JP2006092503A - Multi-instance in-memory database - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To divide single data into sizes that can be loaded into main storage devices and handle the divisions for search, processing and the like in the same manner as the single undivided data so that an in-memory database is applied to a large scale database. <P>SOLUTION: Data having many tuples are grouped into data sets sized for handling on the main storage devices and are allocated with storage location information assigned. In search, a data set in a storage location is specified, and a data storage device is identified by the location information. If the target data set is on the main storage device, it is accessed direct. If not, the data set is accessed after it is checked in from a secondary storage device or the storage device on another networked electronic computer. The location information about the data sets is set according to characteristics such as a bias at the storage start of stored data and, with time series changes in the characteristics of the data, the relation between the data sets and the data storage devices is coordinated accordingly. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

    The present invention relates to an in-memory database and a grid computing system using the same.

    The basic technology of the database management system has started from the hierarchical type, has reached the object-oriented type through the network type and the relational type, but the relational type is still mainstream.

    In addition, the mainstream relational database management system has as its main component a magnetic disk as a data storage medium that has a large storage capacity and whose stored contents are relatively stable even when power supply is cut off. It has been assumed that a hard disk device is used as a secondary storage device.

    Hard disk devices as secondary storage devices are stable against power outages, but because they have many mechanical sliding parts, the failure occurrence rate is much higher than that of semiconductor storage devices. Access time is orders of magnitude slower. Further, since the magnetic medium that is easily affected by the surrounding magnetic field is used as the storage medium, the memory retention force is unstable with respect to the approach of the magnet. Furthermore, since read / write access is performed by contact of the magnetic header, the magnetic header is always placed in a location very close to the magnetic storage medium. For this reason, it is well known that the memory retention force of a magnetic storage medium is vulnerable to impact.

    Although there is a movement to use a semiconductor storage medium such as a non-volatile memory as a secondary storage device, the price per unit storage capacity is still expensive, and for economic reasons, its spread is used as a database management system. Is not popular.

    With the recent improvement of industrial production technology, volatile semiconductor storage media has been trending toward lower prices and larger capacities. Along with this, personal computers (hereinafter referred to as PCs) with large-capacity volatile semiconductor storage devices on the order of gigabits as main storage devices. ) Has also appeared. On the other hand, in a 32-bit CPU, it is meaningless to install a memory of 2 32 bytes, that is, 4,294,962,722,96 bytes (about 4 GB) or more as a main storage device for addressing reasons. However, with the advent of 64-bit CPUs, it is possible to have a memory space of 2 64 bytes, and it is theoretically possible to handle a terabyte order database in-memory. Actually, it is estimated that the semiconductor memory as the main storage device mounted on one 64-bit CPU is at most several tens GB level due to the limitation of the current integration technology of the semiconductor memory.

    The trend toward lower prices and larger capacities of the volatile semiconductor storage medium is to completely overturn the database management system, the main storage device is the main data storage device, and the secondary storage device is used for recovery in the event of a failure. In-memory database management system, which is positioned as a persistent storage medium, has appeared on the market as a commercial product.

    The in-memory database management system that has appeared on the market is not a database management system for transactions involving complicated update processing by a Maltuser, but a database management for a data warehouse (hereinafter referred to as DWH) for OLAP and data mining. Positioning as a system is reasonable, and it can be said that a database management system having a size of several tens of GB has been successfully implemented in memory.

JP-A-2004-227169 (PCT / JP2003 / 014390) JP 2004-145640 A JP 2000-339390 A

    In order to strengthen EU environmental problem regulations, traceability for proof of non-use of hazardous chemical substances such as industrial products and agricultural products, and notification and collection of messages such as prohibition of use to consumers when used incorrectly. In order to secure it, there is a demand for practical use of a large-capacity ultrahigh-speed database of the order of several hundred terabytes.

    Although the introduction of a 64-bit CPU-equipped PC has made it close to the realization of a large-capacity ultra-high-speed database, the degree of integration of semiconductor volatile storage devices used in main storage devices is in the order of several hundred terabytes per PC. The semiconductor volatile memory device is not installed. Therefore, it is required that the database can be distributed to a plurality of electronic computers arranged on the network. Therefore, a set of large-scale data having the same attribute is divided into data sets of a size that can be processed only by the main storage device on an electronic computer having only a few gigabyte order main storage device, Although it is conceivable to connect a plurality of electronic computers having only a few gigabytes of main storage to a network, a plurality of divided data sets are obtained from a user's electronic computer that performs data processing. The problem was that they could not be integrated and handled virtually as a single data set.

    In addition, it is ideal that all data sets are expanded on a main storage device on an electronic computer arranged on a network, but a secondary storage medium such as a hard disk device due to the limitation of the number of connected electronic computers. In some cases, it is required to efficiently use the secondary storage medium, and the secondary storage medium cannot be used efficiently.

    In addition, the computer connected to the network does not have uniform CPU performance, and the capacity of the mounted main storage device is also various. Even in time series, it can be placed on the network by rearranging or replacing old and new. The number of connected electronic computers changes, the performance of the CPU mounted on the electronic computer changes, and the capacity of the main storage device mounted on the electronic computer also changes. Another problem is that the location information of the divided data sets cannot be dynamically changed to cope with the diversity and changes, and in addition, the optimal arrangement is difficult.

    Furthermore, even if a large-scale data set is simply divided into a plurality of appropriately-sized data sets, a plurality of data sets are used to express forward and backward development of BOM (parts configuration table) in the production management system. It was also difficult to efficiently implement self-joining across the two.

    Therefore, the present invention enables a large data set to be handled as a single large data set by virtually integrating even if a large data set is divided into small data sets in the database system. While operating the database system, the number of one or more electronic computers arranged on the network constituting the database system, the performance of the CPU mounted on the electronic computer, the number of CPUs, or the electronic computer The object is to dynamically and optimally arrange the locations of the divided data sets in response to changes in resources such as the capacity of the main storage device mounted.

    In order to solve the above-described problems, in the present invention, as shown in claim 1, a large-scale data set having the same attribute, a part or all of data information to be stored is hashed or the like. Using an algorithm with specific parameters, the data is divided and arranged in a size that can be processed only by the main storage device of the electronic computer arranged on the network, and each is used as a data set. Each data set is given an identification symbol and location information. When searching or processing, a part or all of the information of the target data is stored in the same parameter using the same algorithm as the above algorithm. Identify the data set. Ideally, all data sets are deployed on the main storage device on a computer located on the network, but the data sets can be saved to other storage media (hereinafter referred to as checkout) and summoned. By having a function (hereinafter referred to as “check-in”), it supports the limitation of the number of connected computers.

    Further, as described in claim 2, by having a function of changing the location information of the data set identification symbol location information conversion unit of the data set, the arrangement of each data set at each time point is reviewed. Enable optimal relocation.

    Further, as described in claim 3, for the size of the arranged data set, the unused storage capacity of the main storage device of the electronic computer in which the data set is arranged is checked in the access target. If a checked out data set can be checked in without checking out the currently checked in data set, the current data set is not checked out and the accessed data set is Check in. This suppresses the use of the secondary storage device and improves performance.

    Further, as shown in claim 4, the CPU utilization rate of each data set and the electronic computer arranged on the network and the unused storage capacity of the installed main storage device are constantly or regularly monitored, Depending on changes in the actual size of the set at that time, the CPU utilization rate of the computers that are installed and usable, and the unused storage capacity of the installed main storage, the location information is automatically changed, By rearranging the set, the use of the secondary storage device is suppressed and the performance is improved.

    Also, as set forth in claim 5, when a data set is referenced from another tuple of its own data set, the data set referenced by the attribute data of the referring tuple By using the data as input data to the data set identification symbol conversion unit, the self part is recursively performed and the parent part described in the original large-scale data set before each data set is divided The BOM is forwardly expanded and reversely expanded based on the child component configuration table.

    As a result, extraction, sorting, merging, joining, and projection from a large-scale data set of several hundred terabyte order level can be executed at high speed.

    Further, even in a large-scale BOM of several tera number order level, it is possible to perform forward and reverse expansion at high speed.

    Even if a small database with a size of several gigabytes at the start of use grows over time and grows into a large database with a few hundred terabytes, the basic structure can be changed. Without increasing the number of computers arranged on the network, or improving the performance or increasing the number of CPUs mounted on individual computers arranged on the network, or arranged on the network. It is possible to construct the database while expanding the database to the optimum scale at that time according to the degree of growth only by increasing the capacity of the main storage device mounted on each individual computer.

    The best mode for carrying out the present invention will be described below with reference to the drawings. Note that the present invention is not limited to these embodiments, and can be implemented in various modes without departing from the scope of the present invention.

(Concept of invention)
FIG. 1 shows the concept of the invention according to claim 1.
Data input device 101,
Data storage specifying device 102,
Data storage devices 103, 104, 105,
Data search processing device 106,
The data output device 107 is
2 is a component of a multi-instance in-memory database according to the present invention.
Data set identification symbol converter 121,
The data set identification symbol location information conversion unit 122
It is a component of the data storage unit specifying device 102 that is a part of the components of the multi-instance in-memory database.
Data storage units 131, 141, 151,
Resource management unit 132, 142, 152,
Checkout / data saving spaces 133, 143, 153, 154
It is a component of the data storage devices 103, 104, 105 that are part of the components of the multi-instance in-memory database.
Data search unit 161,
Data processing unit 162,
Workspace part 163,
The resource management unit 164
It is a component of the data search processing device 106 that is a part of the components of the multi-instance in-memory database.

    The data input device 101 may import data from other electronic computers on the network.

    The data input device 101 imports data created on a removable storage medium such as a floppy disk, a CD, a DVD, or a memory disk by a network computer or a computer not connected to the network from the storage medium. May be.

    The data input device 101 may import data directly input from an input device such as a keyboard or a mouse connected to the computer on which the data input device 101 is mounted.

    The data set identification symbol conversion unit 121 reads the data imported by the data input device 101 for each tuple, and stores the tuple as input information using all or part of the information recorded in the attributes constituting the tuple. Convert to the identification (ie, ID) of the data set to be done.

    In order to convert the input information, which is all or part of the information recorded in the attribute constituting the tuple, into the identification symbol of the data set in which the tuple is to be stored, an appropriate parameter for an appropriate hash algorithm is used. You may carry out by giving a value. However, the hash algorithm and parameter values are not changed while the database system according to the present invention is operating.

    The data set identification symbol location information conversion unit 122 and the data set identification symbol to be stored in the data set identification symbol conversion unit 121 obtained by the data set identification symbol conversion unit 121 and the physical data It is a table which matches the location information on the network of the electronic computer arrange | positioned on the said network in which the data storage apparatus 103, 104, 105 in which a set is stored is mounted.

FIG. 2 shows an example of the configuration of a table that associates the identification symbol of the data set with the location information. In this example,
The identification symbol column 201 includes a location information column 211,
The identification symbol column 202 includes a location information column 212,
The identification symbol column 203 includes a location information column 213,
The identification symbol column 204 includes a location information column 214,
The identification symbol column 205 includes a location information column 215,
The identification symbol column 206 includes a location information column 216,
The identification symbol column 207 includes a location information column 217,
Are associated.
Therefore, the data set of the identification symbol 1 in the identification symbol column 201 means having a data storage device on the electronic computer on the network indicated by the location information 192.168.1.11 in the location information column 211. is doing. The identification symbols in the other identification symbol columns 202, 203, 204, 205, 206, and 207 are also associated with the location information in the other location information columns 212, 213, 214, 215, 216, and 217. A plurality of identification symbols may have the same location information. When the plurality of identification symbols have the same location information, a data management check-in or check-out is performed by a resource management unit that is a component of the data storage device as shown in claim 1. . However, as described in claim 3, this is not necessary when another data set requested to be accessed can be checked in without requiring the data set currently stored in the data storage unit to be checked out. is not.
When the example of FIG. 2 is applied to FIG. 1, the location information of the data storage device A 103 is “192.168.1.11”, and the location information of the data storage device B 104 is “192.168..11”. The location information of the data storage device C of “1.12” and 105 may be “192.168.1.13”.

    The resource management units 132, 142, and 152, which are part of the elements constituting the data storage apparatuses 103, 104, and 105 in FIG. 1, are part of the elements that constitute the corresponding data storage apparatuses 103, 104, and 105, respectively. Based on the unused main storage capacity of the data storage units 131, 141, and 151 and the size of the data set for which a new access request has occurred, the data set currently stored in the data storage unit is Evaluate whether or not the checkout / data saving spaces 133, 143, 153, and 154 that are part of the elements constituting the data storage devices 103, 104, and 105 need to be saved, and if necessary, save them. If the data set for which a new access request has occurred is a new data set, a new data set is created and checked out. To summon to check-in if the data sets that are saved in the chromatography data saved space 133,143,153,154. If the data set for which an access request has newly occurred is an existing data set by the function of the data set identifier conversion unit, the data to which the data set for which the access request has occurred belongs to It is guaranteed that it is saved in a checkout data saving space connected to the storage device.

    There may be a plurality of checkout / data saving spaces such as checkout / data saving spaces 153 and 154 shown in the data storage device C105 of FIG.

    Check if the original data set before being divided is sufficiently small, or the main storage capacity of the computer connected to the network is sufficiently large, or the number of computers connected to the network is sufficiently large All data sets may be stored on the main storage device without using a secondary storage device such as a hard disk device as an out data saving space.

    As shown in FIG. 1, the data search unit 161 constituting a part of the data search processing device 106 applies a search key after specifying a data set to be searched by the data storage unit specifying device 122. A data set in the data store is retrieved. If the specified data set exists in the data storage device 103, the data set is saved when it exists in the data storage unit 131 in the data storage device 103, and is saved when it exists in the checkout data save space 133. If it is necessary to check the data set into the data storage unit 131 for searching and processing, the data processing unit 162 in the data search processing unit 106 processes the search result as a processing target. The work space 163 in the data search processing device 106 is used as a work area by the data processing unit 162. Further, the resource management unit 164 in the data search processing device 106 may manage the capacity of the work space unit in which the required amount increases or decreases with the processing operation of the data processing unit 162.

    The data output device 107 may export the data to other electronic computers on the network.

    The data output device 107 creates a removable storage medium such as a floppy disk, a CD, a DVD, or a memory disk by using an electronic computer on the network or an electronic computer not connected to the network, and uses the storage medium to create data. You may export.

    The data output device 107 may directly export the data to an output device such as a CRT device or a printer device connected to the computer on which the data output device 107 is mounted.

    The data input device, the data storage unit specifying device, the data storage device, the data search processing device, and the data output device shown in claim 1 may be wholly or partially on the same electronic computer.

    In FIG. 2, the identification symbols are stored in the identification symbol columns 201, 202, 203, 204, 205, 206, and 207, and the location information is stored in the location information columns 211, 212, 213, 214, 215, 216, and 217. As shown in claim 2, it is possible to change the physical storage location of the data set by changing the contents described in the identification symbol field and the location information field. Show.

    FIG. 3 shows a state in which the data set 304 exists in the data storage unit 303 in the data storage device 301 and is assigned to the data storage device 301 and saved in the checkout / data saving space 305. This shows a case where a new access request is generated for the data set that has been stored and the resource management unit 302 determines that the data set 304 needs to be checked out to the checkout data save space 305, After the data set 304 is checked out to the checkout data save space 305, the data set for which the access request is newly generated from the checkout data save space 305 is checked into the data storage unit 303. FIG. This indicates that it is possible to assign a plurality of data sets to one data storage device as shown in claim 1.

    FIG. 4 shows that the data storage device 401 has a data set 404 currently in the data storage unit 403 and is assigned to the data storage device 401 and saved in the checkout data saving space 405. This shows a case where a new access request is generated for the data set that has been stored and the resource management unit 402 determines that the data set 404 does not need to be checked out to the checkout data save space 405. Without checking out the data set 404 to the checkout data save space 405, the data set for which the access request is newly generated from the checkout data save space 405 is transferred to the data storage unit 403. Check in as set 406 Diagrams are the cornerstone. This is the case when the data set to be newly checked in is sufficiently small and / or when the data storage is sufficiently large, or both, as shown in claim 3. This shows that the use of a secondary storage device can be suppressed, and a plurality of data sets can be assigned to the data storage unit of one data storage device and accessed at high speed.

    FIG. 5 shows that in the data storage device 501, the belonging data set list 503 and the data storage unit unused main storage capacity register 504 are arranged in the resource management unit 502, and the data A data set for which an access request is newly generated is generated by constantly monitoring increase / decrease in the data amount of the set, calculating an unused main storage capacity of the data storage unit, storing it in the unused main storage capacity register 504 of the data storage unit Is retrieved from the belonging data set list 503, and is compared with the data storage unit unused main storage capacity stored in the data storage unit unused main storage capacity register 504, and used main memory If the device capacity is larger than the data set to be checked in, the check Shows detail than FIG. 4 that it is possible to check in without the door.

    FIG. 6 shows that an electronic computer 601 and an electronic computer 611 arranged in a network 621 have data storage devices 602 and 612, respectively, and resource management units 603 and 613 inside the data storage devices 602 and 612, respectively. The resource management units 603 and 613 have identification symbols and data size correspondence tables 604 and 614 and allocated main storage capacity, respectively, and exchange information between the electronic computer 601 and the electronic computer 611. And automatically optimizing the relationship between the data size of the belonging data set and the capacity of the data storage unit main storage device, and relocating the belonging data set. Data set and checkout data saving space stored in 606,616 It indicates that the data set stored in 07,617 can be optimized relocated respectively. Here, two or more computers may be arranged on the network for exchanging information.

    FIG. 7 shows that in the data set identification symbol conversion unit 701, a top level (hereinafter referred to as L0) part number corresponding to a product number in a parts configuration table (hereinafter referred to as BOM) in the production management system is used as an input value. Of the data storage device of the electronic computer arranged on the network indicated by each location information, converted into location information by the data set identification symbol location information conversion unit 702 It shows that the data set of the parts configuration table divided in the storage units 703, 704, and 705 is arranged.

    FIG. 8 shows a BOM forward development table 801 using the parent parts of the parent part numbers 3, 16, and 18 as products from the data set of the divided parts configuration table shown in FIG. In FIG. 7, the product given the parent part number 3 obtains the identification number 0 as the remainder system of 3 by the data set identification symbol conversion unit 701. For the identification number 0, the data set identification symbol location information conversion unit 702 obtains location information 192.168.1.10. Part numbers 8, 10, 12 from a part of the parts configuration table in the data set stored in the data storage unit 703 in the data storage device on the electronic computer given the location information of 192.168.1.10. Get the child parts. Since the part having the part number 8 as the parent number is not included in a part of the part configuration table in the data set stored in the data storage unit 703, the part number 8 is input to the data set identification symbol conversion unit 701. The identification symbol 2 is obtained by giving it as a value. The location information 192.168.1.30 is obtained from the identification symbol 2 by the data set identification symbol location information conversion unit 702. Part numbers 11, 13, 20 from a part of the parts configuration table in the data set stored in the data storage unit 705 in the data storage device on the electronic computer given the location information of 192.168.1.30 Get the child parts. Since the data of the part configuration table having the part number 11 as the parent number exists in the data set stored in the data storage unit 705, the data is searched as it is to confirm that it is the terminal part having no child parts. Since the part having the part number 13 as the parent number does not exist in the data set stored in the data storage unit 705, the part number 13 is identified by giving the part number 13 again as the input value of the data set identification symbol conversion unit 701. The symbol 1 is obtained. The location information 192.168.1.20 is obtained from the identification symbol 1 by the data set identification symbol location information conversion unit 702. A terminal that has no child parts from a part of the parts configuration table in the data set stored in the data storage unit 704 in the data storage device on the electronic computer given the location information of 192.168.1.20 Confirm that it is a part. Since the data of the part configuration table having the part number 20 as the parent number exists in the data set stored in the data storage unit 705, the data is searched as it is and the part number 17 is obtained as the child part. Since the data of the part configuration table having the part number 17 as the parent number exists in the data set stored in the data storage unit 705, the data is searched as it is to confirm that it is the terminal part having no child parts. The same process is performed for parts having part numbers 10 and 12, and as shown in claim 5, the value of level 0 (hereinafter referred to as L0) in the BOM regular expansion table 801 shown in FIG. 3, that is, it is possible to obtain a BOM positive development table of a product having a part number 3 at the top level.

    The data storage units 703, 704, and 705 in FIG. 7 may be replaced with commercially available, shareware, freeware, or originally developed in-memory database engines. .

    In order to strengthen EU environmental problem regulations, traceability for proof of non-use of hazardous chemical substances such as industrial products and agricultural products, and notification and collection of messages such as prohibition of use to consumers when used incorrectly. In order to secure it, there is a demand for practical use of a large-capacity ultra high-speed database.

    With the present invention, it is possible to connect multiple PC servers equipped with an in-memory database, which is an ultra-high-speed database engine, via a network, and to scale up and scale up as grid computing by linking small databases. In the meantime, the EU is trying to cope with tightening regulations on environmental issues in the EU.

Overall configuration diagram Overall configuration diagram according to claim 1 of the present invention Example of Data Set Identification Symbol Location Information Conversion Unit Example of Data Set Identification Symbol Location Information Conversion Unit Based on Claims 1 and 2 of the Present Invention Conceptual diagram of check-out and check-in of data set Conceptual diagram showing that a plurality of data sets can be assigned to one data storage device according to claim 1 of the present invention. Check-in conceptual diagram without data set check-out Conceptual diagram showing that it is possible to check-in without checking out, according to claim 3 of the present invention. Detailed conceptual diagram of check-in without check-out of data set Detailed explanatory diagram showing in detail that it is possible to check-in without check-out according to claim 3 of the present invention Configuration diagram of automatic optimization mechanism of storage location of data set Based on claim 4 of the present invention, automatically optimizes the relationship between the data size of the affiliated data set and the capacity of the data storage main storage device, Configuration diagram showing optimization by relocation Cross-sectional integration self-join mechanism explanatory diagram of multiple data sets after splitting FIG. 7 is a table showing a result of simulating the normal development of the divided parts configuration table based on claim 5 of the present invention.

Claims (5)

Placed on the network,
One or more data input devices for inputting external data;
One or more data storage unit specifying devices for specifying the storage destination of the input data;
One or more data storage devices for storing the input data on a main storage device;
One or more data search processing devices for searching and processing the stored data;
One or more data output devices for outputting the searched and processed data to the outside;
A multi-instance in-memory database system comprising:
The data storage unit specifying device includes:
Using part or all of the information of the input data as a data set identification symbol for specifying the data set where the input data is stored, using an algorithm such as hashing with specific parameters A data set identifier conversion unit for the input data to be converted;
A data set identification symbol location information conversion unit of the data set that associates data set location information specifying the location of the data set corresponding to the identification symbol and the data set identification symbol;
Have
The data storage device includes:
In the data storage device specified by the data storage unit location information, a data storage unit for storing the input data in a data set existing on a main storage device on the data storage device;
A secondary storage device or network that currently has the same data set location information as the data set present on the main storage device on the data storage device and is currently connected to the data storage device When an access request such as insertion or extraction occurs in another data set saved on the main storage device or the secondary storage device on the other electronic computer above, the current data storage device The data set residing on the main storage device is saved on a secondary storage device connected to the data storage device, on a main storage device on another electronic computer on the network, or on a secondary storage device. On a secondary storage device currently connected to the data storage device or on a main storage device or a secondary storage device on another electronic computer on the network And resource management portion in which the access request has been avoided to summon a data set is generated,
Have
The data search processing device includes:
Data search in which one or more data sets on one or more data storage devices are traversed while specifying the data set to be searched using the same algorithm with the same parameters as the algorithm And
The data specified and extracted by the data search unit is changed or deleted while specifying the data set to be processed using the same algorithm as the algorithm with the same parameters, or one or more of the data A data processing unit that performs data processing across data sets such as joins, projections, sorts, and merges of data groups across sets;
A work space part which is a main storage space provided for the processing of the data search part or the data processing part;
When resources of the main storage space provided to the data storage unit, the data search unit, the data processing unit, and the work space unit are insufficient, a secondary storage device such as a hard disk device or another electronic computer on the network A resource management unit that performs resource management by using resources of main storage devices and secondary storage devices of
Having
Multi-instance in-memory database system. The data storage unit specifying device has a function of changing the location information of the data set identification symbol location information conversion unit of the data set, the main storage capacity of the data storage device and the size of the stored data set The multi-instance in-memory database system according to claim 1, wherein the relation can be optimized and rearranged. The data storage device is attempting to newly create another data set having the same location information as the data set currently stored on the main storage device on the data storage device, or When an access request is made to a data set saved on a secondary storage device connected to the data storage device, on a main storage device on another electronic computer on the network, or on a secondary storage device The data set currently stored on the main storage device on the data storage device is stored on the secondary storage device currently connected to the data storage device or on the other electronic computer on the network. There is enough main storage capacity in the data storage device to create or recall without saving on the device or secondary storage device In this case, one or more data sets can be stored on the main storage device on the data storage device without saving the data set currently stored on the main storage device on the data storage device. The multi-instance in-memory database system of claim 1, which is possible. Monitor the capacity of each main storage device and the size of each data set on one or more data storage devices placed on the network so that the number of evacuation or checkout, summoning or checkin is minimized. 4. The multi-instance in-memory database system according to claim 1, wherein the location information is automatically rearranged. When the data set is referenced from another data set or another tuple of its own data set, the data set of the data set referenced by the attribute data of the referring tuple 4. The join result can be created as a new data set by using the input data to the set identification symbol conversion unit. 5. The multi-instance in-memory database system according to 4.

Images