JP3073889B2 - Data transfer method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a high-speed data transfer method from a computer system to a database device and a method for estimating a data transfer amount in the method.

[0002]

[Prior Art] "Information Processing" published by Information Processing Society of Japan
December, 1992 (Vol. 33, No. 12) pp. 1388
-1440 describes a database system provided with a conventional database device. FIG.
This is a data transfer method from a computer system to a database device in a database system having a conventional database device introduced on pages 1416-1423. In the figure, 1 is a computer system, 2 is all data of the computer system 1, 4 is a database device connected to the computer system 1, 5 is data to be finally selected from the data 2, and 8 is a database. Data 15 finally selected via the device 4 is data to be transferred.

In the conventional data transfer method, in order to reduce the computer load on the computer system 1, all the data 2 included in the file in the selection process is transferred from the computer system 1 to the database device 4 via the connection device. Was.
The operation of selecting the data 8 satisfying the selection condition has been performed by the database device 4 after the above data transfer operation.

The above-mentioned literature describes a method for saving the time required for data transfer from the computer system 1 to the database device 4 in database processing involving data selection, particularly a method for saving data transfer time using an index. Not.

[0005]

In the conventional data transfer method as described above, when the amount of data that satisfies the selection condition is extremely small, the data transfer time from the computer system to the database device is reduced by the user inputting the data selection condition. After that, there was a process that took up most of the time until the search results were obtained. A database user who performs such processing on a large scale on a daily basis has been demanded to reduce the data transfer time from the computer system to the database device.

In addition, the ratio of the data transfer time from the computer system to the database device to the time required for the database processing increases as the number of selected data items increases. As the use of databases has spread in recent years, the amount of data stored in various databases has exploded, and the user wants to reduce the time from entering data selection conditions to obtaining search results. The request was widespread.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-described problems, and reduces the amount of data transferred from a computer system to a database device, and from the time when a user inputs data selection conditions to when a search result is obtained. The purpose is to reduce the time of

[0008]

According to a first aspect a method of data transfer in a computer system and database system by connecting the database apparatus in connection device, if the database processing including data selection condition is given, for the data selection condition It is determined whether an index can be used. If the index can be used, a data block group including all data satisfying the data selection condition in the computer system is obtained using the index, and the data block group is transferred from the computer system to the database device. Is what you do.

[0009] The second data transfer Okukata method, in a computer system and database system by connecting the database apparatus in connection device, determine the number of entries in the root block of the index used to determine the selection criteria data, The number of data that satisfies the selection condition is estimated by the number obtained by multiplying this number by the number of entries included in one block of the index by the number obtained by subtracting 1 from the number of levels in the index.

A third data transfer Okukata method, computer in the system and the connected database system database device connection device, a 1 under the level of entry than the root block of the index used to determine the selection criteria data The number is obtained, and the number of data satisfying the selection condition is estimated by the number obtained by multiplying the number by the number of entries included in one block of the index by the number obtained by subtracting 2 from the number of levels of the index.

[0011] The fourth data transfer Okukata method, in a computer system and the connected database system database device connection device, is used to examine the selection criteria data included in one level below the root block of the index The number of sub-blocks having the head entry is obtained, this number is multiplied by the number of entries included in one sub-block, and the number is further reduced by 2 from the number of levels in the index. The number of data that satisfies the selection condition is estimated by the number multiplied by the number of subtractions.

[0012] The fifth method of data transfer in the first data transfer method, the second or third or fourth data transfer
It estimates the number of selected data that satisfies the Okukata method predicts the output time of the block of level 0 of the index portion required for the processing to determine the data to be transferred to the address of the data from the computer system retrieves the database device If the former is longer than the latter, as compared with the execution time of the input / output command required to transfer all the data from the computer system to the database device, the high-speed processing is stopped.

[0013] The sixth method of data transfer in the first data transfer method, the position of the data block group containing all the data satisfy the conditions were selected for indexed in a computer system using the index obtaining process in the middle The number of I / O instructions used to read the index in this process
The predicted value of the fifth data transfer method is updated, and compared with the execution time of the input / output instruction required for the process of transferring all data from the computer system to the database device, if the former is longer than the latter, the speed is increased. Is stopped.

[0014] Data transfer method seventh in a computer system and database system by connecting the database apparatus in connection device, logical operators (and, or, no
Given a database processing that includes the data selection conditions linked in t), if an index is available for the conditions, then the logical operator of the data selection conditions
Or a combination of two consecutive unit elements in the data selection condition, with a separator of the applicable range of the logical operator, a conditioner not including the logical operator, or a terminal symbol of the condition group as one unit element. A condition group that needs to calculate a condition group affected by at least a certain number of the logical operators in order to clarify the applicable range of the logical operator based on the classified information and reduce the data transfer amount. Is characterized in that the processing is stopped.

The data transfer method of the eighth, the computer system and the connected database system database device connection device, if the database processing including tied data selection condition by logical operators are given for the condition index Can be used, the indexed ones of the data selection conditions are combined, and a set of conditions that includes all the data that satisfies the given selection conditions of all data is selected, and an indexed selection in the computer system is selected. The maximum value and the minimum value of the address of a continuous data block group including all data satisfying the above conditions are obtained by using an index, and the block group sandwiched by blocks having this address is transferred from the computer system to the database device. It is characterized by.

The ninth method of data transfer in a computer system and the connected database system database device connection device, if the database processing including tied data selection condition by logical operators are given for the condition index Can be used, the indexed ones of the data selection conditions are combined, and a set of conditions including all data satisfying the selection conditions of all given data is selected, and a block in the block in the computer system is selected. With respect to a data block having additional information for knowing the maximum value and the minimum value of the key value of the data, all data blocks including all data satisfying the selected condition with an index are obtained using the index and the additional information. Transferring the block group from the computer system to the database device. And butterflies.

[0017]

[0018]

In the first data transfer method, when a database process including a data selection condition is given, it is determined whether or not an index can be used for this condition. If an index can be used, the condition in the computer system is satisfied. A data block group including all data is obtained using an index, the data block group is transferred from the computer system to the database device, and the data transfer amount is reduced compared to the case where all data is transferred from the computer system to the database device. Speed up processing.

In the second data transfer method, the number of entries in the root block of the index used to check the data satisfying the selection condition is obtained, and the number of entries included in one block of the index is added to this number. The number of data that satisfies the selection condition is estimated by the number obtained by multiplying by 1 the number of times.

In the third data transfer method, the number of entries one level lower than the root block of the index used for checking data satisfying the selection condition is obtained, and the number of entries included in one block of the index is calculated. The number of data that satisfies the selection condition is estimated by the number obtained by multiplying the number of index levels by two.

In the fourth data transfer method, the number of sub-blocks having a head entry used for checking data satisfying the selection condition one level lower than the root block of the index is obtained, and this number is used as the sub-block 1 Is multiplied by the number of entries included in the index and the number is multiplied by the number of entries included in one block of the index by the number obtained by subtracting 2 from the number of levels of the index to estimate the number of data satisfying the selection condition. .

In the fifth data transfer method, the number of data satisfying the selection condition is estimated, the address of this data is taken out, and the level 0 block of the index unit necessary for determining the data to be transferred from the computer system to the database device. If the former is longer than the latter, it is compared with the execution time of the input / output command required for the transfer of all data from the computer system to the database device, and the speed-up process is stopped.

In the sixth data transfer method, the index is read by this process during the process of using the index to find the position of a data block group including all data satisfying the selected condition with an index in the computer system. If the former is longer than the latter, the predicted value is updated by calculating the number of I / O instructions used for the command, and compared with the execution time of the I / O instructions required to transfer all data from the computer system to the database device. Stops the speed-up process.

The seventh data transfer method clarifies the application range of the logical operator based on information that classifies the arrangement of combinations of two consecutive unit elements in the data selection condition,
Processing is stopped for a condition group that needs to calculate a condition group affected by a certain number or more of logical operators in order to reduce the data transfer amount.

The data transfer method of the eighth, the computer maximum address of data block set and minimum contiguous contains all the data satisfy the conditions were selected for indexed in the system be used to calculate the index, the address The block group sandwiched between the blocks having the block is transferred from the computer system to the database device.

The ninth method of data transfer, the maximum value of key values of data in blocks in the computer system, the data block with the additional information to know the minimum value, all the data satisfy the conditions were selected for indexed All data blocks are obtained using the index and the additional information, and this data block group is transferred from the computer system to the database device.

[0027]

[0028]

【Example】

Embodiment 1 FIG. First Embodiment A first embodiment of the present invention will be described with reference to FIGS. FIG. 1 is a diagram in which the speed of a database system including a database device is increased. In the figure, 1 is a computer system, 2 is a data part of the computer system 1, 3
Is an index unit provided by the computer system 1, and 4 is a database device. Reference numeral 5 denotes data to be finally selected, which satisfies all the data selection conditions given to the database system. 6 is data that satisfies the selection condition, 7 is data that satisfies the selection condition that can use the index, 8 is data that is finally selected via the database device 4, 9 is data that is transferred from the computer system 1 to the database device 4. The flow 10 is data satisfying the selection conditions selected by the database device 4, 11 is individual data (data record), 12 is a data block, 13 is an index entry, and 14 is an index block (particularly a root block). ), 16 is an index group entry group, 17 is a data group, 1
8 is a data block group, and 19 is a data group.

In the conventional processing shown in FIG. 8, after transferring all data 2 from the computer system 1 to the database device 4,
The database device 4 has selected data 5 that satisfies the selection conditions. On the other hand, in the database of FIG. 1 realized by using the present invention, only the data 6 that satisfies the selection condition for using the index is sent from the computer system 1 to the database system. The difference between the data transfer amount between the database system using the present invention and the conventional system indicated by the data 6 satisfying the selection condition and the data 15 transferred, respectively, is as follows.
The time required for data transfer is saved, and the speed of database processing by the database system using the present invention is increased. FIG. 5 specifically shows the structures of the data section 2 and the index section 3. The data section 2 is a set of individual data 11. The individual data 11 is grouped by a constant number (three in FIG. 5) to form a data block 12. The ordinary computer system 1 uses the data block 12 as an input / output unit. Hereinafter, even in the database system of this example, it is assumed that data transfer from the computer system 1 to the database device 4 is performed in units of the data blocks 12. The index unit 3 includes an entry 13 that holds a value (key value) of a specific location in each data 11, and a plurality of entries 13 (three in this example) are collected to form a block 14 (a block of the index unit). Form.
In particular, the block 14 at the top of the index unit 3 is called the root block of the index unit 3. The entire index part 3 is shown in FIG.
As shown in the figure, the blocks have a tree structure. The horizontal arrangement of the index section 3 is called a level, a level near the root block is called an upper level, a level away from the root block, and a level near the data section is called a lower level. As shown in FIG. 5, the entries 13 in the same level of the index unit 3 are arranged in ascending order (or descending order) of the held key values.
In the example of FIG. 5, a data group having a value from 1 to 100 as a key value, and indexes for these data groups are shown. The entry at the lowest level (referred to as level 0) of the index portion has a one-to-one correspondence with each piece of data, and holds a pointer to each piece of data and a key value of the corresponding data. Individual entries other than level 0 are
It has a one-to-one correspondence with the block of the index part at the next lower level, and the maximum value of the key value of the entry in the corresponding block (the minimum value if the entries are arranged in descending order of the key value) , And a pointer to the corresponding lower-level block.

When a value of data is specified and extracted from the data portion, any data can be extracted at high speed by using an index according to the following procedure. The index section 3 is sequentially traced from the root block 14, the key value of the entry in the index section 3 is compared with the value of the data to be searched, the first entry having a key value equal to or greater than the value of the search data is found, and the pointer is traced. To retrieve the next block until the search data is reached. When a data value range is given as a data selection condition, a level 0 entry of the index unit 3 having a key value corresponding to the condition value range is searched, and data having a value within the given range is searched. The pointer to the group can be retrieved. To search for this level 0 entry, the index unit 3 is added to the root block 14.
In the blocks other than level 0, the first entry having a key value equal to or greater than the minimum value of the condition and the key value greater than the maximum value (or the last entry if the entries are arranged in descending order of the key value) )
The pointer to the lower level held by the entry between them is traced, and the block at the lower level indicated by the pointer is fetched until the block at level 0 is fetched, and the search for entries in the block is repeated. The present invention makes use of this fact when an index can be used for data selection conditions.
The data to be transferred from the computer system 1 to the database device 4 is reduced to only those satisfying the selection condition for using the index, thereby saving the data transfer amount.

For example, in FIG. 5, if "data having key values of 80 to 90" is given as a selection condition, the entry group 16 in the figure searches for a level 0 entry having a key value satisfying the condition. Is a group of entries of the index unit 3. This corresponds to the portion indicated by the data 7 in FIG. 1 that satisfies the selection condition for using the index. This entry group 1
6 are examined sequentially from the root block 14 in accordance with the above-described procedure, and a level 0 entry having a key value of 80 to 90 is searched. The data group 17 corresponds to these entries. This corresponds to the data group 5 shown in FIG. The data block group 18 or the data block group 18 from which the data group 19 has been removed is transferred from the computer system 1 to the database device. This corresponds to the portion indicated by data 6 that satisfies the selection condition in FIG. The search for data satisfying the selection condition is performed when the level 0 entry of the index unit 3 has a pointer to the next entry or when the position of the next entry can be easily obtained. , A level 0 entry corresponding to the minimum value, and checking the level 0 entry between the two. FIG.
Applying this method to the example of the above, the entry at the end of each level can be determined without tracing the entire entry group 16 (the entry having the key value of 90 which is one level higher than the entry group of the level 0 is excluded). Is checked, a level 0 entry group corresponding to the data group 17 can be extracted.

Next, the operation of this embodiment will be described with reference to the flowchart shown in FIG. When a data selection condition is given from the user to the database system of this embodiment, it is checked whether an index can be used for the selection condition (step 101). If the index can be used, a level 0 entry of the index portion corresponding to the minimum value of the selection condition is searched (step 102) (corresponding to the leftmost entry at level 0 in the entry group 16 in FIG. 5), and (Step 103) (corresponding to the address of the leftmost data in the data group 17 in FIG. 5). A group of entries of level 0 of the index unit 3 is sequentially extracted in ascending order of the key value (step 10).
5) The addresses of the data corresponding to these entries are extracted (step 103). In FIG. 5, this corresponds to examining the entries at level 0 in the entry group 16 one by one from left to right, and extracting the address of the data group 17. At this time, if an entry of level 0 of the index part corresponding to the maximum value of the selection condition is found (step 1)
04), a data block including all of these data is sent from the computer system 1 to the database device 4 from the addresses of all of the extracted data (step 10).
6) End the processing (the order of checking the entry at level 0 may be the descending order of the key value from the maximum value). At this time, for example, the maximum value and the minimum value of the address of a continuous data block group including all data satisfying the conditions are obtained by using an index, and the block group sandwiched by the block having this address is obtained from the computer system 1 in the database. Device 4
In the block group 18 of FIG. 5, the block 19 that does not include data that satisfies the condition is also transferred.
This is because, depending on the design of the database system, there is no guarantee that the data belonging to the data group 17 is always contained in a single block. Therefore, it is generally necessary to transfer the block containing the data to the database device 4 as a continuous area. Because there is.

As described above, according to this embodiment, the database processing including one data selection condition that can use an index is reduced by reducing the amount of data to be transferred from the computer system to the database device as compared with the prior art. This has the effect of speeding up.

Embodiment 2 FIG. In the database of this embodiment,
When a plurality of selection conditions linked by a logical operator are given by the user, the database of this embodiment performs a structural analysis of the selection conditions shown in the flowchart of FIG. To explain this, the logical structure of a data selection condition group will be described with reference to FIGS. The condition group for data selection consists of conditions on the values of individual data and logical operators connecting them, and the logical structure analysis of the condition group clarifies the scope of each of these logical operators, To extract the outer logical operator.
Here, the logical operator is an operator represented by and, or, not. When examining the outermost logical operator, only the and, or is targeted, and is expressed by a logical expression including not. By inverting individual conditions using de Morgan's law, the logical operator not is previously removed from the logical expression. The “outermost logical operator (outermost operator)” that connects the condition groups is a single object representing the content of the parentheses by enclosing the applicable range of the logical operators of the same type that are arranged in pairs. (A single condition without a logical operator, or a group of conditions connected by a single logical operator)
It is defined as the operator that remains in the finally obtained conditional expression. For example, in the following example, assuming that one uppercase alphabetic character from A to F is a condition for selecting data that does not include a logical operator in itself, the outermost operator of the condition AandB is and. Condition Aand (BorC) a
The outermost operator of ndD is also and. Condition ((Aor
B) The outermost operator of and (CorD)) or (EndF) is or.

The outermost logical operator connecting the condition groups is a
The policy for selecting the condition that can be used for reducing the data transfer amount among the conditions that can use the index is changed depending on whether the condition is nd or or. That is, in the case of the condition AandB, even when the index can be used for only one of the conditions A and B,
Although the present invention can be implemented, in the case of the condition A or B, the processing of the present invention can be executed only when the index can be used for both the conditions A and B.

The analysis in the case where the outermost logical operator is "and" will be described with reference to FIG. 6 by taking as an example a process of selecting data satisfying the condition "AandB". In FIG. 6, 2
Numerals 0 and 21 indicate the same computer system, but are divided into two for convenience of explanation. 22 is a computer system 2
Data groups satisfying the condition A, 0 and 21 are data groups satisfying the condition B, and 24 is a data group satisfying the condition AandB. As shown in FIG.
Is included as a part of each of the data groups 22 and 23, even if the index can be used for only one of the conditions A and B, the computer system 1 according to the present invention
These indexes can be used in a process of reducing the amount of data transferred from the database device 4 to the database device 4. When an index can be used for both conditions A and B, a process for reducing the data transfer amount can be executed using both of them.

The analysis in the case where the outermost logical operator is or will be described with reference to FIG. 7 by taking as an example a process of selecting data satisfying the condition of AorB. In FIG. 7, 25,
26, the same computer system is divided into two for convenience of explanation, as in the case of and. 27 is a data group which satisfies the condition A in the computer systems 25 and 26, 28 is a data group which satisfies the condition B, and 29 is Ao
This is a data group satisfying rB. As shown in FIG. 7, the data group 29 is not included as a part of any of the computer systems 25 and 26, and the entire data group 29 can be obtained only after evaluating both the conditions A and B. Therefore, only when an index can be used for both the conditions A and B, the processing for reducing the amount of data transferred from the computer system to the database device according to the present invention can be executed.

Returning to FIG. 3, which shows the structural analysis of the selection conditions performed by the database when a plurality of selection conditions are given by the user to the database of this embodiment. First, it is checked whether an index can be used among the given selection conditions (step 201). If there is a selection condition for which an index can be used, the logical structure of the condition group is analyzed as described above, and it is checked whether the outermost logical operator connecting the condition group is and or or (step 20).
2).

Using the results of the analysis in step 202, the conditions in which indexes can be used, which are used in the processing in steps 203 to 211 to reduce the amount of data transferred from the computer system to the database device, are selected. The data blocks to be transferred from the computer system 1 to the database device 4 in the processing shown in FIG.
Actually transfer. First, as a result of the analysis in step 202,
When the outermost operator is and, this and
It is checked whether there is a condition directly usable for the index (a condition that the index can be used without including a logical operator by itself) (step 203). If any, the processing for reducing the amount of data transfer from the computer system 1 to the database device 4 according to the present invention can be executed. Choose to use for (Step 2
04). The method of selecting the condition in step 204 depends on the policy of designing each database system using the present invention.

If the condition group connected to the outermost and is a condition that includes all logical operators or a condition that cannot use an index that does not include a logical operator, step 20 is executed.
Go to 5. When the condition group connected to the outermost and that includes a logical operator by itself can be selected recursively by applying the condition analysis of FIG. 3 to a condition group that can be used in the process of reducing the data transfer amount according to the present invention. (Step 2
Even if YES in 03, the processing of steps 205 to 211 can be applied). Whether this recursive application is performed depends on the design policy of each database system using the present invention.

When the outermost operator is or as a result of the analysis in step 202, each of the condition groups directly connected to this or is extracted (step 20).
6, 207), it is checked whether an index can be used alone for all of them (step 208).

In step 208, a condition including a logical operator in any one of the condition groups connected to the outermost or
Alternatively, if there is a condition that does not include a logical operator but cannot use an index, the processing cannot be accelerated using the present invention, so the processing is stopped. However, when this condition includes a logical operator, the processing may be speeded up by recursively applying the structural analysis of FIG. 3 without stopping the processing. Whether a process is stopped or a structural analysis is applied recursively when a condition including a logical operator is found in step 208 depends on the design policy of each database system implemented using the present invention.

The outermost operator is and, step 2
04, when one or more conditions that can use the index alone are selected, or when the outermost logical operator is or,
When it is determined in step 10 that the condition group connected to the or is a condition in which all of the indexes can be used independently, a process of determining a data block to be transferred from the computer system to the database device shown in FIG. 4 is performed. Note that, depending on the database design policy, these conditions may include conditions under which the index can be used by applying the structural analysis of FIG. 3 recursively. This processing will be described. Each condition selected using the structural analysis of FIG. 3 is extracted (step 301),
A search is made for a level 0 entry of the index part corresponding to the minimum value of the data under each condition (step 302), and the address of the data corresponding to this entry is extracted (step 303). The level 0 entries of the index part are sequentially extracted in ascending order of the key value (step 305), and the addresses of data corresponding to these entries are extracted (step 303). At this time, when a level 0 entry of the index portion corresponding to the maximum value of the selection condition is found (steps 304 and 305), a data block including all of these data is read from the computer system based on the addresses of all of the extracted data. The data is sent to the database device (step 306), and the processing is terminated (similar to the case where the number of conditions is one, the order of checking the level 0 entry may be the descending order of the key value from the maximum value).

After obtaining the addresses of the data groups satisfying the respective conditions by performing the above processing, a logical operation is performed by applying a logical operator connecting the conditions, and the data is actually transferred from the computer system to the database device. The address of the block group is determined (step 307). For example, there are conditions A and B for data values, data that satisfies condition A is present in blocks from address 1 to address 1,000, and data that satisfies condition B is from address 100 to 10,000.
When it exists in the block up to the address 00, when the selection condition of AandB is given, 100
If a selection condition of AorB is given from the address to the address 1,000, the data part block from the address 1 to the address 10,000 is transferred. Conditions for each index (A, B in the above example)
), The address range of the data block containing the data satisfying the following (addresses 1 to 1,000 for A, B
After obtaining the addresses 100 to 10,000 of the above, a logical operator (here, an
Based on (d or or), the address of the block of the data section that includes data that satisfies the conditions of the actually given conditional expression is calculated. This number corresponds to the block group 18 in the example of FIG.

In the logical structure analysis of the condition group in FIG. 3, for the condition where the logical structure analysis is applied recursively, the above-described logical operation for obtaining the block number is also performed recursively (the logical Operator) to determine the addresses of the blocks to be transferred. The data blocks including the blocks obtained in step 307 are actually transferred from the computer system to the database device, and the process ends.

As described above, according to this embodiment, a plurality of data selection conditions combined by a logical operator are analyzed in a logical structure, a necessary set of conditions is selected, and a database system is selected from the computer system more than in the prior art. There is an effect that database processing is speeded up by reducing the amount of data transferred to the device.

Embodiment 3 FIG. In the first or second embodiment, the number of data to be transferred when the process of reducing the data to be transferred from the computer system 1 to the database device 4 is completed until the end of the process, or the data including these data. The number of blocks can be estimated.

In the index part of the database system shown in FIG.
The entry used for checking data that satisfies the selection condition of “data having a key value of 90” is one at the right end holding a key value of 100 (generally, level 0 in the index portion).
The entries used to check data satisfying the selection condition at each level other than are the entries between the first entry having a key value equal to or greater than the minimum value and the key value greater than the maximum value. In the example of FIG. 5, the same entry holding the key value of 100 is the first entry having the two key values. If the entries are arranged in descending order of the key value, the entry used to check the data that satisfies the selection condition must be between the first entry with the key value equal to or less than the maximum value and the key value smaller than the minimum value. It becomes an entry. In the following description, the order of the entries in the index is in ascending order of the key value.)

In the index section 3 shown in the example of FIG. 5, when the pointers pointing to the entries belonging to the entry 16 are traced in the direction of the upper level (from the block of level 0 to the direction of the root block), all the right ends of the root block 14 are obtained. To the entry at.

Conversely, starting from the entries of the root block 14 of the index that are used to check the data satisfying the selection condition (for example, the rightmost entry of the root block 14 in FIG. 5), the lower level included in these entries By repeatedly tracing the pointer to the lower level contained in the entry in the block pointed to by the pointer to the lower level, data such as entry 16 in FIG. Through all entries that are In particular, the finally reached level 0 entry group always includes all entries corresponding to data satisfying the selection condition.

As a result, the entries belonging to the root block 14 of the index unit 3 and having the key value equal to or greater than the minimum value and having the key value greater than the maximum value are respectively located from the first entry group to the lower level. Can be used as an estimated value of the number of data transferred by the processing according to the present invention.

In the example of FIG. 5, the number of entries 16 belonging to the root block 14
Is used to estimate the number of entry groups belonging to level 0.

Therefore, in the first embodiment or the second embodiment, after the logical structure analysis of the conditions shown in FIG. 3 between the steps 101 and 102 in FIG. Before starting the process, the number of data transferred from the computer system to the database device when the present invention is implemented can be estimated.

Each level entry is one level lower.
One-to-one correspondence with blocks. Therefore, if it is determined how many entries are included in one block of the index part, one level lower than the number of entries used to check data having a key value satisfying the selection condition among the entries belonging to the root block. The number of entries can be obtained. By using this number and the total number of levels of the index, the estimated value of the number of data is calculated.

More specifically, the number of levels of the index is obtained by multiplying the number of entries included in the root block 14 of the index and used for checking data satisfying the selection condition by the number of entries included in one block of the index. Is greater than or equal to 2, the estimated number of data that satisfies each condition is calculated by multiplying this number by the number of times obtained by subtracting 1 from the total number of index levels.

That is, the estimated number of data can be obtained by Equation 1.

[0057]

(Equation 1)

In the example of FIG. 5, the number of entries used for checking data satisfying the selection condition included in the root block of the index is 1, the number of entries per block of the index is 3, and the number of levels of the index is 3 Therefore, the estimated number of data is 1 × 3 ^3-1 = 9 (pieces).

In a database system, since the unit of data input / output is often a data block, the calculation when the number of index levels is 2 or more is multiplied by the value obtained by subtracting 2 from the total number of index levels. By revising to match, it is possible to calculate the estimated number of data blocks including data satisfying each condition.

That is, the estimated number of data blocks can be obtained by Equation 2.

[0061]

(Equation 2)

In the example of FIG. 5, since the number of index levels is 3, the estimated number of data blocks is 1 × 3 ³⁻² = 3.
(Pieces).

In a general database system,
Since data is deleted / updated and the entries in the index part are also deleted / updated accordingly, the number of entries included in one block of the index part is not constant. In such a case, for example, the initial value of the number of entries included in one block of the index part can be used as the number of entries included in one block of the index part in a general case.

As described above, according to this embodiment, based on the data selection conditions and the contents of the root block of the index,
In the initial stage of processing for realizing the present invention, there is an effect that the amount of data transferred from the computer system to the database device by this processing is estimated, and the effect of speeding up is predicted.

Embodiment 4 FIG. In the third embodiment,
Instead of the number of entries belonging to the root block, the entries between the first entry having a key value equal to or more than the minimum value of the selection condition and the key value exceeding the maximum value among the entries in the index part one level lower than the root block The number of data can be estimated more accurately by using the number of groups and using a value obtained by subtracting 2 from the total number of levels of the index part instead of the total number of levels of the index part.

That is, the estimated number in this case is obtained by Expression 3.

[0067]

(Equation 3)

In the example of FIG. 5, the number of entries used to check data satisfying the selection condition included in the level one level lower than the root block of the index is 3, the number of entries per block of the index is 3, Since the number of index levels is 3, the estimated number of data is 3 × 3 ^3-2 = 9 (pieces). In this example, among the 16 entry groups, those belonging to the level one level lower than the root block of the index are included in the same block, so the estimated value of the number of data is the level lower than the root block in the third embodiment. Is the same as the case where the number of entries is approximated by (the number of blocks) × (the number of entries per block). However, in general, the first entry having a key value equal to or greater than the minimum value of the selection condition and a key value greater than the maximum value are not necessarily at the end of the block. (Number of blocks) x (block 1
(The number of entries per piece). In particular, when the number of entries per index block increases,
Since there is a possibility that the number of entries satisfying the selection condition can be considerably reduced just by being one level lower than the root block,
The estimated value is more accurate than in the third embodiment.

As described above, according to this embodiment, the contents of the entry of the index part one level lower than the root block are used in place of the contents of the root block in the third embodiment, so that the computer system can convert the database. There is an effect that the number of data transferred to the device is more accurately estimated than in the third embodiment.

Embodiment 5 FIG. In the fourth embodiment,
When the number of levels of the index is two or more and each block of the index part is divided into sub-blocks (or groups and sections) for every fixed number of entries, one level below the root block of the index corresponding to each selected condition In the entry, the number of sub-blocks having the head entry used for checking data satisfying the selection condition is obtained, and the number is multiplied by the number of entries included in one sub-block in the index, thereby obtaining:
When the number of index levels is 3 or more, each condition is further multiplied by the number obtained by multiplying the number of entries contained in one block of the index by 2 times the number of entries in the index and the number of times. The number of data that satisfies can be estimated. The estimated value is more accurate in the fourth embodiment than in the present embodiment, but the load for calculating the estimate is smaller in the present embodiment than in the fourth embodiment.

That is, the estimated number of data can be obtained by Expression 4.

[0072]

(Equation 4)

As described above, according to this embodiment, when the number of index levels is two or more and each block of the index part is divided into sub-blocks (or groups and sections) for every fixed number of entries, Instead of the contents of the root block in the third embodiment, the number of data transferred from the computer system to the database device is more accurate than in the third embodiment, using the contents of the sub-blocks included in the block one layer below the root of the index. In addition, there is an effect that the calculation load is estimated to be smaller than that of the fourth embodiment.

Embodiment 6 FIG. The execution time of the input / output instruction in the first or second embodiment is predicted in the middle of the first or second embodiment using the estimated value of the third, fourth, or fifth embodiment. If the calculation for realizing the database system processing to which the present invention is applied is expected to negate the effect of the acceleration compared to the execution time, the processing can be stopped (exactly, the processing is stopped) It is only necessary to determine the ratio of the execution time of the input / output instruction in the processing of the present embodiment and the conventional processing.

The processing power of the computer system which requires the most computing power in the processing of the first or second embodiment is that an entry having a key value which satisfies the selection condition among the entries of the level 0 of the index section is examined, and This is an index block input / output instruction for extracting an address. The time required to execute these input / output instructions is proportional to the number of data to be transferred because of the one-to-one correspondence between level 0 entries and data. The estimated value of the number of data to be transferred in the first or second embodiment is obtained for each data selection condition in the third, fourth, or fifth embodiment. Therefore,
The sum of the estimated values is a predicted value of the number of executions of the input / output instruction for the level 0 entry of the index unit for extracting the data to be transferred.

If the ratio between the predicted value and the total number of data transferred in the conventional processing (the estimated value is the total number of each condition, it can exceed the total number of data), then the embodiment The ratio of the calculation time required for the processing 1 or 2 itself to the time required for the conventional processing is calculated, whereby the former is expected to exceed the latter, and it is expected that the effect of speeding up the database processing according to the present invention will be negated. If it does, the process can be stopped.

Further, since the input / output of the normal computer system is performed not in units of entries or data but in units of blocks, the above-mentioned entries in the index section are indexed as described in the third, fourth or fifth embodiment. , And replace the number of all data with the total number of blocks in the data section, predict the time required for each input / output processing,
This can be used to determine whether to cancel the effect of speeding up.

Note that, depending on the design of the computer system, a dedicated instruction for collectively inputting / outputting a large amount of data may be provided in addition to the usual input / output in block units. In such a case, the input / output instructions used in the first or second embodiment and the conventional processing may be different.

At this time, the ratio of the number of entries in the index unit in the first or second embodiment to the number of data to be transferred in the conventional processing is multiplied by the ratio of the execution time of each input / output instruction, thereby obtaining the first or second embodiment. The calculation time of the processing itself of 2,
The ratio of the calculation time of the conventional processing is obtained.

In FIG. 2 showing the operation of the first embodiment, this embodiment is realized by estimating the calculation time of the part of the repetition processing of steps 102 to 105 at the stage when the processing of step 101 is completed. In the example of FIG. 5, processing for transferring all data in the data section 2 in FIG. 5 (conventional processing)
And the execution time ratio of the input / output instruction in the process of inputting the level 0 entry of the index unit belonging to the entry 16 (the process of the first embodiment of the present invention), and determine whether or not to suspend the process. I do.

Further, in the second embodiment, at the stage where the processing of FIG. 3 is completed, the calculation is performed by estimating the calculation time of the part of the repetition processing of steps 302 to 306 of FIG. First, for each condition selected by the condition group logical structure analysis shown in FIG. 3 in the second embodiment, a value that estimates the number of data that satisfies each condition is calculated. Next, the sum of these estimated values is obtained, the ratio of the execution time of each input / output instruction in the second embodiment and the conventional processing is calculated, and it is determined whether or not to stop the processing.

After the processing for reducing the data transfer amount is stopped according to the present embodiment, the processing can be switched to the processing for transferring all data from the computer system to the database device as in the conventional method, and the database processing can be continued.

As described above, according to this embodiment, the execution time of the input / output instruction in the first or second embodiment is predicted using the estimated value of the third, fourth, or fifth embodiment, and If the calculation time required for retrieval is expected to negate the effect of the high-speed database processing according to the present invention, there is an effect of stopping the processing.

Embodiment 7 FIG. Note that, in the sixth embodiment,
The execution time of an input / output instruction for inputting / outputting an entry of the index unit or a block of the index unit in order to extract the address of the data to be transferred was estimated. Is executed, the estimated value is updated each time an I / O instruction is actually issued, and the calculation time required to retrieve the data to be transferred is stopped before the effect of the database processing acceleration according to the present invention is negated. Can be. This corresponds to steps 102 to 10 in FIG. 2 showing the operation of the first embodiment.
5 or the loop of steps 302 to 306 in FIG. 4 in the second embodiment, the sum of the time spent for the processing up to that time and the time required for the remaining processing is the time required for the conventional processing. This is realized by stopping the processing when it is found that the number exceeds the threshold.

The predicted value of the execution time of the input / output instruction obtained in the sixth embodiment is the sum of the predicted values of the execution time of the input / output instruction for each condition selected to reduce the data amount using the index, Or, they are required to have been subjected to logical operations. Therefore, as for the loop portion of steps 302 to 306 in FIG. 4 showing the operation of the second embodiment, each time a condition is extracted and the index is checked, the condition obtained in the sixth embodiment is satisfied. It can be calculated by replacing the predicted value of the data input / output time with the actual input / output time. Example 1 is described in Example 2
And the condition can be realized as one.

As described above, according to this embodiment, the estimated value of the data transfer amount obtained by applying the third, fourth, or fifth embodiment is updated every time an I / O instruction is actually issued. There is an effect that the processing is stopped before the calculation time required to retrieve the data to be processed cancels out the effect of accelerating the database processing according to the present invention.

Embodiment 8 FIG. 3. A logical structure analysis of a group of conditions necessary for determining the condition of the outermost operator in step 202 of FIG.
An example of how to select the outermost operator will be described.
Define the logical operators, conditionals, conditional expressions and their redundancy necessary for this explanation. The logical operators are and, o
Although the operator is represented by r and not, when examining the outermost logical operator, only and and or are targeted.
A condition group represented by a logical expression including not is to be removed in advance from the logical expression by inverting each condition using De Morgan's law.
A conditioner is a condition that does not include an operator.
A condition object is a condition element, a group of condition elements connected by a logical operator and the logical operator, or a group of condition objects connected by the logical operator and the logical operator. (In the following example, it is assumed that one letter of an uppercase alphabet from A to F is a conditioner.) In this embodiment, the data selection condition is a database in the form of a conditional expression in which a conditioner and a logical operator are combined. Shall be given to the system. Note that the selection condition need not be given in the form of an expression as long as the logical relationship between the conditioners (the relationship between the logical operators) can be determined. For example, a method may be used in which each condition element is separately input and the logical structure is input later.

When different types of operators are described on an equal basis (when the priority is not specified in parentheses) for one conditioner, it is assumed that and acts more preferentially than or. The reverse may occur depending on the realization policy of the database system. In that case, the following and and or
May be applied in reverse. For example, Aan
dBorC is interpreted as (AandB) orC.
Aand (BorCandD) is Aand (Bor (C
andD)) (Aand (BorC)
andD etc.).

Redundant parentheses in conditional expressions refer to cases where logical operators acting inside and outside parentheses are of the same type. For example, the parentheses of Aand (BandC) are redundant.

In order to take out the outermost operator and to select one that can be used to reduce the number of data to be transferred under the condition that the index can be used, it is necessary to remove parentheses with redundancy like (AandB) andC. Also, when logical operators of different types are described equally for one conditional element (or a logical object enclosed in parentheses) as in B in the expression AandBorC, the applicable range of the logical operator is clarified. There is a need. Therefore, a logical structure analysis of the condition group is required.

The logical structure analysis of the conditional expression necessary for the condition determination of the outermost operator in step 202 of FIG. 3 and the method of selecting the outermost operator can be realized by using, for example, the following method.

First, a stack is prepared, and when a logical operator appears in an expression, the logical operator is put on the stack. When the scope of a logical operator ends (the end of parentheses, or a conditional object immediately following an unparenthesized and, or a terminator of a conditional expression),
Pops the top element off the stack.

Each element of the stack is actually composed of a list of logical operators and condition objects to which the logical operators are applied (if the condition objects to which the logical operators are applied can be specified, they are not necessarily lists). .

At this time, if the operator at the top of the elements remaining on the stack is of the same type as the fetched element, the fetched operator is included in the scope of the operator remaining on the stack. Fused operators are applied.
Thereby, redundant parentheses can be removed.

Further, when analyzing the conditional expression AorBandC, it is not determined whether B is directly connected to or only by looking at or next to A. For this reason, the conditioner or the logical operator or the delimiter of the scope of the logical operator (open parenthesis "(" or close parenthesis "(") or the terminal symbol of the conditional expression (notation omitted) One unit of the logical structure analysis is performed, and the processing is performed in different cases, and the unit elements of the analysis are sequentially extracted from the head of the conditional expression.

In the case of the AorBandC analysis described above,
With A and or, or and B, B and and, and and C as one unit element of the analysis, processing is determined for each unit element in order from the top of the conditional expression, and the analysis is advanced.

That is, the logical structure analysis of the conditional expression in this embodiment is based on two of the elements of the conditional expression, which are the unit elements of the analysis.
The execution is performed based on three pieces of information: a tuple, a stack, and a scope list of each logical operator stacked on the stack. The execution operation is a two-tuple of a conditional expression element that is a unit element of the analysis. Is divided into cases.

When the above processing is repeated to analyze even the terminal symbol of the conditional expression, since the application range of all logical operators is always terminated, only one element remains on the stack. The logical operator of the elements remaining on this stack is the outermost operator of this conditional expression.

Note that the classification of the analysis operation according to the type of the unit element is specifically described in the following items.

When the unit element is a conditional element and and, if the top element of the stack is not and, the and is stacked on the stack. Since this conditional element is included in the application scope of this and, this conditional element is put in the application scope list of this and. If the top element of the stack is "and", this conditional element is put in the applicable list of the "and" on the stack.

When the unit element is a conditional element or or, if the top element of the stack is not or, the or is stacked on the stack. If this conditioner has not been included in the scope list of the operator stacked on the stack before or, this conditioner is included in the scope list of this or. If this conditional is already in the scope list of the operator before or, do nothing more. If the top element of the stack is or, the conditioner is placed in the scope list of or on this stack (does not push a new or on the stack).

When the unit element is a conditional element and a closing parenthesis (")"), the conditional element is put in the applicable range list of the top element of the stack.

The unit element is a closing parenthesis (")") and an
In the case of d, the top element is first fetched from the stack. If the first operator of the elements remaining on the stack is and and the logical operator of the elements fetched from the stack is and, the scope of the elements fetched from the stack and the elements remaining on the stack are merged. Is the new scope list for the remaining element on the stack. The first operator of the elements remaining on the stack is and
And the logical operator of the element taken out of the stack is o
If it is r, add the scope list of the operator taken out of the stack as a condition object to the scope list of the remaining elements on the stack (the condition object is expressed by parentheses after removing redundancy from the condition group). Corresponding to the separated parts). If the first operator of the elements remaining on the stack is or, the and in the analysis unit element is pushed onto the stack. At this time, if the operator of the element taken out of the stack is and, this and
Is an application range list of and newly loaded on the stack. Alternatively, if the operator of the element taken out of the stack is or, the scope list of the taken out operator is added as a condition object to the scope list of the elements remaining in the stack. If there are no remaining elements on the stack, the and in the analysis unit element is pushed onto the stack. At this time, if the operator of the element taken out of the stack is and, this and
Is an application range list of and newly loaded on the stack. Alternatively, if the operator of the element taken out of the stack is or, the application range list of the taken out operator is added as a condition object to the application range list of and newly loaded on the stack.

The unit element is a closing parenthesis (")") and or
In the case of, first, the top element is taken out of the stack. If the head operator of the element remaining on the stack is and and the logical operator of the element fetched from the stack is and, the element fetched from the stack and the scope list of the head element remaining on the stack are added. Merge into a new coverage list of the remaining and on the stack (this process is equivalent to removing redundant parentheses). Further, the element is successively taken out of the stack, and the or in the unit element of the analysis is loaded on the stack. An element newly extracted from the stack (having a scope list fused for two and's) is added as a condition object of the scope list of this or. Also, the first operator of the elements remaining on the stack is and
And the logical operator of the element taken out of the stack is o
If r, the scope list of the operator extracted from the stack is added as a condition object to the scope list of the elements remaining on the stack. Further, elements are successively taken out of the stack, and a new or
Stack. An element newly extracted from the stack (an element obtained by adding a condition object including “or” to an application range list of an element including “and”) is added as a condition object in the application range list of “or” to be newly stacked on the stack. If the first operator of the elements remaining on the stack is or and the operator of the element fetched from the stack is and,
The scope list of d is set to or which is the top element of the stack.
(Or is not newly placed on the stack). Also, if the first operator of the element remaining on the stack is or and the operator of the fetched element is or, the new applicable scope list of the first element remaining on the stack will be Fuse the scope list into a new scope list for the remaining elements. If there are no remaining elements on the stack, the or in the unit element of the analysis is pushed onto the stack. At this time, if the operator of the element taken out of the stack is or, the scope list of this or is set as the scope list of or newly loaded on the stack. Alternatively, if the operator of the element taken out of the stack is "and", the application range list of the taken out operator is added as a condition object to the application range list of "or" newly loaded on the stack.

When the unit element is a conditional symbol or a terminal symbol of a conditional expression, first, the applicable range list of the top element of the stack is
Insert a conditional. Then, the processing in the case of the delimiter indicating the end of the application range of the above-described conditional element and the logical operator,
The analysis is repeated until there is only one element on the stack.

When the unit element is and and a condition element, a
nd is either already on the stack or removed by a redundancy check. Since the next element in the conditional expression determines which operator applies to the conditional element, the stack and each element on the stack are not changed here.

When the unit element is or and a conditional element, or
Is already on the stack or is removed by a redundancy check. Since the next element in the conditional expression determines which operator applies to the conditional element, the stack and each element on the stack are not changed here.

If the unit element is and and open parenthesis ("("), no change is made here to the stack and each element stacked on the stack. However, the first logical operator appearing after the open parenthesis is always Push onto the stack (redundancy check, comparison with the top element of the stack is omitted in this case. Redundancy is determined by removing the closing parenthesis corresponding to this open parenthesis, or the symbol at the end of the condition group) This is done to determine the scope of the logical operator in parentheses and then determine the redundancy).

If the unit element is or and an open parenthesis ("(")
Is the same as for and and open parenthesis above.
Here, the stack and each element on the stack are not changed. However, the first logical operator that appears after parentheses always pushes onto the stack.

When the unit element is an open parenthesis ("(") and a conditional element, as in the case of the above and and open parenthesis, no change is made here to the stack and each element stacked on the stack. The first logical operator that appears after parentheses always pushes onto the stack.

If the unit element is two closing parentheses (")"), the top element is first taken out of the stack. And the first operator of the elements remaining on the stack is and
And the logical operator of the element taken out of the stack is a
If nd or both are or, the element taken from the stack and the scope list of the first element remaining on the stack are merged into a new scope list for the operator remaining on the stack. . If the first operator of the elements remaining on the stack is and and the logical operator of the element taken out of the stack is or, or the two are reversed, the applicable range list of the elements remaining on the stack Is added as a condition object with the scope list of the operator extracted from the stack.

If the unit element is a closing parenthesis and a terminal symbol of a conditional expression, the processing of paragraph number 111 is repeated until one element remains in the stack.

If the unit element is two open parentheses ("("), no change is made to the stack and each element on the stack. It is always put on the stack like.

If a combination other than the above paragraph numbers 100 to 113 appears in the conditional expression, a syntax error occurs. For example, a conditional expression must not have two consecutive conditionals or logical operators. Open parenthesis and closing parenthesis are one-to-one.
Must be supported.

An example in which the analysis of the conditional expression AorBandC is executed by using the analysis method based on the case of the paragraph numbers 100 to 114 will be described. First A and or
Is taken out, and according to paragraph number 101, o
r is pushed onto the stack. Since A is not yet included in any applicable scope list, A is put at the head of this applicable scope list.

Next, unit elements or and B are extracted. No change is made to the stack and each element stacked on the stack according to the paragraph number 107, and B and and, which are the next unit elements of the analysis, are subsequently extracted.

At this time, since the top element of the stack is or, the and in the unit element is stacked on the stack according to paragraph number 100. In addition, B is put at the head of the list indicating the application range of this and.

Subsequently, a pair of and and C is extracted.
Paragraph number 106 keeps the stack and each element on the stack unchanged (since it is not possible at this point to determine whether C falls within the directly applicable range of this and) and retrieves the next unit element of the analysis.

The next unit element to be extracted is C and "the terminal symbol of the conditional expression" (the description is omitted). The elements of the stack are taken out from the top according to the paragraph number 105 and merged with the scope of the logical operator which is the element under the stack, or as the condition object, the scope of the logical operator which is the element under the stack is added as the condition object. I will put in the elements above.

In this example, at this time, the stack is and and from the top in order (B is included in the applicable range list),
or (A is included in the applicable range list).

First, "and" is taken out of the stack. Then, since the type of the operator is different from that of or which is the first (upper) element remaining on the stack, the application range list of and which was on the stack is put as a condition object in the application range of or. Can be

Since only this or remains in the stack, the analysis ends. This or is the outermost operator of this conditional expression.

Next, Aan, which is an expression including redundancy,
The analysis of dBand (CandD) will be described. First, A and and are taken out, and and is stacked on the stack by paragraph number 100. Then, A is put into the applicable range list of this and.

Subsequently, a pair of and and B is extracted. However, the stack is not changed by the paragraph number 100, and the next unit elements B and and are immediately extracted. The newly fetched and according to paragraph number 106 is compared with the top element of the stack.

In this case, since the newly fetched and is the same type of operator as the top element of the stack, the newly fetched and is not stacked on the stack, and B is added to the application scope list of and already stacked on the stack. (A and B will be included in the coverage list).

Next, and and open parenthesis are taken out. Regarding this pair of and and open parenthesis, no change is made to the stack according to paragraph number 108, and the pair of open parenthesis and C, which is the next unit element in the analysis, is extracted.

Also in this case, the stack is not changed by the paragraph number 110, and the next set of the unit element C and and is taken out.

Here, and is an operator of the same type as the top element of the stack, but since the logical operator appearing immediately after the opening parenthesis according to paragraph number 110 is always loaded on the stack, this and is loaded on the stack. , C is this and
In the scope list. At this point, and (and C in the applicable range list), an
d (A, B in the applicable range list) is stacked.

Subsequently, a pair of and and D is extracted. The stack is not changed by the paragraph number 106.

Subsequently, D and a closing parenthesis are extracted. According to paragraph 102, D is the top element of the stack and
In the scope list. At this point, and (from top to bottom, C and D in the coverage list),
and (A and B in the applicable range list) are stacked.

Finally, the closing parenthesis and the "terminal symbol of the conditional expression" are extracted. According to the paragraph number 112, first, "and", which is the top element of the stack, is extracted. At this time, since the fetched and is a logical operator of the same type as the first element remaining in the stack (and which first appears in the condition), the applicable scope lists of both are merged, and (A, B, C and D are included)
It becomes a new scope list for the remaining elements on the stack, and.

Here, since there is a “terminal symbol of the conditional expression” in the unit element of the analysis, the analysis is completed, and it is found that the last remaining “and” is the outermost operator.

As described above, paragraph numbers 92 to 11
By executing the analysis described in 4 above, it is possible to extract information about the outermost operator of the condition group and information obtained by removing redundancy from the logical structure of the condition group. here,
If it takes too much calculation time to select data that satisfies the condition that can use the index, the calculation time may be reduced by transferring less data, canceling the saved time, and the effect of speeding up may be lost. In order to avoid this, when an indexed condition requiring complicated processing needs to be used to reduce the data to be transferred, the processing can be stopped.

The complexity of the logical structure for a given conditional is defined as the number of logical operators acting on the conditional. This corresponds to a value obtained by adding 1 to the overlapping depth of parentheses for a certain part of a conditional expression when the conditional expression representing the logical structure of the condition group is modified so as not to have redundancy. For example, in AandB, the complexity of the logical structure of both A and B is 1. In Aand (BorC), the complexity of the logical structure of A is 1, and the complexity of the logical structure of B and C is 2. Aand (BandC) becomes AandBandC when logical redundancy is removed, so that the complexity of the logical structure is 1 for A, B, and C. Aand ((Band
C) orD), the complexity of the logical structure of A is 1, D
The complexity of the logical structure of B is 2 and the complexity of the logical structure of B and C is 3.

Thus, for example, when the outermost operator is “and”, it is necessary to calculate a conditioner having a logical structure having a complexity of 4 or more in order to determine the data to be transferred. When it is necessary to calculate a conditioner having a logical structure with a complexity of 3 or more in the case of (1), stop the processing before performing the complicated logical calculation and canceling out the effect of increasing the speed of the database processing according to the present invention. Can be. The degree of complexity of the logical structure when it is necessary to calculate a conditional element to stop processing depends on the design policy of the database system using the present invention.

After the processing is stopped, the processing can be switched to the processing for transferring all data from the computer system to the database device, as in the conventional processing.

Although the stack is used for the logical structure analysis here, the same processing can be performed by generating a syntax tree of a conditional expression using a tree structure. In addition, the notation of the condition group does not need to be the above method as long as it has the same descriptive power as the above conditional expression.

As described above, according to this embodiment, information on the outermost operator of the condition group and the information obtained by removing the redundancy from the logical structure of the condition group is extracted, and the indexed condition is added to this processing. A logical structure analysis for examining the complexity of the processing at the time of use is realized, and when it is necessary to perform processing that is too complicated to reduce the data to be transferred, the processing is stopped.

Embodiment 9 FIG. In the first or second embodiment, each data block in the data section of the computer system is provided with additional information (a header portion of the block, etc.) for knowing the maximum value and the minimum value of the key value of the data in the block. For example, the number of data blocks actually transferred from the computer system to the database device can be reduced by simply adding a simple process to the first or second embodiment.

In the first or second embodiment, in the example shown in FIG. 5, although the block 19 does not include any data group 17 to be transferred to the database device 4, it is a continuous block including data satisfying the selection condition. Since it was included in the area, it was transferred from the computer system 1 to the database device 4. However, for each data block,
If there is additional information for knowing the maximum and minimum key values of the data in the block, the additional information of each block belonging to 18 is checked, and the block 19 does not include any data group 17 (data group 17). Is out of the block, and a part thereof is not included in the block 19). Therefore, the transfer of the block 19 to the database device 4 can be omitted without checking all the data in the block group 18.

As described above, according to this embodiment, in each of the first and second embodiments, each data block in the data section of the computer system has an additional function of knowing the maximum and minimum key values of the data in the block. If information (a header portion of a block, etc.) is provided, the first or second embodiment is used.
By simply adding simple processing to the above, there is an effect of reducing the number of data blocks actually transferred from the computer system to the database device.

Embodiment 10 FIG. Note that, in the first or second embodiment,
, Each block belonging to level 0 of the index unit 3 has additional information (header part of the index block, etc.) including the maximum value and the minimum value of the data block address (pointer to the data block) held in the block. ), The number of data blocks actually transferred from the computer system to the database device can be reduced by simply adding simple processing to the first or second embodiment.

In the first or second embodiment, in the example shown in FIG. 5, if the additional information of the index unit 3 is used, the block 19 becomes the data group 1 at the time of extracting the address of the data block.
7 is not included (the data belonging to the data group 17 protrudes from the block, and a part thereof is not included in the block 19). Therefore, the transfer of the block 19 to the database device 4 can be omitted without checking all the data in the block group 18.

As described above, according to this embodiment, in each of the first and second embodiments, each block belonging to the level 0 of the index unit 3 has the data block address (data block address) stored in the block. ), The additional information (the header part of the index block, etc.) including the maximum value and the minimum value of the pointers is simply added to the first or second embodiment, and the database is actually transferred from the computer system to the database. This has the effect of reducing the number of data blocks transferred to the device.

[0145]

According to the first data transfer method, in a database system in which a computer system and a database device are connected by a connection device, if a database process including a data selection condition is given, an index can be used for the data selection condition. Judgment, when usable, a data block group including all data satisfying the data selection condition in the computer system is obtained using the index, and the data block group is transferred from the computer system to the database device. Therefore, the data transfer amount can be reduced and the database processing can be executed at a higher speed than in the case where all the data in the conventional processing is transferred to the database device.

[0146] The second data transfer Okukata method, in a computer system and database system by connecting the database apparatus in connection device, determine the number of entries in the root block of the index used to determine the selection criteria data, Since this number is multiplied by the number of entries included in one block of the index by the number obtained by subtracting 1 from the number of levels in the index, the number of data satisfying the selection condition is estimated. At this stage, the amount of data transferred from the computer system to the database device can be estimated, and the effect of speeding up can be predicted.

[0147] The third data transfer Okukata method, computer in the system and the connected database system database device connection device, a 1 under the level of entry than the root block of the index used to determine the selection criteria data The number of entries satisfying the selection condition is estimated by multiplying the number by the number of entries included in one block of the index by the number obtained by subtracting 2 from the number of levels of the index. The amount of data transferred from the system to the database device can be more accurately estimated.

[0148] The fourth data transfer Okukata method, in a computer system and the connected database system database device connection device, is used to examine the selection criteria data included in one level below the root block of the index The number of sub-blocks having the head entry is obtained, this number is multiplied by the number of entries included in one sub-block, and the number is further reduced by 2 from the number of levels in the index. Since the number of data that satisfies the selection condition is estimated by multiplying the number of subtractions, the number of data transferred from the computer system to the database device can be estimated more accurately and the calculation load can be estimated to be small.

[0149] The fifth method of data transfer in the first data transfer method, the second or third or fourth data transfer
It estimates the number of selected data that satisfies the Okukata method predicts the output time of the block of level 0 of the index portion required for the processing to determine the data to be transferred to the address of the data from the computer system retrieves the database device In comparison with the execution time of the input / output command required for the process of transferring all data from the computer system to the database device, if the former is longer than the latter, the process of accelerating is stopped, so that the database processing is performed. If the effect of speeding up is predicted to be small, the processing can be stopped.

[0150] The sixth method of data transfer in the first data transfer method, the position of the data block group containing all the data satisfy the conditions were selected for indexed in a computer system using the index obtaining process in the middle The number of I / O instructions used to read the index in this process
The predicted value of the fifth data transfer method is updated, and compared with the execution time of the input / output instruction required for the process of transferring all data from the computer system to the database device, if the former is longer than the latter, the speed is increased. , The calculation time required to retrieve the data to be transferred can be stopped before the effect of the high-speed database processing is negated.

The seventh data transfer method uses a logical operator (and, or, no) in a database system in which a computer system and a database device are connected by a connection device.
Given a database processing that includes the data selection conditions linked in t), if an index is available for the conditions, then the logical operator of the data selection conditions
Or a combination of two consecutive unit elements in the data selection condition, with a separator of the applicable range of the logical operator, a conditioner not including the logical operator, or a terminal symbol of the condition group as one unit element. A condition group that needs to calculate a condition group affected by at least a certain number of the logical operators in order to clarify the applicable range of the logical operator based on the classified information and reduce the data transfer amount. Is configured to stop the processing, so that when it is necessary to perform too complicated processing to reduce the data to be transferred, the processing can be stopped.

[0152] data transfer method of the eighth, the computer system and the connected database system database device connection device, if the database processing including tied data selection condition by logical operators are given for the condition index Can be used, the indexed ones of the data selection conditions are combined, and a set of conditions that includes all the data that satisfies the given selection conditions of all data is selected, and an indexed selection in the computer system is selected. The maximum value and the minimum value of the address of a continuous data block group including all data satisfying the above conditions are obtained by using an index, and the block group sandwiched by the blocks having this address is transferred from the computer system to the database device. Database processing by reducing the amount of data transferred. The can be executed at high speed.

[0153] ninth method of data transfer in a computer system and the connected database system database device connection device, if the database processing including tied data selection condition by logical operators are given for the condition index Can be used, the indexed ones of the data selection conditions are combined, and a set of conditions including all data satisfying the selection conditions of all given data is selected, and a block in the block in the computer system is selected. With respect to a data block having additional information for knowing the maximum value and the minimum value of the key value of the data, all data blocks including all data satisfying the selected condition with an index are obtained using the index and the additional information. In a configuration in which blocks are transferred from the computer system to the database device Since, it is possible to reduce the number of data blocks to be transferred from the computer system to the database device.

[0154]

[Brief description of the drawings]

FIG. 1 is a configuration diagram illustrating a database system that is Embodiment 1 of the present invention.

FIG. 2 is a flowchart of the first embodiment of the present invention.

FIG. 3 is a flowchart illustrating an operation of analyzing a logical structure of a condition group according to the second embodiment of the present invention.

FIG. 4 is a flowchart illustrating an operation of reading data satisfying each selection condition and transferring the data to a database device according to the second embodiment of the present invention.

FIG. 5 is a detailed diagram of a data part and an index part according to the first embodiment of the present invention.

FIG. 6 is a diagram illustrating an operation principle when the outermost operator is “and” in the logical structure analysis of the condition group in the second embodiment of the present invention.

FIG. 7 is a diagram illustrating the principle of operation when the outermost operator is or in the logical structure analysis of a condition group according to the second embodiment of the present invention.

FIG. 8 is a configuration diagram of a database system including a conventional database processing device.

[Explanation of symbols]

1 Computer system, 2 data (data section, all data), 3 indexes, 4 database devices, 5 data section, data to be finally selected, 6 index entries satisfying selection conditions, 7 selection conditions that can use indexes , Data finally selected via the database device, data flow transferred from the computer system to the database device according to the present invention, selection conditions selected by the database device and satisfying selection conditions Data, 11 individual data (data record), 12
Data block, 13 Index section entry, 14 Index section block (particularly root block), 15 Flow of data transferred from the computer system to the database device in the conventional manner.

──────────────────────────────────────────────────続 き Continuation of the front page (56) References JP-A-5-81340 (JP, A) JP-A-64-42731 (JP, A) JP-A-3-2938 (JP, A) 233720 (JP, A) Tsuchida, Mutoh, "Proposal of Optimization Method for Embedded Query in RDB" Proc. Of the 36th (Early 1988) National Convention on Information Processing, Proceedings of the National Convention (▲ I ▼), p. 505-506 (Showa 63-3-16) (58) Field surveyed (Int. Cl. ⁷ , DB name) G06F 17/30 G06F 12/00 512 JICST file (JOIS)

Claims (9)

(57) [Claims] In a database system in which a computer system and a database device are connected by a connection device, when a database process including a data selection condition is given, the computer system indexes the data selection condition.
Part is determined whether or not the use, when the index portion is used, determine the memory addresses of all data blocks containing data that meets the data selection condition with the index portion of the data block group Consecutive memory addresses are all included
The data blocks in the address area are transferred to the database device , and the database device is transferred from the computer system.
Data blocks in consecutive address areas
Each data of the received data block group is
Data selection conditions are met, and
Data that does not satisfy the data selection conditions from the
And select data that satisfies the data selection conditions
Data transfer method and outputting. 2. Whether the index unit of the computer system can be used.
Whether the judgment determines the number of root block entry index portion is used to examine the selection criteria data, the number of levels in the index portion of the number of entries in the one block of the index portion to the number estimate <br/> the number of the selection criteria data by the number obtained by multiplying the number of times obtained by subtracting 1 also Ri, the number of estimated data takes the data to be transferred
Of I / O instruction to entry of index part to issue
By setting the predicted value of the number of times,
Claim 1 Symbol, which comprises carrying out the Rukoto determined between
Data transfer Okukata method of mounting. 3. An index unit of the computer system can be used.
Whether the judgment determines the number of entries one level below the root block of the index portion is used to examine the selection criteria data, the index unit the number of entries included in one block of the index portion to the number also Ri, the number of estimated data transfers estimates the number of the selection criteria data by the number obtained by multiplying the number of times obtained by subtracting 2 from the number of levels of
Entry / exit of index part entry to retrieve data
By using the predicted value of the number of executions of
Data transfer Okukata method according to claim 1, characterized in that the Rukoto seek predicted input time. 4. An index section of the computer system can be used.
To determine whether or not the number of sub-blocks having a head entry used for examining data that satisfies the selection condition included one level below the root block of the index part , the number is included in one sub-block. Multiplied by the number of entries
Further Ri also the number of the selected data that satisfies the number of the number of entries by multiplying the number of times obtained by subtracting 2 from the level number of the index portion included in one block of the index portion to the number estimate, estimated data The number of data to be transferred
I / O instructions for index entry to retrieve
The predicted value of the number of executions of
Claims, which comprises carrying out the Rukoto determined output time
Data transfer Okukata method claim 1. 5. The method according to claim 1, further comprising determining an estimated input / output time of said index unit.
Is the index part level 0 (most distant from the root part of the index part)
A layer which is a block of the layer) containing a pointer to the data portion
It is to obtain the prediction output time for access, this
Predictive I / O for accessing level 0 block of index unit
Compared time, the execution time of the input-output instructions necessary all data from the computer system to the processing to be transferred to the database system, if the former is longer than the latter placing claim 1
Stops the process of the computer system, wherein all data
5. The data transfer method according to claim 2, wherein the data is transferred from the computer system to a database device . 6. in the middle of the process for obtaining with the index part storage addresses of all the data blocks including the data selection condition is satisfied data, input and output commands using to actually read the index portion in this process Input / output by calculating the number of times
The time is updated, and the updated predicted input / output time is compared with the execution time of the input / output instruction required for the process of transferring all data from the computer system to the database device, and the former is longer than the latter. Is a computer system according to claim 1.
Stop processing the stem, the computer cis all data
6. The data transfer method according to claim 5 , wherein the data is transferred from the system to the database device . Wherein said computer system, if the logical operators the Jokenko (and, or, not) a database processing including only data selected condition group with knot at a given, there
The complexity of the logical structure of the conditioner, the effect on the conditioner
It is defined as the number of logical operators on, data selection conditions
The complexity of the logical structure of the conditionals in the group
When the complexity of the logical structure is more than a predetermined value,
Stops the logical calculation process for the condition
2. The data transfer method according to claim 1 , wherein all data is transferred from the system to the database device . 8. The computer system according to claim 1, wherein a maximum value and a minimum value of the storage address of the continuous data block group are obtained by using an index unit, and the maximum value of the storage address of the data block group is determined.
2. The data transfer method according to claim 1, wherein a block group sandwiched by the minimum value and the minimum value is transferred to a database device. 9. stored before Symbol computer system data
Block, the maximum value of key values of the data in the data block, has an additional information including the minimum value, the computer system, the maximum value of key values of the data in the data block in the additional information, the minimum value Is used to determine whether or not the data block includes data to be transferred to the database device. If the data block does not include data to be transferred to the database device, the data block 2. The data transfer method according to claim 1, wherein the groups are not transferred to the database device.