the media write forced command next to the oldest write command. The data storage device according to claim 5, wherein A write buffer for temporarily storing write data to be written to the medium;
A table for registering information on write commands corresponding to each of the write data stored in the write buffer;
A cache manager that determines the execution order of write commands registered in the table,
When the cache manager obtains a media write forced command for sending a command completion notification to the host after writing to the media is completed, the write command already registered in the table is determined to determine the execution order. To select an execution order determination method of the media write forcing command from a plurality of determination methods,
Data storage device.   8. The apparatus according to claim 7, wherein the cache manager selects an execution order determination method of the media write forced command based on an empty area of the write buffer.   8. The apparatus according to claim 7, wherein the cache manager selects an execution order determination method of the media write forced command based on the number of cached write commands.   8. The apparatus according to claim 7, wherein the cache manager selects an execution order determination method for the media write forced command based on a write data amount corresponding to a cached write command.   8. The apparatus according to claim 7, wherein the cache manager selects a method of determining an execution order of the media write forced command based on an expected execution time of a write command.   The plurality of determination methods include a first determination method for determining an execution order of media write forced commands at a higher priority than when a non-media write forced command is acquired, a media write forced command, and a non-media write forced command; The data storage device according to claim 7, further comprising: a second determination method that determines an execution order without distinguishing between the two.   The data storage device according to claim 12, wherein the cache manager selects the first determination method on condition that an empty area of the write buffer is equal to or greater than a reference value.   The data storage device according to claim 12, wherein the cache manager selects the first determination method on condition that the number of cached write commands is smaller than a reference value.   The data storage according to claim 12, wherein the cache manager selects the first determination method on condition that a write data amount corresponding to a write command already cached is smaller than a reference value. apparatus.   The cache manager selects the first determination method on condition that a write data amount corresponding to a write command already cached and the media write forced command is smaller than a reference value. Item 13. A data storage device according to Item 12.   The data according to claim 12, wherein the cache manager selects the second determination method on condition that a write command execution time expected until execution of a media write forced command is smaller than a reference value. Storage device. A write cache control method in a data storage device, comprising:
Register information about the acquired write command,
In response to the write command, the write data to be written to the medium is stored in the write buffer,
In determining the execution order of a plurality of registered write commands, when receiving a media write forced command that sends a command completion notification after completion of writing to the media, higher priority than when receiving a non-media write forced command Determine the execution order in degrees,
Write cache control method. A write cache control method in a data storage device, comprising:
Register information about the acquired write command,
In response to the write command, the write data to be written to the medium is stored in the write buffer,
When receiving a media write forced command for sending a command completion notification after the completion of writing to the media, in order to determine the execution order of the write commands to be executed, refer to the write commands already registered, Select execution order determination method from multiple determination methods,
Write cache control method.

Images