JP5919529B2 - Data storage system - Google Patents

Description

  The present invention relates to a data storage system for storing data in a distributed database.

  Conventionally, various databases accessed through a telecommunication line such as the Internet have been operated. In this type of database, data from many terminals is repeated, such as a database that collects observation data related to weather at multiple points on the earth, and a database that collects meter reading data from consumers such as electricity and gas. Some things need to be collected.

  Since the amount of data handled by this type of database is enormous, it is conceived that the database is distributed and stored in a plurality of computer systems instead of constructing the database with one computer system (for example, , See Patent Document 1). Therefore, a database management system for managing this kind of database integrates a function of storing data in a large number of physically existing computer systems and a plurality of computer systems into one computer. It is necessary to have a function to handle like a system. That is, it is necessary to prevent the user from being conscious of a plurality of computers even though the data is stored in a distributed manner. This type of technology is called a distributed database.

Japanese translation of PCT publication No. 2002-528817

  In the distributed database described above, computer systems having different hardware resources may be mixed. In the database system, among the hardware resources, the difference between the performance of a CPU (Central Processing Unit) and the storage capacity of the storage device greatly affects the throughput. Therefore, due to variations in hardware resources in computer systems constituting the distributed database system, variations may occur in the writing time when storing data and the response time when providing data.

  Further, when data is stored in a plurality of computer systems constituting a distributed database system, access may be concentrated on a single computer system. In this case, the corresponding computer system becomes a bottleneck. That is, the overall throughput of the distributed database system is reduced.

  The present invention makes it possible to suppress the occurrence of a bottleneck even if there are variations in hardware resources among a plurality of computer systems constituting a distributed database system. As a result, computer systems with inferior hardware resources are mixed. It is an object of the present invention to provide a data storage system that makes it possible to construct a distributed database that is unlikely to cause a decrease in throughput.

A data storage system according to the present invention is a data storage system for storing data in a distributed database configured using a plurality of computer systems, and includes a plurality of virtual servers having the same storage capacity as constituent units of the distributed database, Selecting means for determining a virtual server for storing data from the virtual server, the virtual server including a data storage unit that divides a storage area for storing data into a plurality of tables having the same unit size, The storage capacity of the virtual server is determined so that the minimum storage capacity of the storage capacity of the plurality of computers is the upper limit, and the selecting means identifies the set to which the data belongs. Data belonging to the same set is stored in the same virtual server and the same table using the data ID expressed numerically. As will be, to determine a virtual server and a table for storing data, the data to be stored in distributed databases, is dispersed in the virtual server, and is dispersed in the table, characterized in that stored.

  In this data storage system, the data storage unit preferably includes a plurality of types of storage units that store data for each type.

In this data storage system, distributed database, storing data in the data storage unit, provides the data from the data storage unit, it is preferable to provide a management function unit that performs deletion of data in the data storage unit.

In this data storage system, a data ID for identifying a set to which the data belongs is added to the data, and the selecting means uses the remainder obtained by dividing the data ID by the number of virtual servers with respect to the data ID. And by specifying the table by the remainder obtained by dividing the data ID by the number of tables, it is preferable to store the data in a distributed manner so that data having the same data ID is stored in the same table.

  According to the configuration of the present invention, even if there are variations in hardware resources among a plurality of computer systems constituting a distributed database system, occurrence of a bottleneck is suppressed, resulting in a computer system with inferior hardware resources. It is possible to construct a distributed database that is less likely to cause a decrease in throughput even if they are mixed.

It is a block diagram which shows embodiment. It is a schematic block diagram same as the above. It is operation | movement explanatory drawing same as the above. It is a block diagram which shows other embodiment.

  In an embodiment described below, a database system stores meter reading data obtained by collecting meter reading values of meters installed in consumers such as electricity, gas, and water through an electric communication line, and provides stored data. The case where it is used for a meter-reading system is assumed. In the remote meter reading system, it is required to store meter reading data obtained every certain time (for example, 10 minutes), and it is necessary to sort and manage meter reading data for each customer. Therefore, in this embodiment, the database system includes a large number of tables in which meter-reading data collected at regular intervals is stored as records, and records corresponding to one customer are stored together in one table. I try to do it. The storage area of each table has the same unit size. In addition, a plurality of customer records are registered in one table. That is, meter reading data of each customer is sorted into a table.

  The embodiment described below is not intended to limit the use of the database system to a remote meter-reading system, but is a use in which many records of the same type are stored in one table and the sorting of records is necessary. Similar techniques can be employed. Applications that handle such data include, for example, the storage of meteorological observation data for a large number of observation points, the storage of operating status monitoring data for a large number of facilities, and multiple locations for disaster prevention and crime prevention. There are uses for storing monitoring data. In these applications, it is necessary to sample and store data regularly or irregularly. These applications are examples, and in applications where fixed point observation of data is performed, the effects described later can be expected by storing data using the technology described in this embodiment.

  Assume that a large number (for example, 2500) tables are provided in the database system, and a large number (for example, 1 million) records are stored in each table. In this numerical example, meter reading data is collected every 10 minutes, and one meter reading data is associated with one record, and if it corresponds to 1 million customers, Thus, the meter reading data of 2500 records can be stored. That is, meter reading data for about 17 days is stored. If the upper limit of the table capacity is 2 GB, 2 kB can be used for one record. Therefore, the capacity required for storing meter-reading data in the entire database system is 5 TB (2.5 billion records).

  As described above, when a large amount of data is generated at a relatively short time interval, if one computer system tries to cope with it, input / output for collecting and providing data becomes a bottleneck, and exclusive control is performed. Increased frequency. Therefore, there arises a problem that a high-performance computer system having high input / output speed and high internal processing capability is required. A distributed database is known as a technique for dealing with this type of problem. That is, the distributed database includes a database management system that accesses a database constructed by a plurality of computer systems as a single database. When a distributed database is used, data is distributed and stored in a plurality of computer systems, so that it is possible to avoid a bottleneck in a database constructed with one computer system.

  However, when there are a plurality of computer systems (hereinafter referred to as “entity systems”) as physical entities constructing a distributed database, entity systems having different hardware resources may be mixed. In a database system, the difference between the performance of a CPU (Central Processing Unit) and the storage capacity of a storage device among hardware resources greatly affects the throughput. For this reason, due to variations in hardware resources of the entity systems constituting the distributed database system, variations occur in the writing time when storing data and the response time when providing data.

  In addition, when data is stored in a plurality of entity systems constituting a distributed database system, access may be concentrated on a single entity system. In this case, the corresponding entity system becomes a bottleneck. That is, the overall throughput of the distributed database system is reduced.

  This embodiment employs the following configuration so that the occurrence of bottlenecks is suppressed even if there are variations in hardware resources among a plurality of entity systems that constitute the distributed database system. In other words, by adopting the following configuration, it is possible to provide a distributed database that is unlikely to cause a decrease in throughput even when a substantial system with inferior hardware resources is mixed.

  As shown in FIG. 2, the present embodiment includes a plurality of (three in the illustrated example) entity systems 21, 22, and 23, and at least one entity system 21 and the other entity systems 22 and 23 communicate with each other. A description will be given using a configuration in which communication is performed through the line NT1. The telecommunications line NT1 is assumed to be a public network that is a wide area network such as the Internet, but may be a dedicated network or a private network. Each of the actual systems 21, 22, and 23 includes a large-capacity storage device such as a hard disk device, a CPU that executes a program, and a communication interface device for communicating with other devices through an electric communication line NT1.

  As shown in FIG. 1, the entity systems 21, 22, and 23 integrally function as the server 20, and construct a client server system using the user terminal 31 connected to the telecommunication line NT1 as a client. This client server system has a three-layer architecture including a presentation layer, an application layer, and a data layer, and the presentation layer is realized by a user terminal 31.

  The server 20 corresponds to the data layer. In the illustrated example, the server 32 corresponding to the application layer is provided separately from the server 20. The server 32 has a function of making a request corresponding to the processing requested from the user terminal 31 to the server 20 and returning a response of the server 20 to the user terminal 31.

  As will be described later, the server 20 is divided into a plurality of virtual servers (hereinafter referred to as “virtual servers”) 10, and the server 32 is formed by integrating a plurality of virtual servers 10. Provided with a function of operating as the server 20. This function is realized by an API (Application Program Interface) provided in the server 32. That is, the API of the server 32 includes a function or instruction for realizing a function of selecting the virtual server 10 corresponding to the request from the user terminal 31. Further, the API of the server 32 has a function of handling a plurality of entity systems 21, 22, and 23 as a single server 20.

  In other words, the server 20 is a distributed database composed of a plurality of entity systems 21, 22, and 23, and the virtual server 10 is a structural unit of the server 20. In the present embodiment, the virtual servers 10 have the same storage capacity.

  Since the present embodiment assumes a remote meter reading system, a collection server 33 for collecting meter reading data and storing it in the server 20 is also connected to the telecommunication line NT1. The collection server 33 is provided with an API for realizing a sorting function unit (sorting means) 331 that determines a virtual server 10 that stores meter-reading data from among a plurality of virtual servers 10. The operation of the sorting function unit 331 will be described later.

  The virtual server 10 includes a data storage unit 101 that stores data, and a management function unit 102 that provides functions of a database management system. The management function unit 102 is realized by an API provided in the entity systems 21, 22, and 23. The management function unit 102 performs transaction management in addition to functions for operating data such as storing data in the data storage unit 101, providing data read from the storage device, and deleting data stored in the storage device. It has functions. This type of function in the management function unit 102 is the same as the function of a general database management system. However, the management function unit 102 includes three layers including an interface layer 1021 that performs transaction management, an upper layer 1022 that performs database control, and a lower layer 1023 that performs database operations.

  As described above, since the server 20 is divided into a plurality of (for example, 100) virtual servers 10, the data storage unit 101 is also divided. Each virtual server 10 includes a management function unit 102. Here, in order to simplify the explanation, it is assumed that the data storage unit 101 and the management function unit 102 have a one-to-one correspondence. In other words, one data storage unit 101 and one management function unit 102 are provided for each virtual server 10.

  In the present embodiment, the storage capacities of the data storage units 101 are set to be equal to each other in the virtual server 10. Accordingly, the storage capacities of the data storage units 101 of the respective virtual servers 10 are equal regardless of the storage capacities of the storage devices of the entity systems 21, 22, and 23 constituting the virtual server 10. In the actual systems 21, 22, and 23, as a general tendency, the storage capacity of the storage device increases as the processing capability increases. From this, it can be considered that by making the storage capacity of the data storage unit 101 equal in the virtual server 10, the throughput of the virtual server 10 becomes substantially equal regardless of the processing capabilities of the entity systems 21, 22, and 23.

  As illustrated in FIG. 3, the data storage unit 101 of one virtual server 10 is divided into storage units having the same unit size, and each storage region of each unit size forms a table 1011. Therefore, the data storage unit 101 includes a plurality (100 in the illustrated example) of tables 1011. In the illustrated example, a 4-digit number for identification is assigned to each table 1011. In the illustrated example, the numbers for identifying the table 1011 are “0001” to “2500”. The last two digits of the number identifying the table 1011 correspond to the number identifying the virtual server 10, and “01”... “99” “00” is used. “00” corresponds to No. 100.

  Further, identification information (hereinafter referred to as “data ID”) is individually set for each consumer, and the illustrated example assumes a case where 1 million consumers exist. That is, 1 million kinds of data IDs are assigned to records registered in the table 1011. Here, records with the same data ID are stored in the same table 1011. Further, the data ID is set as a numerical value or is appropriately converted into a numerical value, and assigned from 1 to 1000000 in order. Here, since the data is meter reading data, the data ID corresponds to the customer. However, if generalized, it can be said that the data ID is given to identify the set to which the data belongs.

  In order to construct the virtual server 10 as described above, it is necessary to determine the number of virtual servers 10, the storage capacity of one virtual server 10, the storage capacity of one table 1011, and the like.

  Hereinafter, a procedure for properly constructing the virtual server 10 will be described. First, in order to determine the scale of the database system, the number of assumed consumers, that is, the maximum number of assumed data IDs is required.

  Further, since the virtual server 10 is configured by the entity systems 21, 22, and 23, it is not preferable that one virtual server 10 is configured across the plurality of entity systems 21, 22, and 23. Therefore, when constructing the virtual server 10, it is necessary to know the storage capacity of each of the entity systems 21, 22, and 23.

  The storage capacity of the data storage unit 101 in the virtual server 10 is limited to the minimum storage capacity of the storage devices included in the real systems 21, 22, and 23. Therefore, the minimum value of the storage capacity assumed for the storage devices in the real systems 21, 22, and 23 is obtained, and the storage capacity of the data storage unit 101 in the virtual server 10 is determined using this minimum value.

  Furthermore, the amount of data (number of records) stored in one table 1011 and the storage capacity of the table 1011 are design values, and are appropriately determined according to the amount of data acquired and stored from a consumer. The upper limit value of the data amount stored in one table 1011 is a value determined as the storage capacity of the data storage unit 101 in the virtual server 10. However, in practice, it is preferable to make it sufficiently smaller than the storage capacity of the data storage unit 101.

As described above, in order to construct a distributed database using the virtual server 10, the following information is necessary as a design condition. That is, the maximum number of data IDs assumed, the number of records stored for each data ID, the storage capacity of the real system, the storage capacity of the virtual server 10, the storage capacity of the table 1011 and the maximum number of records stored in the table 1011 are required. It is.

  When these parameters are obtained, the number of virtual servers 10, the maximum value of the total number of tables 1011, and the number of tables 1011 in one virtual server 10 are obtained. Further, the virtual server 10 can be associated with the actual systems 21, 22, and 23.

  If the storage capacity of one virtual server 10 is Q and the storage capacity of the table 1011 is q, the number of tables 1011 in one virtual server 10 is Q / q. Furthermore, if the maximum number of assumed data IDs is N, the number of records stored for one data ID is m, and the number of records stored in one table 1011 is n, the number of necessary tables 1011 is N × m. / N. From this, the number of virtual servers 10 is (N × m / n) / (Q / q) = Nmq / nQ.

  In the configuration example shown in FIG. 3, the assumed maximum number N of customers (data IDs) is 1 million, the storage capacity q of the table 1011 is 2 GB, and the number of records n stored in the table 1011 is 1 million. Furthermore, the number of records m stored for one consumer is 2500. In the configuration example shown in FIG. 3, the minimum value of the storage capacity in a plurality of physical systems is 50 GB. Therefore, the storage capacity Q of the virtual server 10 is also set to 50 GB. In the configuration example of FIG. 2, three entity systems 21, 22, and 23 are used. However, the configuration example of FIG. 3 uses more entity systems.

  When these numerical values are applied to the relational expression described above, the number of tables 1011 provided in one virtual server 10 is 50 GB / 2 GB = 25. Further, the number of virtual servers 10 configuring the database system is (1000000 × 2500 records × 2 GB) / (1000000 × 50 GB) = 100.

  Here, it is assumed that an entity system having a storage capacity of 100 GB and an entity system having a storage capacity of 50 GB are used as the entity system constituting the database system. If the storage capacity of the virtual server 10 is 50 GB, an entity system that configures only one virtual server 10 and an entity system that configures two virtual servers 10 are mixed.

  In the configuration example described above, it is assumed that the database system is configured using 40 100 GB entity systems and 20 50 GB entity systems. Assuming this, 20 of the 100 virtual servers 10 are allocated to the 50 GB entity system, and 80 are allocated to the 100 GB entity system. In this configuration, the number used to identify the virtual server 10 is configured by an entity system including only one virtual server 10 between the virtual servers 10 configured by the entity system including a plurality of virtual servers 10. It is preferable to assign the virtual servers 10 so that they line up.

  For example, “01” to “40”, “61” to “99”, “00” are used as numbers for identifying the virtual server 10 using the entity system including the plurality of virtual servers 10, and the single virtual server 10 is included. “41” to “60” are used as numbers for identifying the virtual servers 10 using the real system. Here, in an entity system including a plurality of virtual servers 10, the numbers for identifying the virtual servers 10 do not necessarily have to be consecutive, but in the following description, it is assumed that they are consecutive.

  By the way, when storing meter reading data using the database system having the above-described configuration, it is necessary to determine the table 1011 for storing meter reading data. That is, the selection function unit 331 of the collection server 33 obtains a virtual server 10 for storing meter reading data in the procedure described below, determines a substantial system corresponding to the virtual server 10, and further acquires meter reading data. The table 1011 to be stored is determined.

  The virtual server 10 that stores meter-reading data can be easily determined by using a data ID represented by a numerical value. That is, if the data ID is X and the number of virtual servers 10 is K, the number for identifying the virtual server 10 storing meter reading data corresponding to the customer is obtained by the remainder obtained by dividing X by K (that is, X mod K). However, when the remainder becomes 0, it is assigned to the virtual server 10 of “00”.

  For example, if the data ID = 50, (50 mod 100) = 50, so the corresponding meter reading data is stored in the 50th virtual server 10. Further, for example, if the data ID = 36935, (36935 mod 100) = 35, so the corresponding meter reading data is stored in the 35th virtual server 10. The number of the virtual server 10 is associated with the entity system by the lookup table or the like in the selection function unit 331 of the collection server 33, and therefore, which entity system is accessed is determined.

  Next, it is necessary to determine which table 1011 in the virtual server 10 stores the meter reading data. In the configuration described above, it is easy to determine the table 1011 for storing meter-reading data. When the total number of the table 1011 is L, when the data ID is X, the remainder obtained by dividing X by L is the number of the table 1011. (That is, X mod L). However, when the remainder becomes 0, the 2500th table 1011 is allocated.

For example, if the data ID = 50, (50 mod 2500) = 50, so the corresponding meter reading data is stored in the table 511 of the number 0050. Further, for example, if the data ID = 36935, (36935 mod 2500) = 1935, the corresponding meter reading data is stored in the 1935 table 1011 .

  As described above, by defining the virtual server 10 for storing meter-reading data, an entity system for storing meter-reading data is required. Therefore, a storage area corresponding to the virtual server 10 and the table 1011 is included in the range of the corresponding entity system. Can be extracted. That is, meter reading data can be stored in an appropriate table 1011.

  When the table 1011 for storing meter reading data is selected in such a procedure, when the data ID is a continuous numerical value, as shown by arrows in FIG. Stored.

  That is, meter reading data of the first customer is stored in a table 0001 provided in the first virtual server 10, and meter reading data of the second customer is stored in a table 0002 provided in the second virtual server 10. Stored in Similarly, the meter reading data of the 101st customer is stored in the 0101 table 1011 provided in the 1st virtual server 10 and the meter reading data of the 2500th customer is the 2500th of the 100th virtual server 10. It is stored in the table 1011. The meter reading data of the 2501 customer is stored in the 0001 table 1011 in the 1st virtual server 10. In short, the meter reading data whose remainder is 1 obtained by dividing the data ID by 2500 is stored in the table 1011 of the number 0001. Thus, for one million customers, meter reading data of 400 customers is stored in one table 1011.

  With the above-described operation, the meter reading data of each consumer is distributed and stored in different virtual servers 10 as if striping is performed using a plurality of hard disks. That is, access is not concentrated on a single virtual server 10, and meter reading data is stored faster. Further, even when the processing capacity of one virtual server 10 is relatively low, it is possible to construct a distributed database in which bottlenecks are unlikely to occur. That is, in the present embodiment, the processing speed when storing meter-reading data is equalized by equalizing the storage capacities of the virtual servers 10 and leveling the processing capabilities of the virtual servers 10.

  In the above-described configuration example, an example of a database that handles customer meter reading data has been described. However, there is a use in which a plurality of types of data are acquired and stored at one measurement point. For example, it is possible to store not only electricity meter reading data at consumers but also gas meter data and water meter reading data together, and store information related to crime prevention and disaster prevention at consumers. Etc. are also possible. That is, the use of the database is not limited to the use of remote meter reading, and in addition to numerical data, a plurality of types of data such as text data, audio data, and image data may be stored. Thus, when collecting different types of data, not only the data ID relating to the location where the data is obtained but also the data type name is added to the data so that the data type can be identified.

  By the way, this type of data has a different amount of information depending on the type of data. Therefore, when multiple types of data are mixed, if the data is handled in a batch, the type of data included in the batched data is changed. Correspondingly, a large variation occurs in the amount of information. For this reason, when this type of data is stored in one table 1011 in a state where a plurality of types of data are mixed, the access time of the virtual server 10 varies greatly.

  Therefore, as shown in FIG. 4, in the data storage unit 101 of the virtual server 10, a plurality (three in the illustrated example) of storage units 1012, 1013, and 1014 that store the data sorted according to the type of data are stored. It is desirable to provide it. In this case, the virtual server 10 is provided with a classification function unit 103 for distributing data types. The classification function unit 103 is configured using an API in the same manner as the management function unit 102. That is, the data transferred from the collection server 33 to the virtual server 10 is sorted for each data type in the classification function unit 103 of the virtual server 10 and stored in the respective type storage units 1012, 1013, and 1014.

Thus, Ru stored different kinds of data are distributed by type storage unit 1012,1013,1014. Therefore, even if the information amount differs for each data ID, the variation in the information amount (data amount) is suppressed for the data stored in the type storage units 1012, 1013, and 1014. That is, data stored in the type-specific storage units 1012, 1013, and 1014 may be associated with each data ID, and data may not be associated, but data stored in the data ID is associated. In this case, there is little variation in the information amount of the data. Therefore, the processing load for storing data in each type storage unit 1012, 1013, 1014 is equalized for each type storage unit 1012, 1013, 1014.

  In the configuration example of FIG. 4, a plurality of type-specific storage units 1012, 1013, and 1014 are provided in one virtual server 10. However, the type-specific storage units 1012, 1013, and 1014 are distributed to different virtual servers 10. Alternatively, part of the type-specific storage units 1012, 1013, and 1014 may be distributed to different virtual servers 10.

  By the way, when the data ID increases and exceeds the predicted maximum number of data IDs, it is necessary to expand the database system. In the above example, the maximum number of data IDs is 1 million. However, in order to collect data on a larger scale, it is necessary to increase the maximum number of data IDs. In such a case, the above-described configuration may be one unit and the number of units may be increased as necessary. For example, a group name is assigned to each unit, and numerical values given in order from 0 are used as the group name. In the configuration example described above, a total of 2.5 billion records can be stored for 1 million data IDs, so when one group is added, a total of 5 billion records are stored for 2 million data IDs. Is possible. That is, data in the range of data ID 1 to 1000000 may be stored in group “0”, and data in the range of data ID 1000001 to 2000000 may be stored in group “1”. In this example, the group name is a quotient obtained by dividing the data ID by 1000000.

  As described above, by expanding the database in units of one group, the selection function unit 331 of the collection server 33 only adds a procedure for identifying the group without changing other procedures. Data can be distributed. It should be noted that a server for dividing the group may be separately provided before the collection server 33. In other words, the database system may be expanded by providing a dedicated function for grouping and adding groups including the collection server 33.

  Whether or not the database system needs to be expanded depends on the total number of data IDs. Therefore, the collection server 33 or the server 32 may be provided with a function for acquiring the total number of data IDs (usually input from a keyboard or the like). (Acquired by operating the device). The collection server 33 or the server 32 calculates the number of groups necessary for data storage from the total number of acquired data IDs, and if the calculated number of groups is larger than the current number of groups, the database extension is made to the user terminal 31. You just have to give instructions.

  By the way, in general, when the database is expanded, data is stored in a newly provided entity system, so that it is considered that access concentrates on the new entity system. However, in this embodiment, since data is distributed and stored in a plurality of virtual servers 10 using data IDs, even when the database is expanded, access is concentrated on the newly provided entity system. There is nothing to do. In other words, access to the real system is not concentrated, access to the virtual server 10 is leveled, and as a result, the storage capacity of the virtual server 10 and the load due to access can be leveled.

  In the above-described example, a technique has been described in which the virtual server 10 and the table 1011 are determined when data is stored, and the data is transferred to the corresponding actual system. However, a similar procedure is used when data is extracted. For example, if data relating to a specific data ID is necessary, the virtual server 10 and the table 1011 may be determined by giving the data ID, and the data may be extracted from the table 1011 of the corresponding entity system.

  In the above-described configuration example, the collection of data to be stored and the use of the stored data use the telecommunication line, and the entity system is linked by communicating through the telecommunication line between the plurality of entity systems. It is not essential to communicate using a telecommunication line. In other words, data can be collected or used by an input device or an output device attached to the entity system, and a plurality of entity systems can be arranged in close proximity and can be linked appropriately. You may connect via an interface apparatus.

  Note that the API constituting the management function unit 102 may have a function of accepting various parameters related to the above-described database system. In this case, by setting parameters in the management function unit 102, the management function unit 102 may determine the storage capacity of the data storage unit 101 in the virtual server 10, the total number of tables 1011, the storage capacity of the table 1011, and the like. Good.

  In the above example, in order to determine the virtual server 10 and the table 1011 for storing data, the selection function unit 331 divides the data ID, but it may be distributed using other relationships such as random numbers. Good.

10 virtual servers 20 servers (distributed database)
21, 22, 23 Entity system (computer system)
101 Data storage unit 102 Management function unit 331 Sorting function unit (sorting means)
1011 Table 1012, 1013, 1014 Type storage unit

Claims (4)

A data storage system for storing data in a distributed database configured using a plurality of computer systems, wherein a plurality of virtual servers having the same storage capacity as a constituent unit of the distributed database and data from the virtual servers Selecting means for determining a virtual server for storing the virtual server, the virtual server comprising a data storage unit that divides a storage area for storing data into a plurality of tables having the same unit size, and the storage capacity of the virtual server Defines the storage capacity of the data storage unit so that the minimum storage capacity among the storage capacities of the plurality of computers is set as an upper limit, and the selecting means identifies the set to which the data belongs. Data belonging to the same set is stored in the same virtual server and the same table by using numerical data IDs The data storage is characterized in that the virtual server and the table for storing the data are determined, the data to be stored in the distributed database is distributed to the virtual server and stored in the table. system.   The data storage system according to claim 1, wherein the data storage unit includes a plurality of types of storage units that store data for each type.   3. The distributed database includes a management function unit that stores data in the data storage unit, provides data from the data storage unit, and deletes data in the data storage unit. The data storage system described.   A data ID for identifying a set to which the data belongs is added to the data, and the selecting means specifies a virtual server by a remainder obtained by dividing the data ID by the number of the virtual servers. The data is distributed and stored so that data having the same data ID is stored in the same table by specifying the table by a remainder obtained by dividing the data ID by the number of the tables. Item 4. The data storage system according to any one of Items 1 to 3.

Images