Electrical computers and digital processing systems: memory – Storage accessing and control – Specific memory composition
Reexamination Certificate
1999-06-03
2001-04-10
Kulik, Paul V. (Department: 2177)
Electrical computers and digital processing systems: memory
Storage accessing and control
Specific memory composition
C707S793000
Reexamination Certificate
active
06216203
ABSTRACT:
BACKGROUND OF THE INVENTION
The present invention relates to a file management system having a large capacity memory for storing long data, and more particularly to a data processing method and an apparatus therefor in a system for managing a multimedia data base, a map information data base and a knowledge data base in which character texts, figures and image data are stored.
In the system which handles long data such as character texts, figures and image data, since a logical record length of data is increased and an amount of data to be transferred between an input/output buffer set in a main memory in a central processing unit for processing data and an external memory unit for storing data is increased, a sufficient access performance is not obtained due to a limitation of a data transfer rate. Further, by repeating update, addition and deletion of expanded records, invalid areas are scatteringly formed in a file to thereby deteriorate a space efficiency of the file.
In this connection, JP-A-1-106217 discloses a technique that long data is divided and the divided data are stored in a plurality of secondary memory units in parallel. In this technique, the long data is divided to form divided files and information relative to memory locations of the divided files in the secondary memory units is stored in one of the divided files. When the long data is read out, the information relative to the memory location is read out from one of the divided files and the divided files in the secondary memory units indicated by the information are read out in parallel to restore the long data. When the data is updated, the restored data is updated and the updated data is divided again to store in the secondary memory units.
Further, JP-A-2-52305 discloses a partially updating method of image data in which in an image file apparatus for storing a document such as a drawing as an image file, an inputted document is divided into a predetermined number of divided image units to be stored and when any modification occurs in the document, only a divided image unit having contents which are to be modified is modified and stored again. In this technique, when the modification occurs, the whole image is not stored again and accordingly the image file can be updated at high speed.
SUMMARY OF THE INVENTION
The inventors have found the following problems as a result of studying the techniques in the prior arts.
In the former prior art technique, since data processing for all of the divided files is made in parallel, there is a problem in a speed of operation for processing only data of a portion of the divided files independently.
Further, the management of the input/output buffers and the improvement of operation speed thereof are not considered sufficiently. More particularly, the system must provide a large number of input/output buffers in the main memory in order to transfer data in parallel effectively. Since a record length of the multimedia data such as character texts, figures, images or the like is several hundred KB to several hundred MB, a main memory capacity required for input and output of one record is increased. Accordingly, when all of requests for record operation are made in unit of record, multi-processing of a plurality of requests for record operation is restricted by the main memory capacity and a time loss occurs due to waiting for an unoccupied input/output buffer, so that the throughput of the whole system is reduced.
In addition, in a retrieval instruction of record, only a partial item of data is used to make judgment of a retrieval condition which is made by reading data into the input/output buffer in the main memory, although all of data constituting the record is read in the buffer and accordingly useless reading of data having no relation to the retrieval condition is made in the case where the condition is not matched to thereby occupy the input/output buffer uselessly.
In the latter prior art technique, since only the image unit modified actually of the divided image units is recognized and stored again, input and comparison of correction data and unmodified data are necessary, so that useless input of image units having data not updated and useless occupation of the input/output buffer occur similarly.
Furthermore, in both of the prior art techniques, data to be handled is intended to be fixed length data, while handling of variable length data is not described. When the variable length data is divided simply in the same manner as in the prior art, there is the following problem.
When a length of divided sub-records is changed, the capacity of each of the divided files is varied and storage addresses such as physical addresses or relative byte addresses in a file storing the sub-records are also varied for each of the divided files. Accordingly, if the relation between the sub-records constituting one record is not established by holding a storage address of an actual value of a next corresponding sub-record in actual values of the sub-records, a random access to any record can not be made. On the other hand, when the relation between the sub-records is established, a storage address of the sub-record is not understood as far as a sub-record positioned just before the sub-record having a storage address is not inputted except the sub-record including an item which is a key for storage of the record and accordingly parallel input of sub-records can not be made.
On the contrary, when the records are of the variable length type and the records are divided to be stored in a plurality of files on condition of the maximum record length, an actual value of record having a short record length has no or less data to be stored in a rear divided file. Accordingly, actual values of sub-record having all or many useless areas are formed and the space efficiency of the rear divided file is deteriorated.
It is an object of the present invention to provide a data processing method and apparatus capable of processing long data effectively at high speed with minimum input/output process using a limited input/output buffer area.
It is a second object of the present invention to provide a data processing method capable of dividing long data into a plurality of sub-records to be stored and making operation of data in unit of sub-record.
It is a third object of the present invention to provide a data processing method capable of dividing variable length data into a plurality of sub-records effectively.
It is a fourth object of the present invention to provide a data processing method capable of dividing variable length data into a plurality of sub-records, storing the divided sub-records in a plurality of memory units dispersedly and accessing the stored data at high speed.
In order to achieve the above object, according to the present invention, a record constituting a unit of data operation is divided into a plurality of sub-records and the divided sub-records are stored in memory units, each capable of performing independent input and output process, in unit of sub-record. The number of sub-records necessary for required data operation is limited to the minimum and the data operation is performed in unit of one or more sub-records. The size of divided sub-records is desirably determined on the basis of unit of access to the memory unit for storing the sub-records. For example, when an external memory unit is a magnetic disk unit, access of data is controlled in accordance with a physical record length of the magnetic disk unit and accordingly if the size of divided sub-records is determined in accordance with the physical record length, recording/reproducing operation is made effectively at high speed.
The divided sub-records are stored in the memory units, each capable of performing independent input and output process, in unit of sub-record. It is desirable to store the plurality of sub-records in a plurality of memory units in parallel. The memory unit desirably uses an array type magnetic disk unit. Storing of sub-records can be made at higher speed by making storage addresses
Hitachi , Ltd.
Kulik Paul V.
Mattingly Stanger & Malur, P.C.
LandOfFree
Data processing method using record division storing scheme... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Data processing method using record division storing scheme..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Data processing method using record division storing scheme... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2450922