JP2006059502A - Method and device for inputting/outputting data - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a method for inputting/outputting data, suited to an active type database using a compact disk drive, and a device for inputting/outputting data, using the method. <P>SOLUTION: Retrieval information notified from a mobile device 1 is sent to a control unit 4 through an interface 3 in a compact disk drive 2. Before disk access is performed, for an index stored in an index storage part 51, an index optimization part 41 executes index optimization, i.e., sorting of the physical position of a retrieval target record. A head control unit 45 controls a data reading part 62 to sequentially access the physical positions of records indicated by optimized indexes, and to read the record stored in a data storage part 63. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a data input / output method suitable for a small disk drive, and more particularly to a data input / output method suitable for an active database using a small disk drive, and to a data input / output device using the data input / output method. is there.

  In recent years, with the advent of small notebook personal computers, PDAs (Personal Digital Assistants), mobile phones and the like (hereinafter referred to as mobile devices), it has become possible to carry mobile devices on a daily basis. Mobile devices are becoming smaller and more advanced, and are expected to be used as wearable computers with built-in GPS (Global Positioning System) and cameras in the future. At the same time, since the demand for mobile devices to operate as user assistants is increasing, it is necessary for the mobile device to actively present information by judging the information requested by the user according to the situation. Is desired.

  A scene where a mobile device assists a user is widely demanded from a place of daily life to a disaster rescue site. In an environment where a network cannot be used, the mobile device is required to operate alone. In order to operate by a single device, it is necessary to have a large-capacity storage capable of storing all necessary data, and it is appropriate to use a low-cost small disk drive.

In order to actively present information required by the user, it is effective to use an active database that operates based on rules registered in advance and allows customization of information to be presented (for example, non-patent literature). 1).
Sharma Chakravarthy, Barbara Blaustein, Alejandoro P. Buchmann, Michael Carey, Umeshwar Dayal, David Goldhirsch, Meichun Hsu, Rajiv Jauhari, Rivka Ladin, Miron Livny, and Dennis McCarthy: "HiPAC: A RESEARCH PROJECT IN ACTIVE TIME-CONSTRAGEMENTD DATABASE" , Xerox Advanced Information Technology, XAIT-89-02 (1989.7).

  However, compared to a normal database, event processing, condition evaluation and actions are performed automatically by the system in an active database, so disk access occurs frequently, such as data browsing and event monitoring, and processing costs are reduced. Becomes higher.

  In general, a small disk drive has a low disk access speed in the current configuration because the disk rotation speed and head seek cannot be increased due to problems such as power consumption and shock resistance. Therefore, the access waiting time (average seek time + average rotation waiting time) for the head to move to the target record position is several tens of ms, which is a long time when viewed from the system side. On the other hand, the memory density has been increasing year by year, and the sequential access speed has been improved even at the same rotational speed. In other words, in order to efficiently use a small disk drive, how to reduce the head seek time is an important key.

  Therefore, in order to realize an active database using a small disk drive, a file system (data input / output method) with a short head seek time suitable for the characteristics of the small disk drive is required.

  The present invention has been made in view of the above problems, and even in a small disk drive with a low disk rotation speed, it is possible to shorten the access waiting time for the head to move to the target record position, and to It is an object to provide a data input / output method suitable for a type database and a data input / output device using the same.

  The invention according to claim 1 comprises a data storage means for storing data, a head position monitoring means for monitoring the position of a head for reading data from the data storage means, and the data storage means, comprising a magnetic disk. Obtained from the head position monitoring means when reading the target data, which is the data to be read from the data storage means, and a semiconductor storage device having an index storage means for storing an index for designating the address of the data stored in Index optimization means for optimizing an index for designating an address of the target data stored in the index storage means based on the position of the head to be stored, and storing the index storage means in the index storage means; and storing the index storage means in the index storage means The target data is controlled based on the index. Characterized in that it comprises a data reading means for reading, a.

  According to this configuration, the data is stored in the magnetic disk as the data storage means, and the index specifying the address where the data is stored is stored in the index storage means provided in the semiconductor storage device. When reading target data, which is data to be read from the magnetic disk, the index optimizing means corresponds to the target data stored in the index storage means based on the head position obtained from the head position monitoring means. The index is optimized and stored in the index storage means. Then, the data reading unit drives the head based on the index stored in the index storage unit, and reads the target data.

  A second aspect of the present invention is the data input / output device according to the first aspect, wherein the magnetic disk is a small disk drive having a diameter of 1 inch or less.

  According to this configuration, a small disk drive including a magnetic disk having a diameter of 1 inch or less and a semiconductor storage device is used as a data input / output device. First, data is stored in the magnetic disk, and an index for designating an address storing the data is stored in the semiconductor memory device. When reading the target data from the magnetic disk, the index corresponding to the target data in the semiconductor memory device is optimized based on the head position obtained from the head position monitoring means and stored in the semiconductor memory device. Then, the head is driven based on the index stored in the semiconductor memory device, and the target data is read.

  According to a third aspect of the present invention, in the data input / output device according to the first or second aspect, the index optimization unit is configured to access the head at the time of data reading based on the position of the head obtained from the head position monitoring unit. The index is optimized so as to minimize the waiting time.

  According to this configuration, the index optimization unit receives the current position of the head from the head position monitoring unit. Then, it is calculated how the movement time (access waiting time) to the predetermined data becomes the shortest when moving from the head position, and the indexes are rearranged as such.

  According to a fourth aspect of the present invention, in the data input / output device according to any one of the first to third aspects, the data input / output device further includes data writing means for writing data into the data storage means.

  According to this configuration, the data writing unit writes data into the data storage unit.

  According to a fifth aspect of the present invention, in the data input / output device according to the fourth aspect, the data writing unit writes all logs into the data storage unit.

  According to this configuration, all logs are written to the data storage means.

  According to a sixth aspect of the present invention, in the data input / output device according to the fourth or fifth aspect, the data writing means writes data to an address next to the end address of the data stored in the data storage means. It is characterized by.

  According to this configuration, data is sequentially added to the address next to the end address of the data already stored in the data storage means.

  A seventh aspect of the present invention is the data input / output device according to any one of the first to sixth aspects, further comprising an index re-storing unit that causes the data writing unit to periodically store an index in the data storage unit. It is characterized by that.

  According to this configuration, the index is periodically stored in the data storage unit by the index re-storing unit.

  According to an eighth aspect of the present invention, in the data input / output device according to any one of the fourth to seventh aspects, a record ID control for issuing a record ID for specifying input data received together with a command based on the command The apparatus further comprises means.

  According to this configuration, when a command and input data are received, the record ID control means issues a record ID for specifying the input data in accordance with the input command.

  According to a ninth aspect of the present invention, in the data input / output device according to the eighth aspect, based on the input data and the first command, a new record ID for identifying the input data is received from the record ID control means, First processing means for writing the record ID and the input data as continuous data to the data storage means by the data writing means, and creating an index for designating the address of the continuous data, and storing the index in the index storage means And an index assigning means.

  According to this configuration, first, the first processing unit receives a new record ID for specifying input data from the record ID control unit based on the input data and the first command. Then, the record ID and the input data are written as continuous data in the data storage means by the data writing means. Subsequently, the index assigning means creates an index that designates the address of the continuous data, and stores the index in the index storage means.

  The invention according to claim 10 is the data input / output device according to claim 8 or 9, wherein the record ID specifying the input data is received from the record ID control means based on the input data and the second command, A second processing means for writing the record ID and the input data as continuous data in the data storage means by the data writing means; and an index storage means for data having the same record ID as the record ID for specifying the input data. Index changing means for deleting the index, creating an index designating the address of the continuous data, and storing the index in the index storage means is further provided.

  According to this configuration, first, the second processing means receives the record ID for specifying the input data from the record ID control means based on the input data and the second command. Then, the record ID and the input data are written as continuous data in the data storage means by the data writing means. Subsequently, the index changing unit deletes the index of the data having the same record ID as the record ID for specifying the input data in the index storage unit, creates an index for designating the address of the continuous data, and stores the index as the index storage unit. To remember.

  The invention according to claim 11 is the data input / output device according to claim 8 or 9, wherein the record ID for specifying the input data is received from the record ID control means based on the input data and the third command, The data stored in the data storage means and having the same record ID as the input data is compared with the input data to obtain an updated data portion in the input data, and the record ID and the updated data are obtained. A third processing means for writing a portion as continuous data to the data storage means by the data writing means; and an index storage means for adding the continuous data to an index of data having the same record ID as the record ID for specifying the input data. Address as a list and add the index And further comprising a index additional means for storing the unit.

  According to this configuration, first, the third processing means receives the record ID for specifying the input data from the record ID control means based on the input data and the third command. When the input data having the same record ID as the data already stored in the data storage means is stored in the data storage means, the data portion and record updated from the already stored data among the input data The ID is written as continuous data in the data storage means by the data writing means. Subsequently, in the index storage means, the addresses of continuous data are added as a list to the index of the data having the same record ID as the record ID specifying the input data, and stored in the index storage means.

  According to a twelfth aspect of the present invention, in the data input / output device according to any of the eighth to eleventh aspects, the deletion target is based on deletion information including at least information specifying data to be deleted and a fourth command. A fourth processing means for receiving a record ID for identifying the data from the record ID control means and writing the record ID and a mark indicating that the data is to be deleted as continuous data in the data storage means by the data writing means; And the index storage means further comprises index erasure means for erasing the index of data having the same record ID as the record ID for specifying the input data.

  According to this configuration, first, the fourth processing means performs record ID control on the record ID for specifying the data to be deleted based on the deletion information including at least information specifying the data to be deleted and the fourth command. Receive from means. Then, the record ID and the mark indicating the data to be deleted are written as continuous data in the data storage means by the data writing means. Subsequently, the index erasure unit erases the index of the data having the same record ID as the record ID for specifying the input data in the index storage unit.

  The invention according to claim 13 is the data input / output device according to any one of claims 8 to 12, wherein the search object is based on search information including at least information specifying data to be searched and a fifth command. The data ID is received from the record ID control means, the index of the data having the record ID is specified in the index storage means, and the data reading means is based on the index stored in the index storage means The method further comprises fifth processing means for reading the search target data from the data storage means.

  According to this configuration, first, the fifth processing means performs record ID control on the record ID for specifying the search target data based on the search information including at least the information specifying the search target data and the fifth command. Receive from means. Subsequently, the index of the data having the record ID is specified in the index storage means. The data reading means reads the data to be searched from the data storage means based on the index stored in the index storage means.

  According to a fourteenth aspect of the present invention, in the data input / output device according to any one of the fourth to thirteenth aspects, the semiconductor memory device further includes a buffer unit, and the data input / output device inputs input data to the buffer unit. The buffer writing means for storing, the buffer monitoring means for monitoring the storage area of the buffer means, and when the buffer monitoring means detects that the storage area of the buffer means is filled, the buffer means is stored in the buffer means. And a first flash control means for causing the data writing means to write the data stored in the data storage means.

  According to this configuration, the input data is not directly written in the data storage means but is temporarily stored in the buffer means by the buffer writing means. When the buffer monitoring means detects that the storage area of the buffer means is full, the first flash control means stores all input data stored in the buffer means in the data storage means. .

  According to a fifteenth aspect of the present invention, in the data input / output device according to any one of the fourth to thirteenth aspects, the semiconductor memory device further includes a buffer unit, and the data input / output device inputs input data to the buffer unit. As soon as a predetermined attribute is detected in the input data by the buffer attribute monitoring means for monitoring the attribute of the data stored in the buffer means, and the buffer attribute monitoring means, the buffer And a second flash control means for causing the data storage means to write the data stored in the data storage means to the data storage means.

  According to this configuration, the input data is not directly written in the data storage means but is temporarily stored in the buffer means by the buffer writing means. As soon as input data having a predetermined attribute is detected by the buffer attribute monitoring means, for example, data having a high importance level that is problematic if it is lost due to a failure, the second flash control means performs buffering. All input data stored in the means is stored in the data storage means.

  According to a sixteenth aspect of the present invention, in the data input / output device according to any one of the first to fifteenth aspects, the magnetic disk as the data storage means is provided while the data input / output device is powered on. It is always rotating.

  According to this configuration, the magnetic disk always rotates while the data input / output device is powered on regardless of whether the disk access is performed.

  According to a seventeenth aspect of the present invention, in the data input / output device according to any one of the first to sixteenth aspects, the data input / output device is detachable from a mobile device.

  According to this configuration, a mobile device including the data input / output device according to the present invention can be realized.

  According to a twenty-eighth aspect of the present invention, in the data input / output method, a head position monitoring step for monitoring a position of a head for reading data from a magnetic disk, and an index for designating an address of data stored in the magnetic disk are provided as semiconductors. An index storage step for storing in the storage device, and when reading target data, which is data to be read from the magnetic disk, based on the position of the head obtained from the head position monitoring step, the index storage step causes the semiconductor to Optimizing an index that specifies an address of the target data stored in a storage device and storing the index in the semiconductor storage device; and controlling the head based on the index stored in the semiconductor storage device; Data reader for reading the target data Characterized in that it comprises a and.

  According to this configuration, the index specifying the address of the data stored in the magnetic disk is stored in the semiconductor memory device by the index storage process. When reading the target data from the magnetic disk, the index corresponding to the target data in the semiconductor memory device is optimized by the index optimization step based on the head position obtained from the head position monitoring step, and stored in the semiconductor memory device. Then, in the data reading step, the head is driven based on the index stored in the semiconductor memory device to read the target data.

  According to the first aspect of the present invention, the data is stored in the magnetic disk, and the index is stored in the semiconductor memory device separately from the data. Since the semiconductor memory device has a higher access speed than the magnetic disk, the optimization processing of the index corresponding to the target data can be performed at high speed when the target data is read from the magnetic disk. As a result, the access waiting time for the head to move to the target record position can be shortened, and the reading speed of the target data is improved.

  According to the second aspect of the present invention, the data is stored in the magnetic disk having a diameter of 1 inch or less, and the index is stored in the semiconductor memory device separately from the data. Since the semiconductor memory device has a higher access speed than the magnetic disk, the optimization processing of the index corresponding to the target data can be performed at high speed when the target data is read from the magnetic disk. As a result, even in a small disk drive with a low disk rotation speed, it is possible to reduce the access waiting time for the head to move to the target record position, and the target data reading speed is improved.

  According to the third aspect of the invention, since the index is optimized so as to minimize the access waiting time of the head, sequential access is possible and data reading can be performed at high speed.

  According to the fourth aspect of the present invention, since data writing means is further provided, writing in an arbitrary format is possible.

  According to the fifth aspect of the present invention, since all logs are stored, all data can be recovered to a consistent state even if the system crashes. Furthermore, a temporal query can be realized by arranging data in chronological order.

  According to the sixth aspect of the present invention, the area in which data is additionally written is determined in the data storage means, and there is no need to search for a storable free area. Therefore, it is possible to add data efficiently and at high speed.

  According to the seventh aspect of the present invention, even if an index disappears from the index storage means due to a system failure, the index can be read from the data storage means. The latest index can be restored by examining these time stamps.

  According to the eighth aspect of the invention, the record ID control means issues a record ID for specifying the input data in accordance with the input command, and therefore, between the input data and the other stored data. The relationship becomes clear.

  According to the ninth aspect of the invention, the processing is simple and the index does not increase beyond the number of data records, so that memory can be saved. Also, since all input data is stored in the data storage means, all data can be recovered to a consistent state even if the system crashes. Furthermore, a temporal query can be realized by arranging data in chronological order.

  According to the invention described in claim 10, since the index does not increase more than the number of data records, memory can be saved. Further, since the data is rewritten by changing the index instead of actually updating the data, the process is simple. Furthermore, since all input data is stored in the data storage means, all data can be recovered to a consistent state even if the system crashes. Furthermore, a temporal query can be realized by arranging data in chronological order.

  According to the eleventh aspect of the invention, since only the changed portion of the data is stored in the data storage means, it is possible to save the disk area. Also, since all input data is stored in the data storage means, all data can be recovered to a consistent state even if the system crashes. Furthermore, a temporal query can be realized by arranging data in chronological order.

  According to the twelfth aspect of the present invention, it is possible to make the system seem to have deleted data by deleting only the index without deleting the data from the data storage means. In fact, since all input data is stored in the data storage means, all data can be recovered to a consistent state even if the system crashes. Furthermore, a temporal query can be realized by arranging data in chronological order.

  According to the thirteenth aspect of the invention, the speed of searching for the target data is improved. In the case of the combination with claims 9, 10, 12, and 13, or the combinations with claims 9, 11, 12, and 13, the data input / output processing can be closed on the data input / output device. A data input / output device having a high-level interface that does not depend on the connected mobile device can be provided.

  According to the fourteenth aspect of the present invention, it is not necessary to frequently access the data storage means, and the access time can be kept short.

  According to the fifteenth aspect of the present invention, it is not necessary to frequently access the data storage means, the access time can be kept short, and important data can be prevented from being lost due to a failure.

  According to the sixteenth aspect of the present invention, it is not necessary to wait until the disk is started from the disk stop state, and frequently arriving data can be stored. Furthermore, since the power necessary for starting and stopping the disk is not required, the power consumption can be kept low.

  According to the invention described in claim 17, a mobile device provided with the data input / output device according to the present invention can be realized.

  According to the invention described in claim 18, the data is stored in the magnetic disk, and the index is stored in the semiconductor memory device separately from the data. Since the semiconductor memory device has a higher access speed than the magnetic disk, the optimization processing of the index corresponding to the target data can be performed at high speed when the target data is read from the magnetic disk. As a result, the access waiting time for the head to move to the target record position can be shortened, and the reading speed of the target data is improved.

  Hereinafter, embodiments according to the present invention will be described with reference to the drawings.

[Outline of this embodiment]
In the present embodiment, in order to realize a high-speed database for wearable computers, a log file system is used, and a file system that sequentially adds all changed data as a log is adopted. In other words, the log is change data stored when a change occurs in the database by a database operation.

  In the present embodiment, since the data update is performed by sequential access, the speed of the disk I / O can be increased. However, since the data is frequently updated, the target data set does not always exist continuously, and the data search is not always performed sequentially. In the present embodiment, this problem is solved by using a disk control processor and (buffer) memory in the small disk drive to process optimization of disk access at the time of data retrieval in the small disk drive.

  In the present embodiment, the disk control processor in the small disk drive is also used as a processor for processing a database management system (DBMS). Then, an index indicating a physical storage position (address) of data stored (stored) in the disk (magnetic disk) is developed on the memory (semiconductor storage device). In general, an index is used to perform high-speed processing when retrieving data on a disk. However, since indexes are frequently updated, creating an index on a disk requires disk access when reading, updating, or changing the index, resulting in random access and a reduction in processing speed.

  In the present embodiment, this problem is dealt with by expanding the index into the memory on the small disk drive, and the sequential access can be fully utilized.

  Assume that a small disk drive is actually installed in a wearable device. FIG. 15 is an external configuration diagram showing an operation surface side of a mobile phone connected to the data input / output device according to the present invention. FIG. 15 shows an example in which a small disk drive 95 (data input / output device) such as a microdrive having a magnetic disk diameter of 1 inch or less is mounted on a disk mounting portion 94 of a mobile phone as a wearable device.

  The mobile phone shown in FIG. 15 includes a mobile phone main body 90, a display unit 91 for notifying the user of call information and the like, a key input unit 92 for the user to input telephone numbers and database operation instructions, and for transmission and reception. Antenna 93 and various processing circuits (not shown) in the mobile phone main body 90. In addition, the mobile phone is provided with a mounting portion 94 for mounting the small disk drive 95 on the side surface of the mobile phone main body 90. For this reason, the user can mount the small disk drive 95 in the mounting portion 94 to enable a mode of use in which data is exchanged between the mobile phone main body 90 and the small disk drive 95.

  For example, consider a case where a user uses a schedule function. The schedule of a certain day of the user is stored in the small disk drive 95 by operating the key input unit 92 of the mobile phone main body 90 in advance. When the user inserts (adds) a new schedule, the new schedule is stored in the small disk drive 95 by operating the key input unit 92. Similarly, when the stored schedule is updated (changed), the existing schedule is deleted, or the schedule is inquired (searched), the small disk drive 95 is operated by operating the key input unit 92 in the same manner. It is possible to update, delete, or retrieve records stored in.

  Here, a record is one piece of data formed by collecting attributes (fields). For example, in the case of an address book, attributes such as name, address, and telephone number are collected to form one record.

  Thus, a person using a mobile phone as a wearable computer can use the database function of the small disk drive 95 by selecting information and operating the key input unit 92. In addition, this is not only a case of using a schedule function provided in a mobile phone. For example, in the case of a mobile phone with a camera, if it is data acquired by the mobile phone, such as a captured still image or a moving image, it is small. It can be stored in the disk drive 95.

  In addition, the user of the mobile phone not only selectively selects information, but the mobile phone itself actively (automatically) outputs data to the small disk drive 95, and conversely, the small disk For example, data may be input from the drive 95. For example, in the case of a mobile phone having a GPS function, the position of the owner (the position of the mobile phone) is automatically changed from the mobile phone main body 90 to the small disk drive 95 without requiring the owner's operation. It is possible to make it output to. Thereby, it becomes possible to reproduce the action of the day in chronological order.

  In the future, it is assumed that a configuration including a wireless IC tag (RFID tag) on an object will be generalized. For example, IT home appliances such as an intelligent refrigerator that automatically identifies foods when they are bought and put in a refrigerator and makes a list of the foods held and informs the expiration date are conceived. If a mobile device equipped with a data input / output device according to the present invention is used as a wearable computer, it is possible to sequentially store data sent from an object equipped with the wireless IC tag and configure it as a database. At this time, as will be described later, since the DBMS can be closed on the small disk drive, the DBMS independent of the mobile device is realized.

  The data input / output device according to the present invention can be applied to other mobile devices, such as mobile computers such as notebook personal computers and PDAs, without being limited to mobile phones. In addition, not only data obtained by using functions provided in mobile phones, but also data received from other digital devices by wire or wirelessly can be used mainly by users or by mobile devices. 95 can be stored.

  Here, as an example, let us consider storing GPS information in a small disk drive, and estimate the required memory capacity. Currently, the accuracy that can be measured by GPS is about 10 cm in the horizontal direction. Assuming that the user uses the mobile device on the go, the accuracy of 10 cm is not necessary, and it is considered that there is no problem at about 10 m. Therefore, data is added once every 10 seconds during walking.

  Examples of the data table used when going out include a position information table, a store information table, and a profile table that stores user actions. Assuming that data is added to the three tables every 10 seconds, about 26,000 records are stored in 24 hours. The index size for this record is 105 KB, which can be stored in a memory in a small disk drive. Moreover, since the initial record taken out from the home is considered to be personal data of the user, it is unlikely that the total number of records will increase significantly.

  Furthermore, there is no need to continue to use data for several days in a data table in a mobile device equipped with a data input / output device according to the present invention. It can also be used for output to a computer. An example of such usage is a configuration in which data of the day is automatically output to a PC when the mobile device is charged. For example, if the configuration is such that after data is output to the PC, data other than the data specified by the user is deleted, an increase in the required memory amount can be suppressed, and all the indexes can be built on the memory. It becomes possible.

  FIG. 1 is a functional block diagram of an embodiment of a data input / output device according to the present invention. The mobile device 1 is connected to a small disk drive 2 and exchanges various data between them. In the present embodiment, the small disk drive 2 corresponds to a data input / output device.

  The mobile device 1 is a PDA, a mobile phone, or a small computer such as a small notebook computer.

  The small disk drive 2 is, for example, a micro drive manufactured by Hitachi GST (HGST), an interface 3 for communication with the mobile device 1, a control unit 4 that performs various controls and calculation processing in the small disk drive 2, A memory 5 composed of a semiconductor memory, a disk unit 6 for storing data, and a code storage unit 7 for storing a code (program) for executing a DBMS function are provided.

  The interface 3 is used for communication with the mobile device 1. However, according to the embodiment of the present invention as described later, it is possible to have a unified function independent of the mobile device 1. It is.

  The control unit 4 includes, for example, a CPU (Central Processing Unit), and is a functional unit having functions of disk control and a database management system (DBMS). The control unit 4 includes an index optimization unit 41 (index optimization unit), a head position monitoring unit 42 (head position monitoring unit), a record ID control unit 43 (record ID control unit), a storage data generation unit 44, and a head control unit. 45, end address storage unit 46, database processing unit 47, index re-storage unit 48 (index re-storage unit), index deletion unit 49, index addition unit (not shown), index assignment unit 81 (index provision unit) and difference data The calculation unit 82 is provided.

  The head position monitoring unit 42 is at which address in the data storage unit 63 the write head that writes data to the data storage unit 63 (data storage unit) and the read head that reads data from the data storage unit 63 are currently located. Is a functional unit for monitoring

  When reading data from the data storage unit 63, the index optimization unit 41 optimizes the index corresponding to the data in the index storage unit 51 (index storage unit) based on the head position obtained from the head position monitoring unit 42, This is a functional unit stored in the index storage unit 51.

  Here, the optimization of the index means, for example, rearranging the pointer included in the index, that is, the physical position (address) of the record indicated by the index so that predetermined processing can be optimally performed.

  The record ID control unit 43 is a functional unit that determines processing requested from the database based on the received command and issues a record ID (RID) accordingly. For example, if the requested process is a record insertion, a new RID is issued. If the requested process is a record update, delete, and search, the corresponding RID is acquired from the index storage unit 51. It is a functional part. The RID is a code that identifies data.

  The storage data generation unit 44 is a functional unit that arranges the RID, the new record, and the record length thereof as one continuous record and arranges them into a format for storage in the data storage unit 63. In addition, the storage data generation unit 44 also performs update processing such as change / correction of records when updating records.

  The head control unit 45 is stored in the head position information notified from the head position monitoring unit 42, the end address information of the record stored in the storage unit 63 notified from the end address storage unit 46, or the index storage unit 51. This is a functional unit that controls the data writing unit 61 (data writing unit) or the data reading unit 62 (data reading unit) based on the index information to access the data storage unit 63.

  The end address storage unit 46 is a functional unit that stores the end address of the record stored in the data storage unit 63.

  The database processing unit 47 determines the processing content requested of the small disk drive 2 via the input unit of the mobile device 1 (for example, the key input unit 92 in FIG. This is a functional unit that notifies a predetermined functional unit. For example, if SQL (Structured Query Language) is used as a database processing language, record insertion, update, deletion, or search is controlled by commands INSERT, UPDATE, DELETE, and SELECT, respectively. A functional unit 47 identifies these commands.

  Further, the database processing unit 47 receives a new record ID from the record ID control unit 43 based on the new data and the insert command received from the mobile device, and inserts the record ID and the new data in the data storage unit 63. Process.

  Further, the database processing unit 47 receives, from the record ID control unit 43, the record ID corresponding to the update data among the data stored in the data storage unit 63 based on the update data and the update command received from the mobile device. This is a functional unit that performs an update process for storing the record ID and update data in the data storage unit 63.

  Furthermore, the database processing unit 47 deletes the record to be deleted from the data stored in the data storage unit 63 based on the deletion information and the deletion command including at least information specifying the record to be deleted received from the mobile device. Is a functional unit that performs a deletion process of receiving the record ID corresponding to the item ID from the record ID control unit 43 and storing at least the record ID in the data storage unit 63.

  Further, the database processing unit 47 selects the search target record from the data stored in the data storage unit 63 based on the search information and the search command including at least information specifying the search target record received from the mobile device. Is a functional unit that performs a search process for receiving the record ID corresponding to the item ID from the record ID control unit 42 and reading the data having the record ID from the data storage unit 63.

  In the present embodiment, a relational type will be described as an example from the data model, but the present invention is not limited to this.

  The index re-storage unit 48 is a functional unit that stores the index stored in the index storage unit 51 in the data storage unit 63 every time a predetermined time elapses.

  When the input (update) data having the same record ID as the data already stored in the data storage unit 63 is stored in the data storage unit 63, the index deletion unit 49 stores the data stored in the index storage unit 51. This is a functional unit that deletes the index and stores the index of the input data having the new record ID in the index storage unit 51.

  When the input data having the same record ID as the data already stored in the data storage unit 63 is stored in the data storage unit 63, the index addition unit (not shown) updates the input data from the stored data. This is a functional unit that stores only the data portion that has been stored in the data storage unit 63, and additionally stores the index of input data having a new record ID in the index storage unit 51.

  The index assigning unit 81 is a functional unit that creates an index indicating the head of the address stored in the data storage unit 63 and stores the index in the index storage unit 51 in the memory 5.

  The difference data calculation unit 82 is a functional unit that compares the data stored in the data storage unit 63 with the input data and obtains a data portion (update attribute value) updated in the input data.

  The index erasing unit 49 and the index assigning unit 81 have a function as an index changing unit. The index adding unit (not shown) and the index assigning unit 81 have a function as index adding means. Further, the index erasing unit 49 has a function as index erasing means.

  The memory (semiconductor storage device) 5 includes an index storage unit 51 and a buffer unit 52 (buffer means). The index storage unit 51 is a functional unit that stores an index that specifies an address of data stored in the data storage unit 63. The buffer unit 52 is a functional unit for temporarily storing data other than the index.

  The disk unit 6 includes a data writing unit 61, a data reading unit 62, and a data storage unit (magnetic disk) 63. The data writing unit 61 is a functional unit that writes data to the data storage unit 63 using a writing head. The data writing unit 61 has a function of writing all logs to the data storage unit 63. The data reading unit 62 is a functional unit that reads data from the data storage unit 63 using a reading head. Hereinafter, for the sake of simplicity, both the write head and the read head are referred to as heads.

  The data storage unit 63 is a functional unit that stores data. For example, the data storage unit 63 has a structure in which a number of aluminum or glass disks coated with a magnetic material are stacked at regular intervals. Further, since the data input / output device according to the present embodiment is assumed to be detachable from a mobile device such as a cellular phone, the diameter of the data storage unit 63 that is a magnetic disk may be 1 inch or less. preferable.

  The code storage unit 7 is composed of a flash memory, for example, and is a functional unit that stores a code for executing the DBMS function.

  In the conventional small disk drive 2, the DBMS function is mainly performed by the DBMS 1, and the control unit (CPU) 4 in the small disk drive 2 mainly performs disk control. By using the data input / output method according to the present invention for the small disk drive 2 and providing the control unit 4 with the DBMS function, it is possible to perform a closed process on the small disk drive 2.

  This is realized, for example, by loading the code stored in the code storage unit 7 into the database processing unit 47 of the control unit 4 when the mobile device is turned on. This eliminates the need for the mobile device 1 to have a DBMS function, and as a result, a DBMS that does not depend on the OS (operating system) of the mobile device 1 is realized.

  Therefore, the operation instruction from the mobile device 1 to the small disk drive 2 can be high in abstraction. Therefore, according to the embodiments described later according to the present invention, a high-level interface can be provided, and the high portability of the small disk drive 2 is realized. As a result, for example, a database created on a Windows (registered trademark) machine using the small disk drive 2 can be used on another UNIX (registered trademark) machine without any special processing.

  Further, the code for executing the DBMS may be stored in the data storage unit 63 or the like instead of the code storage unit 7 configured by a flash memory or the like and arranged in the small disk drive 2. .

  Two examples of data storage methods are shown below, but these are different methods for creating indexes. The index uses the record position stored on the disk as a key.

  The first data storage method is a method in which the entire record is duplicated and added at the time of data update, and the address (pointer) indicated by the index is the head address of the added record. This is called a replication method. In this replication method, when input data having the same record ID as the data stored in the data storage unit 63 is written into the data storage unit 63 by the data writing unit 61, the address of the data stored in the index storage unit 51 The index specifying the address of the input data is stored in the index storage unit 51.

  The second data storage method is a method of appending only the attribute information updated at the time of data update. The pointer to the original record is newly created while the pointer to the original record is kept, and the original record Add as a list to the pointer to. A record can be created by a pointer to the original data and a pointer to the storage destination of the update data. This is called a differential method. In this differential method, in the input data having the same record ID as the data stored in the data storage unit 63, only the data portion updated from the stored data is written into the data storage unit 63 by the data writing unit 61. At this time, an index for designating the address of the input data is further added and stored in the index storage unit 51.

  Hereinafter, as an example of a mobile device, a mobile phone including a small disk drive (data input / output device) according to the present invention is considered. The user who carries the mobile phone is called the former A. It is assumed that each time the user adds a new schedule to the predetermined schedule or changes the schedule, the contents are output as data and additional information from the mobile phone to the small disk drive. Further, the cellular phone also has a GPS function, and description will be made assuming that it has a function of adding location information data and the like to the small disk drive together with the above data.

[First Embodiment]
First, the above-described duplication method will be described as the first embodiment. In this method, when a data change occurs, the entire record to be changed is duplicated, the changed data (input data) is reflected, and then added to the disk.

  FIG. 2 is a schematic diagram showing the correspondence between the data storage and the index of the duplication method which is an embodiment of the data input / output device according to the present invention. (A) is the initial state before the record is inserted or updated, (B) is when the record is inserted, (C) is when the record is updated, and (D) is stored in the index and data storage unit 63 when the record is deleted. The correspondence with the recorded records is shown respectively.

  FIG. 3 is a flowchart showing a processing flow of a replication method which is an embodiment of the data input / output device according to the present invention.

[initial state]
First, the initial state of (A) will be described. On this day, suppose that, for example, A creates data A at 8 o'clock and B at 10 o'clock at home, and stores them in the small disk drive 2 respectively. For this reason, the data storage unit 63 of the disk unit 6 stores, as record ID (RID) -001, data A created by the former at 8 o'clock along with time / position information. Similarly, as RID-002, data B created by the former at 10:00 is stored along with time / position information as well as the data. Further, fine information other than data is stored as additional information.

  An index indicating the head of the address where the records of RID-001 and 002 are stored is stored in the index storage unit 51 in the memory 5. This is indicated by arrows from the topmost index in FIG. 2A to RID-001 and from the second index from the top to RID-002. Further, the end address of the data of RID-002 is stored in the end address storage unit 46.

Insert record
First, a case where a record is inserted (added) in the initial state of (A) ((B)) will be described (first processing). Suppose that A moved from his home and was in Namba at 11:00. Then, it is assumed that data C is created at 11:00 in Namba and is inserted into the small disk drive 2 as a new record (input data) (S10). Since the application software of the cellular phone issues an SQL command corresponding to the operation of the former, in this case, the new record is a small disk drive along with a command (first command) corresponding to the insertion processing (first command). 2 is sent to the control unit 4 through the interface 3 in the system 2.

  When the control unit 4 receives the command and the new record, the database processing unit 47 determines the contents of the database processing. In this case, it is determined to receive a command [INSERT] and insert a record (S11 and S12). Since it is a new record, a new RID is issued in the record ID control unit 43 (S131). In this case, since RID-002 is stored in the data storage unit 63, the newly issued RID is 003. The RID, the new record, and the record length thereof are collected as one new continuous record in the storage data generation unit 44 (S132).

  Subsequently, the head position monitoring unit 42 notifies the head control unit 45 of the current head position. Further, the end address storage unit 46 notifies the head control unit 45 of the end address of the record stored in the data storage unit 63. Based on the information notified from the head position monitoring unit 42 and the end address storage unit 46, the head control unit 45 stores the head end addresses of all records stored in the data storage unit 63, in other words, RID-002. Move to the next address after the end address of the record.

  Then, a new continuous record generated by the storage data generation unit 44, that is, a record of RID-003 is sent to the disk unit 6, and is written into the data storage unit 63 by the data writing unit 61 (S133). At this time, since the head is located at the address next to the end address of all stored records, a new continuous record is added to the previous record in the data storage unit 63.

  After storing the new record, the index giving unit 81 creates an index indicating the head of the address stored in the data storage unit 63, and the index is stored in the index storage unit 51 in the memory 5 (S134). Then, the end address of the data of RID-003 is stored in the end address storage unit 46. Thereby, the insertion of the new record is completed (S135).

  In the conventional DBMS, when a new record is inserted, random access occurs in order to search for a storable free space, and storage takes time. According to the present embodiment, random access does not occur because the additional recording location for storing the record is determined in advance. Therefore, it becomes possible to write data efficiently.

[Record Update]
Next, a case where the record is updated ((C)) in the state after the record insertion ((B)) will be described (second processing). The former rewrites the data A (RID-001) at 13:00 and updates the additional information from A-1 to A-2 (S10). The mobile device 1 outputs a record (input data) including the rewritten data A and position information to the small disk drive 2. The record (update record) is sent to the control unit 4 via the interface 3 in the small disk drive 2 together with a command [UPDATE] corresponding to the update process (second command).

  When the control unit 4 receives the command and the update record, the database processing unit 47 determines the content of the database processing. In this case, it is determined to receive a command [UPDATE] and update the record (S11 and S12).

  The record ID control unit 43 receives the fact that the requested process is a record update, specifies data to be updated using the index stored in the index storage unit 51, and determines the RID (in this case, the RID -001) is acquired (S141). The RID, the update record, and the record length thereof are collected as one changed continuous record in the storage data generation unit 44 (S142).

  Subsequently, in the same manner as in the above-described record insertion, the head control unit 45 sets the head to the end address of all records stored in the data storage unit 63, in other words, the end address of the record of RID-003. Move to address.

  Then, the updated continuous record generated in the storage data generation unit 44, that is, RID-001, is sent to the disk unit 6, and is written in the data storage unit 63 by the data writing unit 61 (S143). At this time, since the head is located at the address next to the end address of all the stored records, the updated continuous record is added to the previous record in the data storage unit 63.

  After storing the update record, the index assigning unit 81 creates an index indicating the head of the address stored in the data storage unit 63, and the index is stored in the index storage unit 51 in the memory 5. At this time, the index deletion unit 49 deletes the index for RID-001 (additional information A-1) already stored in the index storage unit 51, and the index for the updated RID-001 (additional information A-2). Only is stored (S144).

  Further, the pre-update record remains stored in the data storage unit 63 without being erased. Further, the end address of the data of RID-001 (additional information A-2) is stored in the end address storage unit 46. This completes the record update process (S145).

[Delete Record]
Next, a case where a record is deleted ((D)) in the state after this record update ((C)) will be described (fourth processing). In order to delete the information of the data B (RID-002) stored in the data storage unit 63, the user performs a predetermined operation on the mobile device 1 (S10). The mobile device 1 outputs deletion information (input data) including information specifying at least a record to be deleted to the small disk drive 2. The deletion information is sent to the control unit 4 via the interface 3 in the small disk drive 2 together with the command [DELETE] corresponding to the deletion process (fourth command).

  When the control unit 4 receives the deletion information, the database processing unit 47 determines the contents of the database processing. In this case, it is determined to delete a record in response to a command [DELETE] (S11 and S12).

  The record ID control unit 43 receives the fact that the requested process is record deletion, and acquires the RID (RID-002 in this case) of the record to be deleted from the index storage unit 51 (S151). The RID and a mark (for example, [DELETE]) indicating data to be deleted are collected as one continuous record in the storage data generation unit 44 (S152).

  Subsequently, in the same manner as the above-described record insertion and record update, the head control unit 45 sets the head to the end address of all records stored in the data storage unit 63, in other words, the end of the record of RID-001. Move to the next address.

  Then, the continuous record generated in the storage data generation unit 44, that is, RID-002 is sent to the disk unit 6, and is written in the data storage unit 63 by the data writing unit 61 (S153). At this time, since the head is located at the address next to the end address of the stored record, the continuous record is added to the previous record in the data storage unit 63.

  After the continuous records are stored, the index deleting unit 49 deletes only the index corresponding to the RID of the record to be deleted (S154). At this time, the record to be deleted, that is, the record of RID-002 is not actually erased but remains stored in the data storage unit 63. Thus, the record deletion process is completed (S155).

  The conventional DBMS actually deletes the record from the data storage unit 63 when executing the deletion, but in this embodiment, the deletion process is also handled as a record addition. That is, in this embodiment, the record is not actually deleted from the disk, but only the deletion information is added to the disk. When searching for records, the index is used to determine the record position, so deleting the index makes it appear to the system that the record has been deleted.

[Record Search]
Finally, a case where a record is searched in the state after this record update ((C)) will be described (fifth process). In order to search the information of the data A (RID-001) stored in the data storage unit 63, the user performs a predetermined operation on the mobile device 1 (S10). The mobile device 1 outputs search information (input data) including at least information specifying a record to be searched to the small disk drive 2. The search information is sent to the control unit 4 via the interface 3 in the small disk drive 2 together with a command [SELECT] (fifth command) corresponding to the search process.

  When the control unit 4 receives the search information, the database processing unit 47 determines the contents of the database processing. In this case, it is determined to receive a command [SELECT] and search for a record (S11).

  Then, before performing disk access, the index optimization unit 41 performs index optimization, that is, rearranges (sorts) the physical positions of the search target records in the index stored in the index storage unit 51 ( S161). At this time, based on the head position obtained from the head position monitoring unit 42, for example, the index optimization unit 41 minimizes the access waiting time (the time required for the head to start disk access from the current position). In the index storage unit 51, the index is optimized. In other words, the optimization of the index in the present embodiment means rearranging the physical position (address) of the record pointed to by the index, that is, the pointer so that the access waiting time is minimized. Then, the optimized index is stored in the index storage unit 51.

  The index optimization by the rearrangement of pointers will be described with reference to FIG. FIG. 4 is an explanatory diagram of the operation of the search method of the replication method that is an embodiment of the data input / output device according to the present invention. 4A shows a state before index optimization, and FIG. 4B shows a state after index optimization.

  A record specified by the RID is stored in the disk area. In this example, only RID-001, 002, 102, 101 and 103 are shown. In the memory area, an index corresponding to each RID is stored.

  Assume that the index corresponding to the RID stored in the disk area is as shown in FIG. That is, for example, the pointer of the first index is address 3000 indicating the start address of RID-101 on the disk, and the pointer of the second index is address 2800 indicating the start address of RID-002. As described above, when the record is updated or inserted several times, the order of the pointers of the indexes stored in the memory area and the order of the records stored in the disk area are complicatedly entangled.

  Therefore, if reading is performed in the state shown in FIG. 4A, the index pointers are irregularly arranged as 3000, 2800, 3300... Access occurs and reading speed decreases. In the present invention, the index pointers are rearranged to allow sequential access.

  Now assume that the head is located at an address closer to the head than address 1800. Therefore, the index optimizing unit 41 rearranges the index pointers in FIG. 4A from the smallest address on the disk to the larger one in the index storage unit 51, as shown in FIG. 4B. At this time, the correspondence between the address and RID described in the index remains unchanged. That is, the RID of the end point with respect to the start point of the arrow in FIGS. 4 (a) and 4 (b) is the same. When reading is performed in the state shown in FIG. 4B, random access does not occur and sequential access is performed. As a result, data can be read from the disk at high speed.

  In the present embodiment, the index pointer is described as being rearranged from the smallest address on the disk to the larger address, but the present invention is not limited to this, and the head position at the time of starting reading is used as a reference. For example, it is possible to optimize the index so that the access waiting time is the shortest.

  Also, for example, when writing data with high importance and searching for data with low importance are processed at the same time, the write processing should be prioritized and the index optimized to perform the search processing after that. Is also possible. Furthermore, for example, while processing is based on the order of importance, the importance order and the shortest access order are used so that the less important data in the middle of moving the head is read and the access time is shortened. It is also possible to combine them.

  After the index optimization, the head position monitoring unit 42 notifies the head control unit 45 of the current head position. Further, the optimized index is output from the index storage unit 51 to the head control unit 45. The head control unit 45 controls the data reading unit 62 based on the information notified from the head position monitoring unit 42 and the index storage unit 51 to access the physical position of the record indicated by the optimized index. Then, the record stored in the data storage unit 63 is read (S162). This reading is continued until the data reading is completed (S163), and when the data reading is completed, the record search process is completed (S164).

  In addition, this replication method is simple, and the index does not increase beyond the number of records, so that memory can be saved.

[Second Embodiment]
Subsequently, the above-described difference method will be described as a second embodiment. This method is a method in which, when data change occurs, only the difference between records to be changed is added.

  FIG. 5 is a schematic diagram showing the correspondence between the difference type data storage and the index which is an embodiment of the data input / output apparatus according to the present invention. (A) is the initial state before the record is inserted or updated, (B) is when the record is inserted, (C) is when the record is updated, and (D) is stored in the index and data storage unit 63 when the record is deleted. The correspondence with the recorded records is shown respectively.

  FIG. 6 is a flowchart showing a processing flow of a replication method which is an embodiment of the data input / output device according to the present invention.

  In the difference method shown in FIG. 5, (A) initial state, (B) record insertion (S131 to S135) and (D) record deletion (S151 to S155) are the same as the duplication method shown in FIG. Therefore, the description is omitted, and only (C) record update and search processing will be described.

[Record Update]
First, a case where a record is updated ((C)) in a state after record insertion ((B)) will be described (third process). The former rewrites the data A (RID-001) at 13:00 and updates the additional information from A-1 to A-2 (S20). The mobile device 1 outputs a record (input data) including the rewritten data A and position information to the small disk drive 2. The record (update record) is sent to the control unit 4 through the interface 3 in the small disk drive 2 together with a command [UPDATE] (third command) corresponding to the update process.

  When the control unit 4 receives the command and the update record, the database processing unit 47 determines the content of the database processing. In the present case, it is determined to receive a command [UPDATE] and update the record (S21 and S22).

  The record ID control unit 43 receives the fact that the requested process is a record update, and acquires the RID (RID-001 in this case) of the record to be updated from the index storage unit 51 (S241). Then, the difference data calculation unit 82 compares the data of RID-001 stored in the data storage unit 63 with the input data, and obtains a data portion (update attribute value) updated in the input data. The RID and the update attribute value are collected as one changed continuous record in the storage data generation unit 44 (S242). That is, in this difference method, the attribute (data A) that has not been changed is not included in the record stored in the data storage unit 63.

  Subsequently, the head control unit 45 moves the head to the end address of all the records stored in the data storage unit 63, in other words, to the next address of the record of RID-003 in the same manner as the above-described record insertion. Let

  Then, the updated continuous record generated in the storage data generation unit 44, that is, the record of RID-001 is sent to the disk unit 6, and is written in the data storage unit 63 by the data writing unit 61 (S243). At this time, since the head is located at the address next to the end address of all the stored records, the updated continuous record is added to the previous record in the data storage unit 63.

  After storing the update record, the index assigning unit 81 creates an index indicating the head of the address stored in the data storage unit 63, and the index is stored in the index storage unit 51 in the memory 5. At this time, the index pointer for RID-001 (additional information A-1) already stored in the index storage unit 51 is kept as it is by the index addition unit (not shown) in the control unit 4, and the updated RID- An index pointer for 001 (additional information A-2) is further added and stored as a list (S244). That is, in the present embodiment, the storage location of the update attribute value can be known by following the list added to the index in the index storage unit 51. Further, the end address of the data of RID-001 (additional information A-2) is stored in the end address storage unit 46. This completes the record update process (S245).

  The above-described index erasure unit 49 is a functional unit having a function necessary for the duplication method according to the first embodiment. However, by using an index addition unit instead of the index erasure unit 49, the present embodiment It is possible to execute the difference method according to the above.

  In this way, in the differential method, only changed (updated) data is appended to the disk, so that less disk capacity is required.

[Record Search]
Next, a case where a record is searched in the state after this record update ((C)) will be described (fifth process). In order to retrieve information of data A (RID-001) stored in the data storage unit 63, the user A performs a predetermined operation on the mobile device 1. The mobile device 1 outputs search information (input data) including at least information specifying a record to be searched to the small disk drive 2. The search information is sent to the control unit 4 via the interface 3 in the small disk drive 2 together with a command [SELECT] (fifth command) corresponding to the search process.

  When the control unit 4 receives the search information, the database processing unit 47 determines the contents of the database processing. In this case, it is determined to receive a command [SELECT] and search for a record (S21).

  Then, before performing disk access, the index optimization unit 41 performs index optimization, that is, rearranges (sorts) the physical positions of the search target records in the index stored in the index storage unit 51 ( S261). At this time, the index optimizing unit 41 optimizes the index in the index storage unit 51 based on the head position obtained from the head position monitoring unit 42 so that, for example, the access waiting time is minimized. Then, the optimized index is stored in the index storage unit 51.

  However, in this difference method (this embodiment), unlike the above-described duplication method (first embodiment), only the updated attribute is added, and the physical position of the updated attribute value is added to the index. A list to indicate is added. Therefore, when rearranging the physical position of the search target record in the index, it is necessary to optimize the index by including the amount added to this list in the search target record.

  Subsequently, the head position monitoring unit 42 notifies the head control unit 45 of the current head position. Further, the optimized index is output from the index storage unit 51 to the head control unit 45. The head control unit 45 controls the data reading unit 62 based on the information notified from the head position monitoring unit 42 and the index storage unit 51, and sequentially accesses the physical position of the record indicated by the optimized index. The record stored in the data storage unit 63 is read (S262). This reading is continued until the data reading is completed (S263), and when the data reading is completed, the record search process is completed (S264).

  In the differential method, the data stored in the disk is only the updated attribute value, so that the disk area can be saved.

  In both the first and second embodiments described above, sorting is performed in the order of the physical positions of the search target records using the index. Since the index is stored in the memory 5, this rearrangement can be performed in a shorter time than the access time to the disk. Further, random access can be converted into sequential access by this rearrangement, so that a high-speed search can be performed when the record interval is not wide.

  As described above, according to the embodiment of the present invention, it is possible to shorten the access waiting time for the head to move to the target record position even in a small disk drive with a low disk rotation speed. A data input / output method suitable for an active database and a data input / output device using the same can be provided.

[Other preferred embodiments]
(1) When small-size data is frequently written to the data storage unit 63 (magnetic disk), the overhead of disk access becomes high, so that the advantage of sequential access may not be sufficiently obtained. Therefore, a buffer unit 52 (log buffer) is provided on the memory 5, and input data is temporarily stored in the buffer unit 52 by a buffer writing unit 83 (buffer writing means). After that, it is desirable to adopt a group commit method that writes data to the data storage unit 63 together to realize efficient data processing.

  Data handled by the wearable device includes data such as position information that can be estimated from previous and subsequent records even if the data is lost. Since these data are low in importance, there is no problem even if data loss occurs. On the other hand, for example, if highly important data such as possession information is stored in the memory 5, there is a risk that the data may be lost due to a failure. Thus, by giving priority (importance) to the data, it is possible to control the disk flash.

  FIG. 7 is a schematic diagram for explaining group commit in an embodiment of the data input / output device according to the present invention. One form for realizing group commit is shown as “data record 1” in FIG. In this embodiment, when the input data is written to the buffer unit 52 by the buffer writing unit 83, the buffer monitoring unit 84 (buffer monitoring unit) monitors the storage area of the buffer unit 52. When the buffer monitoring unit 84 detects that the storage area of the buffer unit 52 is filled, the flash control unit 86 (first flash control unit) controls the data writing unit 61 and stores it in the buffer unit 52. The stored data is written into the data storage unit 63.

  Another form for realizing the group commit is shown as “data recording 2” in FIG. In this embodiment, when the input data is written to the buffer unit 52 by the buffer writing unit 83, the buffer attribute monitoring unit 85 (buffer attribute monitoring means) monitors the attribute of the input data. When the buffer attribute monitoring unit 85 detects that the input data has a high priority (predetermined attribute), the flash control unit 86 (second flash control means) controls the data writing unit 61 to The data stored in the unit 52 is immediately written into the data storage unit 63.

  This priority can be realized, for example, by pressing a predetermined button when storing data in the small disk drive 2 via the mobile device 1. In other words, the priority may be given in the form of data attributes by pressing a predetermined button or the like.

  (2) In the above embodiment of the present invention, since the system is assumed to be used by individuals, the total number of records is small. Therefore, for example, an index can be developed on the memory of a small disk drive, and search performance can be improved. However, the data on the memory may be lost when a power failure occurs or when the process ends abnormally. In another embodiment of the invention, the index is periodically stored on a small disk drive.

  For example, when a predetermined time elapses, the index restart unit 48 reads a set of indexes from the index storage unit 51 and stores them in the data storage unit 63 with a time stamp via the data writing unit 61. This is realized.

  When a system failure occurs, first, all stored indexes are read from the data storage unit 63 via the data reading unit 62. Then, the time stamp of the read index is checked to identify the latest index group. Subsequently, the data is restored by reading the log data after the latest index is stored from the data storage unit 63. Thereby, it is possible to solve the fault tolerance.

  (3) In the above embodiment of the present invention, it is assumed that a temporal query for retrieving information from the past to the present is executed. Therefore, data is not deleted, and all changed data is stored in the disk. Is added to the next address. Therefore, unused space does not occur in a distributed manner, and the disk can be used efficiently.

  (4) In the above embodiment of the present invention, the relational data model has been described. However, the present invention is not limited to this, and can be applied to other data models such as a card type and an object type. .

  In this example, a system (construction and test) was performed on Linux using a personal computer (PC) connected to a small disk drive. Note that Gentoo 1.4.1 (Kernel Version 2.6) is used as Linux, IBM ThinkPad (CPU: Pentium (registered trademark) 3, Memory: 384MB) as PC, and HGST's Microdrive DSCM-11000 as small disk drive 2. used. Table 1 shows the performance of this microdrive.

  The DBMS processing in this embodiment is performed using the CPU (control unit 4) and the memory 5 inside the PC, and the data is stored in the small disk drive 2.

  In order to construct a write-once file system according to the present invention, a disk is accessed using low-level input / output of C language. With low-level input / output, the disk can be accessed sequentially, and data can be directly transferred to and from the disk without using the buffering function provided in advance by the OS. On the other hand, if high-level input / output is used, the buffering function of the OS is used, and data is stored as long as there is a sufficient buffer capacity from the disk reading location. As a result, when using the records in the buffer, there is an advantage that processing can be performed at high speed without the need for disk access. However, since the data to be stored in the buffer is read, the head moves by the buffer capacity. Although the area unnecessary for the inquiry uses 115 KB as a boundary, the sequential access and the random access are properly used. However, since the head moves by the buffer capacity, the optimum disk access cannot be performed.

  In this embodiment, since the disk is directly controlled, the head is on the disk track where the last access was made, and it is possible to access the record recording location with an average waiting time (8.33 ms). .

  In this embodiment, the disk access performance is verified using a record set used in the Wisconsin benchmark. The Wisconsin benchmark is a test method for reading and writing data with 208 bytes of data as one record. One record is composed of 13 integer type variables (4 bytes) and 3 character type arrays (52 bytes) consisting of 52 characters. When the record size is 208 bytes, the micro drive manufactured by HGST reads or writes about 450 to 700 records in sequential access within the access waiting time (20.33 ms: average seek time 12 ms + average waiting time 8.33 ms). It is possible to execute.

  FIG. 8 is a diagram showing the disk read / write time of 10 MB data in one embodiment of the data input / output device according to the present invention, where the solid line is write, the alternate long and short dash line is read, and thick. A solid line shows the average of those four sections (number of records). As can be seen from this figure, almost the same result is obtained for both writing and reading.

  In the initial stage of data processing, since the buffer in the small disk is used, processing is performed at high speed. Further, since the disk capacity is 1 GB, the processing is completed with 108 records (10 MB × 108 to 1 GB). Since 10 MB of data is read and written in about 5 seconds on average without using the on-disk buffer, the read / write speed is about 2.0 MB / sec. When this is compared with the continuous read / write speed (2.5-4.1 MB / sec) shown in Table 1, it can be seen that performance of 50% to 80% is obtained. If storage at 2.0 MB / sec is possible, it is possible to store at a bit rate of 384 kbps, which is the MPEG4 standard, on a mobile device that stores video and the like. It is also possible to make an inquiry while storing the video.

  Next, a comparison is made regarding the processing time when writing is performed by random access and sequential access. FIG. 9A is a schematic diagram showing a data storage state in random access and FIG. 9B is a sequential access.

  As shown in FIG. 9A, in the case of random access, the next record position is specified from the pointer, the head is moved, and the next record is read. That is, the head moves every time a record is read. In random access, a record ID, data, and a pointer to the next record are handled as one record.

  On the other hand, as shown in FIG. 9B, in the case of sequential access, the data is written continuously, so the head does not move and the data is read continuously. In sequential access, a record ID and data are handled as one record.

  In this embodiment, since the data access speed in random access is significantly inferior to that in sequential access, the access waiting time can be obtained from the difference between the processing speeds of the two.

  FIG. 10 is a graph showing the relationship between the number of records in random access and sequential access and access (response) time (access waiting time + disk access) by 208 byte (1 record) record processing. As can be seen from this figure, the sequential access has a much faster access time than the random access. However, from this figure, it is not possible to compare the access times for sequential access and random access.

  Therefore, FIG. 11 shows the relationship between the number of records and access (response) time when sequential access is performed for 10,000 records. As can be seen from FIG. 10, in the random access, about 55 records are processed per second. That is, the time required to write 100 records by random access is 1,998.4 ms. On the other hand, as can be seen from FIG. 11, in the sequential access, about 9,000 records are processed per second. That is, the time required to write 100 records by sequential access is 14.41 ms. The access waiting time (the value obtained by dividing the time required to write 100 records by the number of records 100) is 19.84 ms, and the access waiting time determined from Table 1 (20.33 ms: theoretical value). ) Is a reasonable value. From this result, it can be seen that the access waiting time requires much time compared to the disk access.

  Subsequently, the disk access processing time in page units and record units is compared. For a record indicating single data of an arbitrary size, a page is a data set of a specific size and generally includes several records of data.

  If disk access is repeated frequently, disk access overhead occurs. For this reason, by using page unit access and writing data collectively, the disk access frequency can be reduced and data processing can be performed efficiently. Note that the access speed is compared with the page size of 2 KB for writing 9 records at a time, 4 KB for writing 19 records at a time, and 8 KB for writing 39 records at a time.

  FIG. 12 is a graph showing the relationship between the number of records and the access (response) time in disk access using a buffer and disk access not using a buffer when the page size is 4 KB. When writing 4,000 records, it takes 68.9 ms for page unit access and 227.7 ms for record unit access. From this result, it is understood that the overhead of disk access can be greatly reduced by using the buffer.

  Next, FIG. 13 is a graph showing the relationship between the number of records and access (response) time when the buffer capacity is changed to 2 KB, 4 KB, and 8 KB. When a 2 KB buffer was used, a result that seems to have caused disk access overhead was obtained. When the buffer capacity is changed to 4 KB and 8 KB, there is no difference in access time between the two. Therefore, it is appropriate to use a 4 KB buffer that stores a small amount of data on the memory.

  From the above, it can be seen that sequential access in units of 4 KB pages is the most efficient processing method.

  In this embodiment, since the index is set using the record position stored on the disk as a key, the record position can be specified in advance. Then, disk access is performed by sorting the storage locations of the records in the order of addresses. Therefore, it is possible to perform a search while minimizing the movement of the head. Further, since the read record interval can be obtained based on the index, random access is adopted when the record interval is wide and the time required for skipping by sequential access exceeds the head movement time by random access. The theoretical access waiting time obtained from Table 1 is 20.33 ms, but as described above, it is known that the access waiting time in this embodiment is 19.84 ms. The amount of data that can be read by sequential access within 19.84 ms is approximately 112 KB, and a 208 byte record corresponds to approximately 600 records. That is, when the skip area is 112 KB or more, it is effective to use random access rather than sequential access because the access time can be shortened.

  A sufficient number of records were prepared, and the access time in 300 times of search processing was measured. In the n-th enforcement, n records are searched. At this time, the interval between the records is artificially given so that {1, 3, 5,..., 2n + 1} is randomly selected without duplication. That is, when n is small, there are densely searched records. For this reason, skipping by sequential access is used. On the other hand, when n is large, the sequential access is selected in the portion where the record interval is dense, but the random access is selected in the sparse portion.

  Specifically, in consideration of the above-mentioned characteristics of the disk, it is set to switch to random access when the record interval is about 112 KB, that is, about 208 pages or more in a 208 byte record.

  FIG. 14 is a graph showing the relationship between the number of records and the access (response) time when the access method is changed. The horizontal axis indicates the number n of records to be searched, and the vertical axis indicates the access time required for the search. In this figure, the solid line is the result of this example. For comparison, the results for all sequential accesses (coarse dotted line) and for all random accesses (thin dotted line) are also shown.

  As can be seen from FIG. 14, sequential access is selected when the number of search records n is small, that is, when the record interval is close. As the number of search records n increases, the portion where the record interval is sparse increases, and the access time due to sequential skipping increases in a quadratic function. When the record interval exceeds 112 KB, the time to reach the target record is faster when the head is moved by random access than by sequential access. When the number of search records further increases from the branch point, it is faster to use random access. As can be seen from FIG. 14, the disk access is optimized according to this embodiment.

  As described above, it is possible to always realize high-speed disk access by properly using random access and sequential access. This embodiment is particularly effective in a system for personal use, in which it is considered that there is little opportunity for search processing with a wide skip area to occur.

  Also, when using a small disk drive in a mobile device, it is necessary to consider the problem of power consumption. The disk drive mainly consumes power when the head is moved or when the disk is started. In this embodiment, since the movement of the head is scheduled and the disk is accessed as sequentially as possible, the speed of data access is increased and the movement of the head is minimized.

  In general, a small disk drive stops the rotation of the disk when the disk is not accessed. However, when a small disk drive is mounted on a mobile device, GPS data and the like are added every few seconds, so repeated stopping and starting of the disk gives torque to the disk. As a result, extra power is consumed. Therefore, it is desirable that the disk is always rotated while the small disk drive is turned on so that frequently arriving data can be stored, so that it does not stop or start. Thereby, power consumption can be kept low. Further, it is not necessary to wait until the disk is started from the disk stopped state, and frequently arriving data can be stored.

It is a functional block diagram in one Embodiment of the data input / output device which concerns on this invention. It is a schematic diagram which shows the correspondence of the data storage of a replication system which is one Embodiment of the data input / output device which concerns on this invention, and an index. It is a flowchart which shows the flow of a process of the replication system which is one Embodiment of the data input / output device which concerns on this invention. It is operation | movement explanatory drawing of the search process of the replication system which is one Embodiment of the data input / output device which concerns on this invention. (A) is before index optimization, and (b) is after index optimization. It is a schematic diagram which shows a response | compatibility with the data storage of a difference system which is one Embodiment of the data input / output device concerning this invention, and an index. It is a flowchart which shows the flow of a process of the difference system which is one Embodiment of the data input / output device which concerns on this invention. It is a schematic diagram for demonstrating the group commit in one Embodiment of the data input / output device which concerns on this invention. FIG. 4 is a diagram showing a disk read / write time of data in units of 10 MB in an embodiment of the data input / output device according to the present invention. (A) is a schematic diagram showing a data storage state in random access, and (b) is a sequential access. 6 is a graph showing the relationship between the number of records in random access and sequential access and access time in one embodiment of the data input / output device according to the present invention. 6 is a graph showing the relationship between the number of records and the access time in sequential access by one-record processing in an embodiment of the data input / output device according to the present invention. 6 is a graph showing the relationship between the number of records in sequential access and the access time when there is a buffer and when there is no buffer in an embodiment of the data input / output device according to the present invention. 6 is a graph showing the relationship between the number of records in sequential access and the access time when the buffer capacity is changed in one embodiment of the data input / output device according to the present invention. 5 is a graph showing the relationship between the number of records and the access time when the access method is changed in one embodiment of the data input / output device according to the present invention. It is an external appearance block diagram which shows the operation surface side of the mobile telephone connected with the data input / output device which concerns on this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Mobile device 2 Small disk drive 3 Interface 4 Control part 5 Memory 6 Disk part 41 Index optimization part 42 Head position monitoring part 43 Record ID issuing part 44 Storage data generation part 45 Head control part 46 Termination address storage part 47 Database processing part 48 index re-storing unit 49 index erasing unit 81 index assigning unit 82 differential data calculating unit 51 index storing unit 52 buffer unit 61 data writing unit 62 data reading unit 63 data storing unit

Claims (18)

A data storage means for storing data, comprising a magnetic disk;
Head position monitoring means for monitoring the position of the head for reading data from the data storage means;
A semiconductor memory device comprising index storage means for storing an index for designating an address of data stored in the data storage means;
When reading target data that is data to be read from the data storage means, the address of the target data stored in the index storage means is designated based on the position of the head obtained from the head position monitoring means Index optimization means for optimizing the index and storing the index in the index storage means;
Data reading means for controlling the head based on the index stored in the index storage means and reading the target data;
A data input / output device comprising:   2. The data input / output device according to claim 1, wherein the magnetic disk is a small disk drive having a diameter of 1 inch or less.   The index optimization means optimizes the index so that the access waiting time of the head is minimized when reading data, based on the position of the head obtained from the head position monitoring means. 2. The data input / output device according to 2.   4. The data input / output device according to claim 1, further comprising data writing means for writing data to the data storage means.   5. The data input / output device according to claim 4, wherein the data writing means writes all logs into the data storage means.   6. The data input / output device according to claim 4, wherein the data writing unit writes data to an address next to a terminal address of data stored in the data storage unit.   7. The data input / output device according to claim 4, further comprising an index re-storing unit that periodically stores an index in the data storage unit by the data writing unit.   8. The data input / output device according to claim 4, further comprising record ID control means for issuing a record ID for specifying the input data received together with the command based on the command. Based on the input data and the first command, a new record ID for specifying the input data is received from the record ID control means, and the record ID and the input data are continuously stored in the data storage means by the data writing means. First processing means for writing as:
Creating an index for specifying the address of the continuous data, and storing the index in the index storage means;
The data input / output device according to claim 8, further comprising: Based on the input data and the second command, a record ID for specifying the input data is received from the record ID control means, and the record ID and the input data are written as continuous data in the data storage means by the data writing means. A second processing means;
In the index storage means, an index of data having the same record ID as the record ID for specifying the input data is deleted, an index for designating an address of the continuous data is created, and the index is stored in the index storage means Change means,
The data input / output device according to claim 8, further comprising: Based on the input data and the third command, a record ID for specifying the input data is received from the record ID control means, stored in the data storage means, and data having the same record ID as the input data and the input data A third processing means for obtaining an updated data portion in the input data, and writing the record ID and the updated data portion as continuous data in the data storage means by the data writing means; ,
In the index storage means, an index addition means for adding the address of the continuous data as a list to the index of the data having the same record ID as the record ID for specifying the input data, and storing the list in the index storage means;
The data input / output device according to claim 8, further comprising: Based on deletion information including at least information specifying data to be deleted and a fourth command, a record ID for specifying the data to be deleted is received from the record ID control means, and the record ID and the data to be deleted are Fourth processing means for writing a mark indicating the presence as continuous data to the data storage means by the data writing means;
In the index storage means, index erasure means for erasing the index of data having the same record ID as the record ID for specifying the input data;
The data input / output device according to claim 8, further comprising:   Based on search information including at least information specifying data to be searched and a fifth command, a record ID for specifying the data to be searched is received from the record ID control means, and an index of data having the record ID is obtained. The apparatus further comprises a fifth processing means for reading the data to be searched from the data storage means by the data reading means based on the index specified in the index storage means and stored in the index storage means. The data input / output device according to claim 8. The semiconductor memory device further comprises buffer means,
The data input / output device includes buffer writing means for storing input data in the buffer means;
Buffer monitoring means for monitoring a storage area of the buffer means;
When the buffer monitoring unit detects that the storage area of the buffer unit is full, the first flash control causes the data writing unit to write the data stored in the buffer unit to the data storage unit The data input / output device according to claim 4, further comprising: means. The semiconductor memory device further comprises buffer means,
The data input / output device includes buffer writing means for storing input data in the buffer means;
Buffer attribute monitoring means for monitoring attributes of data stored in the buffer means;
Second flash control means for causing the data writing means to write the data stored in the buffer means to the data storage means as soon as a predetermined attribute is detected in the input data by the buffer attribute monitoring means; The data input / output device according to claim 4, further comprising:   16. The data input / output device according to claim 1, wherein the magnetic disk as the data storage means is always rotated while the data input / output device is powered on. .   The data input / output device according to claim 1, wherein the data input / output device is detachable from a mobile device. A head position monitoring step for monitoring the position of the head for reading data from the magnetic disk;
An index storage step of storing in the semiconductor storage device an index that specifies an address of data stored in the magnetic disk;
When reading target data, which is data to be read from the magnetic disk, based on the position of the head obtained from the head position monitoring step, the target data stored in the semiconductor storage device by the index storage step An index optimization step of optimizing an index for designating an address and storing the index in the semiconductor memory device;
A data reading step of controlling the head based on an index stored in the semiconductor memory device and reading the target data;
A data input / output method comprising:

Images