JP2007200433A - Data-accumulating device and its control method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To make it possible to quickly execute processing, when a plurality of commands are issued in the same period, in case that data are accumulated by using a plurality of HDDs as in an AV server. <P>SOLUTION: When at least a first and second commands are successively received as the commands instructing data-writing to and/or data-reading from a plurality of hard disk drives when executing control for writing to or reading from a data-accumulation device which accumulates data by using a plurality of hard disk drives, instructions corresponding to the first command and instructions corresponding to the second command are made to be executed by the plurality of the hard disk drives. When the instructions corresponding to the first command is completed in each of the hard disk drives, the instructions corresponding to the second command is made to be executed with individual timing set for respective hard disk drives. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to an apparatus for storing stream data using an HDD (Hard Disk Drive) and a control method applied to the apparatus.

  In recent years, in broadcasting stations and post-production, an HDD array unit that has a large capacity by installing a plurality of HDDs as a storage for storing AV (audio / video) data such as video data and audio data that are stream data. in use.

  For example, such an HDD array unit is also used in an AV server used as an editing system or a transmission system in a broadcasting station. The AV server has a plurality of input / output ports for AV data input / output, and transfers AV data between these input / output ports and the HDD array unit at high speed.

  This AV server is premised on continuous operation for 24 hours and 365 days, and video and audio are not allowed to be interrupted or delayed when the program is on the air. Therefore, high operational reliability and real-time performance are required. .

  However, the HDD is a device whose operation reliability and real-time performance (response performance) are not so high. Therefore, in the AV server, the HDD array unit has a RAID (Redundant Arrays of Inexpensive Disks) configuration with redundancy, and further supports various functions for dealing with failures. Such functions include, for example, correction of data by parity, reconstruction of data by rebuild, data reassignment processing (correction of data in another HDD even if a response delay occurs in one HDD) And a reduction in MTTR (Mean Time To Repair) by installing a spare HDD.

Patent Document 1 describes an example of a data recording / reproducing apparatus that stores AV data having a RAID configuration.
JP 2005-182658 A

  By the way, when AV data that is stream data is read from an AV server and played back, the AV data must be read out in real time so that the AV data can be played back without interruption. When a command is received from the outside, the processing time for the command is preferably as short as possible. However, there are cases where a plurality of commands are continuously received in the storage unit provided with the HDD at the same time, and in such a case, there is a problem that it takes time to process the commands.

  FIG. 7 is a diagram showing an example of conventional processing in the HDD array unit having a RAID configuration when a plurality of commands are issued simultaneously. In this example, it is an example of an HDD array unit composed of 14 HDDs, and it is assumed that the first command and the second command are issued simultaneously. The first command is, for example, a command for instructing writing of certain data, and the second command is, for example, a command for instructing reading of certain data. At this time, the control unit that controls the HDD array unit first sends a first command to each HDD, and executes processing (write processing) Ca1 to Ca14 for each command. Each HDD has a different processing time, such as a seek time for data writing, depending on the state of each drive, so that the processing time (processing Ca2 in the example of FIG. 7) that takes the most processing time is completed. At this stage, the second command, which is the next command, is sent to each HDD. By sending the second command to each HDD, processing (reading processing) Cb1 to Cb14 for each command is executed.

  Even when the second command is issued, the processing time such as the seek time for data reading differs depending on the state of each drive, so the time is different and the processing that takes the longest processing time (in the example of FIG. 7) When the process Cb1) is completed, the second command process ends.

  When a plurality of commands are overlapped at the same time as described above, each command is processed in order, and there is a problem that it takes time to complete the command processing. In particular, in the case of an AV server that records AV data, it is necessary to perform very high time control so that the reading process can be completed within the real time of the AV data at the time of data reproduction. It is necessary to prevent.

  In the present invention, in view of the above points, for example, when data is stored using a plurality of HDDs such as an AV server, processing when a plurality of commands are issued at the same time can be performed quickly. For the purpose.

  The present invention provides a command for instructing writing and / or reading of data to and from a plurality of hard disk drives when performing control of writing or reading to a data storage device that accumulates data using a plurality of hard disk drives. When at least the first command and the second command are received in succession, when the instructions corresponding to the first command are completed in the respective hard disk drives constituting the plurality of hard disk drives, the respective hard disk drives The instruction corresponding to the second command is performed at an individual timing set for each.

  According to the present invention, an instruction corresponding to the second command is issued at an individual timing after the instruction corresponding to the first command is completed in each of the plurality of hard disk drives. As a result, each hard disk drive is processed in the shortest time without waiting time, and the processing time is reduced overall.

  According to the present invention, even if a plurality of commands are generated at the same time, each hard disk drive is processed in the shortest time without waiting time, and the processing time is reduced overall. Therefore, for example, it is possible to effectively prevent a delay in processing when recording and reproducing stream data such as AV data, and it is possible to perform suitable control that prevents a delay in data read as an AV server.

Hereinafter, an example in which an embodiment of the present invention is applied to storage of an AV server used in a broadcasting station will be specifically described with reference to FIGS.
FIG. 1 is a block diagram showing a configuration of a storage unit of an AV server to which the present embodiment is applied. The storage unit 1 includes an FC (Fibre Channel) controller 2, a recording / playback processing system 3, a cache memory 4, an HDD controller 5, a CPU 6, a DRAM 7, and a nonvolatile SRAM 8. Yes. The recording / playback processing system 3 and the HDD controller 5 are connected by a PCI-X standard bus 9.

  The storage unit 1 includes 15 HDDs 10 (HDDs 10 (1) to 10 (15)). Of these HDDs 10, 10 HDDs 10 (1) to 10 (10) are AV data HDDs, and 4 HDDs 10 (11) to 10 (14) are error correction HDDs. One HDD 10 (15) is a spare HDD.

  As these HDDs 10, HDDs manufactured and sold by HDD manufacturers as standard products (HDDs for applications such as external storage devices for computers) are used.

  The FC controller 2 is an interface for exchanging AV data with the outside via a fiber channel. The recording / playback processing system 3 is configured by an FPGA, which is a programmable LSI, and processes AV data to be recorded on the HDD 10 and AV data played back from the HDD 10.

  The cache memory 4 is a DIMM (Dual Inline Memory Module) standard SDRAM for buffering AV data. The HDD controller 5 is a SATA controller when, for example, an HDD that supports SATA (serial ATA) is used as the HDD 10.

  The CPU 6 controls the entire storage unit 1. The DRAM 7 is a main memory of the CPU 6. This DRAM 7 is also used for a command queue (described later) for temporarily storing commands. The nonvolatile SRAM 8 is used not only for the CPU 6 to save log data but also for storing an HDD information table and the like.

  When AV data is recorded (written) in the storage unit 1, the AV data received by the FC controller 2 is sent to the recording / playback processing system 3. The recording / playback processing system 3 generates an error correction code (Reed-Solomon code) from this AV data. Then, the AV data to which this code is added is recorded in the respective AV data HDDs 10 (1) to 10 (10), 10 systems of data (AV data portion), and the error correction HDD 10 (11). 10 to 14 (14), the data is striped into four lines of data (the Reed-Solomon code portion). Then, a total of 14 lines of data are sent to the HDDs 10 (1) to 10 (14) via the HDD controller 5 while being buffered in the cache memory 4, and distributed to the HDDs 10 (1) to 10 (14). RAID configuration for recording data. The HDD controller 5 is also provided with an HDD command queue (described later) for storing instructions for each HDD for each HDD 10.

  When reproducing (reading) AV data from the storage unit 1, AV data and Reed-Solomon codes read from the HDDs 10 (1) to 10 (14) are sent to the recording / reproducing processing system 3 via the HDD controller 5. It is done. The recording / playback processing system 3 performs destriping (returns to the state before striping) while buffering these data and codes in the cache memory 4. Then, the AV data is error-corrected using the Reed-Solomon code, and the error-corrected AV data is transmitted from the FC controller 2 to the outside.

  Further, the recording / reproduction processing system 3 starts rebuilding when a failure or a response delay occurs in any of the HDDs 10 (1) to 10 (14) during data reproduction. That is, the data in the failed HDD among the HDDs 10 (1) to 10 (14) is restored from the data read from the remaining HDDs 10 (1) to 10 (14) using the Reed-Solomon code, and the restoration is performed. Data is recorded in the spare HDD 10 (15).

  FIG. 2 is a diagram showing a system configuration of an AV server installed in a broadcasting station using the storage unit 1 of FIG. In this AV server, a plurality of storage units 1, a plurality of input / output processor units 11, and a relay terminal 12 are connected by an FC (Fibre Channel) switch 13. In addition, each input / output processor unit 11, relay terminal 12, management terminal 14, and maintenance terminal 15 are connected by Ethernet 16 (Ethernet: Ethernet is a registered trademark).

  The input / output processor unit 11 has a plurality of (for example, six) input / output ports, and is connected to the outside in a synchronous transmission format such as SDI (Serial Digital Interface) or an asynchronous transmission format. To input / output AV data.

  The input / output processor unit 11 encodes (compresses) AV data input to the input / output port using a predetermined encoding method, and transfers the AV data to the storage unit 1 via the FC switch 13. The input / output processor unit 11 decodes (decompresses) the data transferred from the storage unit 1 via the FC switch 13 and outputs the data from the input / output port.

  The configuration of the input / output processor unit of the general AV server is well known, and the configuration of the input / output processor unit of the AV server to which the present example is applied may be such a general configuration, and the detailed description thereof is omitted. To do.

  The relay terminal 12 is connected to a plurality of editing terminals 18 via a network 17, exchange of AV data between the storage unit 1 and the editing terminal 18, and the input / output processor unit 11 and the management terminal 14. And relaying of information between the editing terminal 18 and the terminal 18 for editing.

  The management terminal 14 holds information such as a storage address of AV data in the storage unit 1 and a file name / attribute of the AV data, and requests from the input / output processor unit 11 and the relay terminal 12 via the Ethernet 16. In response, the information is sent to the input / output processor unit 11 and the relay terminal 12.

  The maintenance terminal 15 is a terminal operated by maintenance personnel. The maintenance terminal 15 displays log data generated and stored by each storage unit 1 and each input / output processor unit 11 to analyze the state of the AV server, The replacement of the HDD or the like is determined based on a warning displayed when the HDD fails. The maintenance terminal 15 is also used to update an HDD information table in the storage unit 1 as will be described later.

  Next, a configuration example viewed from the command transmission state in the storage unit 1 of this example will be described with reference to FIG. When a command for instructing writing or reading of data is supplied to the CPU 6 from the outside (such as an input / output processor), the command is stored in the command queue 6a in the CPU 6 and an instruction corresponding to the stored command is stored in each HDD 10. (1) to 10 (14). A plurality of commands can be stored in the command queue 6a. The operations of the two timers (first timer 6b and second timer 6c) prepared in the CPU 6 are controlled according to the storage state of the command. The processing using the two timers will be described in the description of the operation described later using a flowchart.

  When a command transferred from the outside is stored in the command queue 6 a in the CPU 6, operation instructions corresponding to the command are individually given to the HDDs 10 (1) to 10 (14) via the HDD controller 5. . At this time, the HDD controller 5 includes HDD command queues 5a to 5n for individually storing commands (operation instructions) for each of the HDDs 10 (1) to 10 (14), and the HDD command queues 5a to 5n. Stores commands according to the states of the corresponding HDDs 10 (1) to 10 (14).

  Next, an example of the operation in response to the issue of such a command will be described with reference to the flowcharts of FIGS. First, an example of operation when a write command or a read command is transferred to the CPU 6 from the outside is shown in the flowchart of FIG. The CPU 6 waits for command input (step S11), determines whether a command has been input (step S12), and if a command is input, whether there is a command currently being executed for each HDD. Judgment is made (step S13). If there is no command being executed, the first timer (FIG. 3) is started (step S14). If there is a command being executed, the second timer (FIG. 3) is started (step S15). Thereafter, a command command corresponding to the input command is issued to each of the HDD command queues 5a to 5n (step S16), and the process returns to waiting for input in step S11.

  If it is determined in step S12 that no command has been input, it is determined whether or not the count time of the first timer has exceeded a predetermined time (step S17). It is determined whether or not the process (hereinafter referred to as the first command) has been completed (step S18). Here, when the process is completed, the process returns to the input waiting in step S11. If not finished, the processing time of the first command is timed out, the processing is interrupted at that time, and time-out processing is performed to reproduce or record using only the data processed at that time (step S19). ), And returns to the input waiting in step S11.

  If it is determined in step S17 that the count time of the first timer has not exceeded the determined time, it is determined whether the count time of the second timer has exceeded the determined time (step S20), if exceeded, it is determined whether or not the processing of the second input command (hereinafter referred to as the second command) has been completed (step S21). Here, when the process is completed, the process returns to the input waiting in step S11. If not finished, the processing time of the second command is timed out, the processing is interrupted at that time, and time-out processing is performed to reproduce or record using only the data processed at that time (step S22). ), And returns to the input waiting in step S11.

  If it is determined in step S20 that the count time of the second timer has not exceeded the determined time, it is determined whether or not the HDD command queue has been completed for all HDDs (step S23). If not, the process returns to waiting for input in step S11. If finished, it is determined whether or not the processing for the first command is finished for all HDDs (step S24), and if finished, the first command is finished (step S25). Return to waiting for input. If not finished, it is determined whether or not the processing for the second command is finished for all HDDs (step S26). If finished, the second command is finished (step S27). In this case, the process returns to the input waiting in step S11.

  Next, an operation example in each of the HDD command queues 5a to 5n prepared for each HDD prepared in the HDD controller 5 will be described with reference to the flowchart of FIG. While waiting for input of a command from the CPU 6 side (step S31), it is determined whether or not a command has been input (step S32). If a command is input, whether or not there is a command currently being executed in the corresponding HDD. Judgment is made (step S33). If it is determined that there is no command currently being executed, the command is sent to the HDD as it is, the read or write process corresponding to the command is executed (step S34), and the process waits for input in step S31 (step S36). . If there is a command being executed in step S33, the command is stored in its own HDD command queue (step S35), and the process waits for input in step S31 (step S36).

  If it is determined in step S32 that there is no command input, it is determined whether or not the current HDD operation is completed (step S37). If not finished here, the process returns to the input waiting in step S11. If it is determined in step S37 that the operation of the HDD has ended, it is determined whether or not a command is currently stored in the HDD command queue (step S38). If the command is stored in the HDD command queue, the stored command is issued to the corresponding HDD (step S39), and the process returns to the input waiting state in step S31. If no command is stored in the HDD command queue, the process directly returns to the input wait in step S31.

  By performing the command processing in this way, when a plurality of commands are input to the storage unit 1 continuously at the same time, the processing is performed as shown in FIG. 6, for example. That is, it is assumed that the first command and the second command are issued simultaneously and input to the storage unit 1. Here, the first command is, for example, a command for instructing writing of certain data, and the second command is, for example, a command for instructing reading of certain data. At this time, the CPU 6 in the storage unit 1 sends a first command to each of the HDDs 10 (1) to 10 (14) via the HDD controller 5, and the respective HDDs 10 (1) to 10 (14) Processing (writing processing) Ca1 to Ca14 for one command is executed. Subsequently, the CPU 6 also sends a second command to the HDD controller 5. At this time, the second command is held in the HDD command queue corresponding to each HDD, and waits until the processing of the first command ends for each HDD. When the processing of the first command is completed, the second command held in the HDD command queue is sent to the HDD at each individual timing, and following the processing Ca1 to Ca14 for the first command, Processes Cb1 to Cb14 are executed.

  Therefore, as can be seen from comparison with FIG. 7 described as the conventional example, there is no waiting time from the end of the processing of the first command to the start of the processing of the second command. As long as the HDD and the HDD that takes time to process the second command are not the same, the time until the processing of the two commands is completed can be shortened. For this reason, particularly in the case of an AV server that records AV data as in this example, it is necessary to complete the recording and playback processing within the actual time for recording and playback. Even in such a case, there is a high possibility that the processing will be completed in real time, and the reliability as an AV server is improved.

  In addition, the processing of each command is performed in order at successive timings as described above, but the processing of the timeout time that defines the maximum processing time of each command is determined individually for each command. As a result, proper timeout processing is possible.

  In the example of FIG. 6, the example in which writing (recording) to the HDD and reading (playback) from the HDD are performed simultaneously as processing by two commands input at the same time is described. The present invention can also be applied to the case where data is written at the same time or the case where two types of data are read simultaneously.

  In the configuration shown in FIG. 3, the HDD controller 5 is provided with HDD command queues 5a to 5n as instruction holding means corresponding to each HDD, and commands are temporarily stored in the HDD command queue. However, the operation state of each HDD is monitored on the other control unit side such as the CPU 6 and a command (instruction) is sent at a timing set individually for each HDD according to the monitoring state. May be.

  In the above description, the processing when two commands are generated at the same time has been described. However, when three or more commands are issued at the same time, each individual HDD is similarly processed. Processing time may be shortened by setting sequential processing timings. In this case, for example, if the number of commands that can be stored in each command queue is increased correspondingly, it can be dealt with.

  Further, in the example of the embodiment described above, the present invention is applied to a storage unit in which a total of 15 HDDs are mounted. However, a storage unit in which a different number of HDDs are mounted, AV data, error correction, spare The present invention may also be applied to storage units in which the number of HDDs for use is other than 10, 4, and 1 respectively.

  In the above example, the present invention is applied to the storage for the AV server. However, the present invention can also be applied to an apparatus using an HDD for storing and reading other data as a storage unit.

It is a block diagram which shows the structure of the storage part of AV server to which one embodiment of this invention is applied. 1 is a system diagram showing a system configuration of an AV server according to an embodiment of the present invention. It is explanatory drawing which shows the structural example seen from the command transmission state by one embodiment of this invention. It is a flowchart which shows the operation example of the command queue by one embodiment of this invention. It is a flowchart which shows the operation example of the command queue of each HDD by one embodiment of this invention. It is a time chart which shows the example of a process of the several command by one embodiment of this invention. It is a time chart which shows the processing example of the conventional multiple command.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 1 Storage part, 2 FC controller, 3 Recording / reproducing processing system, 4 Cache memory, 5 HDD controller, 5a-5n HDD command queue, 6 CPU, 6a Command queue, 7 DRAM, 8 Non-volatile SRAM, 9 Bus, 10 (1 ) -10 (15) HDD, 11 input / output processor section, 15 maintenance terminal

Claims (4)

In a data storage device that stores data using a plurality of hard disk drives,
A command receiving unit for receiving a command for instructing writing and / or reading of data to the plurality of hard disk drives;
When at least the first command and the second command are continuously received by the command receiving unit, the instruction corresponding to the first command is completed in each of the hard disk drives constituting the plurality of hard disk drives Then, a data storage device comprising: a control unit that issues an instruction corresponding to the second command at an individual timing set for each hard disk drive. The data storage device according to claim 1,
For each of the plurality of hard disk drives, a holding unit for storing instructions corresponding to the commands is provided.
In the hard disk drive corresponding to each holding unit, an instruction input when another instruction is being executed is stored in the holding unit, and when the executing instruction is completed, the instruction stored in the holding unit is stored in the hard disk drive Data storage device characterized by being transferred to and executed by The data storage device according to claim 1,
A data storage device, wherein a timeout time is set for each of the commands, and the corresponding timeout processing is performed when the processing for each command cannot be performed within the specified timeout time. In a control method of a data storage device that controls writing or reading to a data storage device that stores data using a plurality of hard disk drives,
When at least a first command and a second command are successively received as commands for instructing writing and / or reading of data to and from a plurality of hard disk drives, a first command is sent to the plurality of hard disk drives. The instruction corresponding to the command of and the instruction corresponding to the second command are performed,
When the instruction corresponding to the first command is completed in each hard disk drive, the instruction corresponding to the second command is performed at an individual timing set for each hard disk drive. Control method of the device.

Images