JPH04358266A - Data retrieving system applying hash method - Google Patents

Abstract

PURPOSE:To eliminate the deviation of the retrieving time caused by the deviation of the number of data retrieved in a subgroup in 8 data retrieving system applying a hash method. CONSTITUTION:The retrieval data (normal retrieval data) in number of pieces almost equal to the key value of the data retrieved in the subgroups sorted by the hash arithmetic are stored in a retrieval memory 4b1 for each subgroup. Meanwhile the remaining retrieval date (overflowed retrieval data) included in each subgroup are stored in an auxiliary retrieval memory 4b2 for each subgroup. Then both memories 4b1 end 4b2 are actuated in parallel with each other to retrieve the data. In such a constitution, a part of the data (overflowed retrieval data) is retrieved by the memory 4b2 in parallel to the memory 4b1 for the subgroup having a large number of retrieval date. Thus the date retrieving time is shortened.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a hardware data retrieval method using a hash method.

0002

2. Description of the Related Art Conventionally, a hash method is one of the methods used to search for one data element from a data group consisting of a large amount of numerically expressed data. The outline is described below. First, there is one piece of data y, which is expressed numerically and is similarly expressed as group A, which is composed of N pieces of data xi (1≦i≦N) with no overlap in numerical expression. Now, consider searching whether data xi having the same numerical expression as key data y exists in group A.

At this time, N pieces of data xi of data A are classified in advance into a plurality of subgroups based on a certain rule. Since the data xi is expressed numerically,
The above classification rules are used here as numerical calculation results. For example, if the classification rule is classified by the remainder Rx obtained by adding data xi with a constant K (1≦K), the remainder Rx takes a value from 0 to K-1, so data group A is divided into K subgroups zz
It can be classified into (1≦j≦K).

[0004] In order to perform the desired retrieval process, first, the same numerical operation as in the above-mentioned classification rule is performed on the key data y, that is, a remainder calculation using a constant K is performed to obtain a remainder Ry. Naturally, Ry also obtains K values between 0 and K-1. Then, one subgroup among the K subgroups zz is specified by j=Ry, and that subgroup zi(i=Ry
) and the numerical representation of the key data, and if there is a matching element, it means that there is data xi with the same numerical representation as the key data y in group A, and if there is no matching element, the group It can be determined that there is no data xi in A that has the same numerical expression as the key data y.

[0005] The above search method is called a hash method, and since the data search process is specified within one subgroup zz (j=Ry), it is a method for performing a simple search on all elements xi in group A. Compared to this, much faster search processing is possible. FIG. 2 is a block diagram showing the configuration of a conventional data search method that implements the above search processing using hardware. The configuration of the conventional data search method will be explained below.

The search memory 4a is a memory with sufficient address space to store all search data, and
A word consists of two fields storing search data x1 and flag f1. All search data is classified in advance into subgroups based on one classification rule, and arranged and stored in the search memory 4a in units of subgroups. Subgroup z1 in the figure is one of them. flag f1
is set to logic "1" only in the last word of subgroup z1, and the logic values in other words are set to "0". In other words, the word whose flag f1 has a logical value of "1" indicates that it is the final word of the subgroup z1.

The computing unit 1a is a computing unit that has a classification function according to the same rules as the classification rules for the search data in the search memory 4a, and the key data y inputted from an external circuit as an input.
1 to generate classification data o1 and input it to the address converter 2a. The address converter 2a generates a subgroup address p1 indicating the starting address of one subgroup z1 in the search memory 4a from the classification data o1, and outputs it to the address counter 3a. There is a one-to-one correspondence between the subgroup address p1 and the classification data o1. The address converter 2a is constituted by a memory in which an address is specified by the classification data o1 and a subgroup address p1 is written in advance in correspondence to the specified address.

[0008] Address counter 3a takes in subgroup address p1 under the instruction of load signal l1 inputted from control circuit 6a, and outputs it to search memory 4a as search memory address a1. It also has a function of counting up the search memory address a1, that is, incrementing it by 1, in accordance with the instruction of the count signal e1 inputted from the control circuit 6a. Search data x1, which is the output data of search memory 4a, is input to comparator 5a. The same key data y1 that is simultaneously input to the arithmetic unit 1a is input to the comparator 5a. Comparator 5a compares search data x1 and key data y1, generates a match signal c1, and outputs it to control circuit 6a. The match signal c1 takes a logic value of "1" if the search data x1 and the key data y1 match, and takes a logic value of "0" otherwise.

The control circuit 6a receives the coincidence signal c1 and the flag f1.
By monitoring the state of , search processing using the search memory 4a is performed, and the results are output as a hit signal h1 and a search end signal s1. The search end signal s1 has a logical value of "0" during the search process, and has a logical value of "1" when the search process is completed. The hit signal h1 has meaning only when the search end signal has a logical value of "1", and the logical value of "1" is the search memory 4a.
A logical value of "0" indicates that there is search data having the same value as the key data y1, and vice versa. Further, the control circuit 6a starts the search process by detecting that the logical value of the activation signal g1 inputted from the external circuit changes from "0" to "1".

The operation of the conventional data retrieval method will be explained below. First, the external circuit inputs the key data y1 to the arithmetic unit 1.
a and gives a transition from "0" to "1" to the logical value of the activation signal g1. When the control circuit 6a detects the above transition of the activation signal g1, the control circuit 6a sets the search end signal s1 to a logical value of "0".
'', and at the same time, the load signal l1 instructs the address counter 3a to load the subgroup address p1.

The address counter 3a takes in the subgroup address p1 and outputs it to the search memory 4a as a search memory address a1. At this time, the value of the search memory address a1 is the starting address of one subgroup z1 in the search memory 4a. will be shown. The search memory 4a reads one word specified by the search memory address a1, sends the search data x1 to the comparator 5a, and sends the flag f1 to the control circuit 6a.
hand over to.

The comparator 5a checks the match/mismatch between the search data x1 and the key data y1 and outputs it as a match signal c1 to the control circuit 6a. Here, the control circuit 6a performs the following processing. First, if the match signal c1 = logical value "0" and the flag f1 = logical value "0", the search memory is sent to the address counter 3a using the count signal e1 in order to search for the next search data of subgroup z1. address a
Give instructions to count up by 1. The counted up search memory address a1 is given to the search memory,
The next word is read out. Repeat the same operation below,
All search data in subgroup z1 and key data y1 are compared. Completion of the comparison process with all search data in the subgroup can be detected by flag f1=logical value "1".

Even if the above comparison operation is repeated until the flag f1=logic value "1" is detected, the coincidence signal c1=
If the logical value "1" cannot be detected, it means that there is no search data that matches the key data y1 in the specified subgroup z1, and the search end signal s1 is set to the logical value with the hit signal h1 = logical value "0". Set to "1" and notify the external circuit.

[0014] Furthermore, if the match signal c1="1" is detected before the flag f1="1" is detected, it is found that search data matching the key data y1 exists in the subgroup z1. At that time, all search processing is completed, and a hit signal h1 and a search end signal s1 are output.
Notify the external circuit as a common logic value "1".

[0015]

[Problem to be Solved by the Invention] However, the hashing method performs an operation (hash operation) on the search data and uses the fact that the result becomes a kind of function to classify the search data into subgroups. This results in bias in the number of search data classified into one subgroup. Therefore, in the conventional data search method shown in FIG. 2, the search time becomes uneven.

FIG. 3 is an explanatory diagram showing the bias in the number of search data in subgroups when a search group is classified into eight types of subgroups using the hash method. The number of search data in each subgroup is calculated and graphed. If the search data in FIG. 3 is searched using the conventional data search method shown in FIG. 2, the search will be completed in R times or less, except for subgroup number 5.
Searching for subgroup number 5 takes longer than searching for other subgroups, resulting in lower search performance.

Such bias in the number of search data within a subgroup can be alleviated to some extent by using an operation that can uniformly classify the search data or a calculation constant in the hash operation, but it is impossible to completely eliminate the bias. I can't. The present invention uses the above-mentioned search data by hash calculation.
It is an object of the present invention to provide a data search method that eliminates bias in search time due to bias in the number of search data within subgroups that occurs when classifying into subgroups, and enables high-speed search processing.

[0018]

[Means for Solving the Problems] In order to solve the above-mentioned problems, the present invention provides a data retrieval method using a hash method, in which at least one of the search data classified into subgroups by a hash operation is selected. It is equipped with a search memory that stores some search data in subgroup units, and one or more auxiliary search memories that store the remaining search data in each subgroup in subgroup units. It was configured to operate in parallel and perform data searches.

[0019]

[Operation] According to the present invention, since the data search method is configured as described above, the search memory stores at least part of each subgroup, for example, approximately the representative value of the number of search data in each subgroup. An equal number of data (normal search data) is stored, and the remaining search data (for example, overflow search data) in each subgroup is stored in the auxiliary search memory. Then, the search is performed by operating the search memory and the auxiliary search memory in parallel.

[0020]

Embodiments Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. FIG. 1 is a block diagram showing the configuration of a data search method according to an embodiment of the present invention. The configuration of a data search method according to an embodiment of the present invention will be described below. Search memory 4b1 and auxiliary search memory 4b2 have the same functions as search memory 4a in FIG. 3, but differ in the following points. First, the search memory 4b1 stores the search data portion in which the search within the subgroup is completed in R times or less, that is, the normal search data ND shown in FIG. 3. Then, the auxiliary search memory 4b2 stores the search data portion that does not end unless the search within the subgroup is executed R times or more, that is, the overflow search data OFD shown in FIG. 3. Here, "R" is
Set approximately equal to the representative value (statistically concentrated value) of the number of data in each subgroup classified by the hash method. The classification regulation and arrangement method of subgroup z2a in search memory 4b1 are the same as in search memory 4a, but subgroups that require more than R searches, such as subgroup number 5 in FIG. 3, are stored in normal search data ND. The portion overflows to the search memory 4b1 and the search data OFD is stored in the auxiliary search memory 4b2. The classification rules and arrangement method of the subgroup z2b in the auxiliary search memory 4b2 are as per the search memory 4b.
Same as 1. Search memory 4b1, auxiliary search memory 4
The word structure of b2 is the same as that of the search memory 4a in FIG. 2, and the meanings of flags f2a and f2b are also the same.

The key data y2 is the arithmetic unit 1b and the comparator 5b.
1, is input to the comparator 5b2. Arithmetic unit 1b is shown in Figure 2.
The arithmetic unit 1a, comparator 5b1, and comparator 5b2 have the same functions as the comparator 5a in FIG. Arithmetic unit 1b outputs classification data o2 and delivers it to address converter 2b1 and address converter 2b2. Address converter 2b1 has the same function as address converter 2a in FIG. 2, and inputs search memory subgroup address p2a to address counter 3b1. Further, the address converter 2b2 converts the classified data o2 into an auxiliary search memory subgroup address p2b which is the start address of the subgroup z2b in the auxiliary search memory 4b2.
and input it to the address counter 3b2, and
A gate signal w indicating whether auxiliary search is necessary is output to AND gate 8 and AND gate 9. This is because subgroups classified using the hash method do not always have overflow search data OFD. Here, the gate signal w has a logic value of "1" indicating that auxiliary search is necessary, and a logic value of "0".
” indicates that the auxiliary search is not required. Note that the address converter 2b2
is constituted by a memory like the address converter 2b1, but data indicating whether or not an auxiliary search is necessary is written in the address designated by the classification data o2 together with the subgroup address p2b.

Address counter 3b1 is connected to control circuit 6b.
The search memory subgroup address p2a is taken in by the load signal l2 inputted from the search memory subgroup address p2a and outputted to the search memory 4b1 as the search memory address a2a. Further, the search memory address a2a is counted up according to the instruction of the count signal e2 inputted from the control circuit 6b. Similarly, the address counter 3b2 takes in the auxiliary search memory subgroup address p2b as instructed by the load signal l2, outputs it to the auxiliary search memory 4b2 as a search memory address a2b, and also receives the auxiliary search memory address p2b as instructed by the count signal e2. Count up a2b.

[0023] Normal search data x2a, flag f2a
is the readout output from the search memory 4b1, and the comparator 5
b1, is input to the control circuit 6b. Similarly, overflow search data x2 is read output from the auxiliary search memory 4b2.
b is input to the comparator 5b2, and the flag f2b is logically ANDed with the gate signal w by the AND gate 9, and then becomes the flag f2, which is passed to the control circuit 6b.

The search memory match signal c2a, which is the output of the comparator 5b1, is input to the OR gate 7. Comparator 5b2
The auxiliary search memory match signal c2b, which is the output of the auxiliary search memory match signal c2b, is input to the AND gate 8, and after being ANDed with the gate signal w, it is input to the OR gate 7, and ORed with the match signal c2a. The coincidence signal c2, which is the output of the OR gate, is delivered to the control circuit 6a.

Similar to the configuration example of FIG. 2, the control circuit 6b monitors the states of the match signal C2, the flag f2a, and the flag f2 while controlling the load signal l2 and the count signal e2, and controls the search memory 4b1 and the auxiliary search memory 4b2. The subgroup z2a and subgroup z2b within the subgroup z2a and z2b are searched in parallel. The hit signal h2, the search end signal s2, and the start signal g2 that are the input of the control circuit 6b have the same functions as the hit signal h1, the search end signal s1, and the start signal g1 in FIG. 2, respectively.

The operation of the data search method according to the embodiment of the present invention will be explained below. First, the external circuit sets key data y2 and gives a transition from "0" to "1" to the logical value of the activation signal g2. When the control circuit 6b detects the above-mentioned transition of the start signal g2, it sets the search end signal s2 to the logical value "0" and simultaneously controls the address counter 3b1 by the load signal l2.
, instructs the address counter 3b2 to load the search memory subgroup address p2a and the auxiliary search subgroup address p2b.

The address counter 3b1 takes in the search memory subgroup address p2a and outputs it to the search memory 4b1 as a search memory address a2a, but the value of the search memory address a2a at this time is one subgroup z2a in the search memory 4b1. This is the first address. Similarly, the auxiliary search memory address a2b, which is the output of the address counter 3b2, becomes the start address of one subgroup z2b in the auxiliary search memory 4b2, but if the overflow search data OFD portion of FIG. 5 does not exist in the subgroup z2a, , a logic value "0" is output to the gate signal w, and the auxiliary search memory match signal c2b, which is the search result in the auxiliary search memory 4b2, is invalidated.

The search memory 4b1 has a search memory address a.
2a is read, normal search data x2a is sent to the comparator 5b1, and the flag f2a is sent to the control circuit 6b.
Enter. Similarly, the auxiliary search memory 4b2 reads out one word addressed by the auxiliary search memory address a2b, and transfers overflow search data x2b and flag f2a to the comparator 5b2.

The output c2b of the comparator 5b2 is logically operated on the gate signal w and the search memory match signal c2a by an AND gate 8 and an OR gate 7, and is then inputted to the control circuit 6b as a match signal c2. After the flag f2b is also subjected to a logical operation with the gate signal w by the AND gate 9, it is inputted as f2 to the control circuit 6a. Here, the control circuit 6b performs the following processing.

If the logical value "1" of the coincidence signal c2 is detected before the logical value "1" of the flag f2a or f2 is detected, the key data y2 and Matching search data or auxiliary search data exists, and the hit signal h2 = logical value "1" and search end signal s2 are generated as in the configuration example of FIG.
=Notify the external circuit using the logical value "1".

[0031] Before detecting the coincidence signal c2, the flag f2a is
If a logical value of "1" is detected, that is, there is no data in subgroup z2a, and subgroup z2a is
If the search for b has not been completed, the count signal e2 causes the auxiliary search memory address a of the auxiliary search memory 4b2 to be
2b is counted up and the search process is repeated until the logical value "1" of the flag f2 is detected. Note that at this time, the count signal e2 is also output to the search memory 4b1, so
The flag f2a, which is the output of the search memory 4b1, is removed from the monitoring target. If the match signal c2 cannot be detected even if the logical value "1" of the flag f2 is detected, it means that there is no data in the subgroup z2b, so the hit signal h2= The external circuit is notified using the logic value "0" and the search end signal s2=logic value "1". On the other hand, if the match signal c2 is detected before the logical value "1" of the flag f2 is detected, it is found that data exists in the subgroup z2b, so the hit signal h2 is detected.
= logical value "1" and search end signal s2 = logical value "1" to notify the external circuit.

On the other hand, if the logical value "1" of only the flag f2 is detected before detecting the coincidence signal c2, the flag f2
is removed from the monitoring target, and search processing is performed until the flag f2a becomes the logical value "1". In this case as well, the search result is notified to the external circuit using two signals, the hit signal h2 and the search end signal s2, as in the configuration example of FIG. Although the search memory 4b1 and the auxiliary search memory 4b2 are shown as separate memories in FIG. 1, the search memory area and the auxiliary search memory area may be provided in a single memory. The same applies to address converters 2b1 and 2b2.

Furthermore, the search memory address a2a and the auxiliary search memory address a at the time of a search hit are sent to the external circuit.
2b, and in the external circuit,
The address is sent to another circuit that is configured similarly to the above and that searches for data related to the data searched in FIG. 1 (for example, searches for a telephone number in FIG. 1 and searches for the name in another circuit). By outputting the data, the search for related data can be completed at the same time.

Note that the present invention is not limited to the above-mentioned embodiments, and various modifications can be made based on the spirit of the present invention, and these are not excluded from the scope of the present invention.

[0035]

[Effects of the Invention] As explained in detail above, according to the present invention, the following effects can be achieved. (1) Under the condition that the number of overflow search data ≦ R, the search result can be determined by searching within R times and at the same time, the search method can be set to 2.
Search speed can be up to twice as fast. (2) Generally, in subgroups in which search data is classified using the hash method, the number of overflow search data takes a value much smaller than the average value of the number of normal search data, so the capacity of the auxiliary search memory 4b2 is smaller than the capacity of the search memory 4b1. A small amount is sufficient compared to . Therefore, if there is enough capacity, the data in the subgroup is distributed to the search memory 4b1 and the auxiliary search memory 4b2 at an appropriate ratio (for example, in FIG. 5, the search data of each subgroup is times(
Store the part exceeding R′<R) in the auxiliary search memory)
This makes it possible to perform parallel searches and improve search speed. (3) Search speed can be further improved by providing two or more auxiliary search memories 4b2 and their search means to perform parallel searches.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing the configuration of a data search method according to an embodiment of the present invention.

FIG. 2 is a block diagram showing the configuration of a conventional data search method.

FIG. 3 is an explanatory diagram showing the bias in the number of search data in subgroups using the hash method.

[Explanation of symbols]

1b Arithmetic unit 2b1, 2b2 Address converter 3b1, 3b2
Address counter 4b1 Search memory 4b2 Auxiliary search memory 5b1, 5b2 Comparator 6b Control circuit 7 OR gates 8, 9 AND gate

Claims (3)

[Claims] Claim 1: In a data search method using a hash method, (a) a search in which at least some search data in each subgroup is stored in subgroup units among search data classified into subgroups by a hash operation; and (b) one or more auxiliary search memories storing the remaining search data in each subgroup in units of subgroups, and performs data searches by operating the search memory and the auxiliary search memory in parallel. A data retrieval method using the hash method, which is characterized by the following. 2. The data search method using a hashing method according to claim 1, wherein the search memory stores a number of data approximately equal to a representative value of the number of search data in each subgroup. 3. The data search method using a hashing method according to claim 1, wherein the number of data to be stored in the search memory and the number of data to be stored in the auxiliary search memory are set for each subgroup.