Database management system and method for query process for the same

• US 5,806,059 A
• Filed: 03/04/1997
• Issued: 09/08/1998
• Est. Priority Date: 01/20/1993
• Status: Expired due to Term

• Current Assignee: Hitachi, Ltd.
• Original Assignee: Hitachi, Ltd.
• Inventors: Kawamura, Nobuo, Tsuchida, Masashi, Torii, Shunichi, Nakano, Yukio, Negishi, Kazuyoshi
• Primary Examiner(s): Amsbury, Wayne
• Assistant Examiner(s): Alam, Hosain T.
• Application Number:
  US08/810,527
• Time in Patent Office: 553 Days
• Field of Search: 395/603, 395/602, 395/675, 395/653, 395/182.08, 395/200.31, 707/1-3, 707/2, 707/8, 707/10
• US Class Current: 1/1
• CPC Class Codes: G06F 16/24532 of parallel queries
• CPC Class Codes: G06F 16/252 between a Database Management System and a front-end application
• CPC Class Codes: Y10S 707/99932 Access augmentation or optimizing
• CPC Class Codes: Y10S 707/99933 Query processing, i.e. searching
• CPC Class Codes: Y10S 707/99937 Sorting

Abstract

A database management system for executing database operations in parallel by a plurality of nodes and a query processing method for it are described. The database management system contains a decision management node for deciding a distribution node for retrieving information so as to analyze a query received from an application program, generate a processing procedure for processing the query, and execute the process and a join node for sorting, merging, and joining the information retrieved by the distribution node. When the query process is executed, the distribution node decided by the decision management node retrieves the information to be processed and the join node decided by the decision management node also obtains the result for the query from the retrieved information. The query result is outputted from an output node and transferred to the application program.

Claims

1. A parallel database management system comprising:

at least one distribution node having storage means for distributing and storing a database to be queried and distribution means for retrieving information from said storage means and distributing the retrieved information to other nodes;

a plurality of join nodes each having its own first private storage means, sorting means for sorting information distributed from said distribution node, merge means for merging said plurality of sorted information, if any, and join means for joining a query on the basis of said merged information, using said first private storage means;

a decision management node having its own second private storage means, analysis means for receiving a query, analyzing said query, and generating a query processing procedure using said second private storage means, and decision means for deciding (i) the distribution node on the basis of the query analysis by said analysis means and calculating an expected processing time in said decided distribution node, and (ii) the join node on the basis of said processing time for performing the execution process;

an output node having output means of outputting the result for the query obtained from said join node; and

a network for transferring data between said nodes, wherein said plurality of nodes process independently and said join nodes input the information distributed from said distribution node successively and process each inputted information.

2. The database management system according to claim 1, wherein said decision means distributes said retrieval information equally to said join nodes on the basis of the expected retrieval information amount in said decided distribution node.

3. The database management system according to claim 2, wherein said decision means uses the hash function which is decided beforehand so as to assign the retrieval information equally to said join nodes.

4. The database management system according to claim 1, wherein to assign the retrieval information equally to said join nodes by said decision means, said decision management node has storage means of storing column value frequency information relating to the information stored in the storage means of said each node.

5. A database management system according to claim 1, wherein said distribution node further comprises sorting means for sorting information distributed by said distribution node.

6. The database management system according to claim 5, wherein from the calculation result of the expected processing time in said distribution node, said decision means decides so that the sorting means of said distribution node sorts information after the distribution process is executed for the distribution node which finishes the process earlier.

7. A database management system according to claim 1, wherein said decision means selects a larger plurality of join nodes than a number of said join nodes which are indicated on the basis of said processing time.

8. The database management system according to claim 7, wherein the sorting means of said join node has a function for executing the merge process after the sorting process is finished.

9. A database management system according to claim 1, wherein said join means has a function for executing said merge process.

10. The database management system according to claim 9, wherein said decision means calculated the expected processing times of said join means and said output means and when the processing time of said output means is greater than the processing time of said join means on the basis of said calculation results, decides so that said join means executes said merge process.

11. A query processing method in a parallel database management system having a plurality of processing nodes which are connected via a network, wherein each of said processing nodes has a storage unit in which a database is stored, said plurality of nodes having distribution nodes each having storage means for use in storing information for retrieving information from said storage means and for distributing and sorting the retrieved information to other nodes, join nodes, having first private storage means, for use in sorting information distributed from said distribution nodes, merging said sorted information, if any, and merging and joining a query on the basis of said merged information, using said first private storage means, and at least one decision management node, having second private storage means, for receiving said query, analyzing said query, and generating the query processing procedure, and for allocating information processing among the distribution nodes and join nodes on the basis of the query analysis result of said analysis means, the method comprising the steps of:

analyzing a query having a condition expression containing a variable in one of said plurality of nodes;

generating a plurality of processing procedure candidates to be executed in said plurality of nodes so as to process said query on the basis of said analysis result;

retrieving database information to be processed for said query from said plurality of nodes on the basis of the value which is substituted for said variable and the information, indicating the data storage status of said database given beforehand and selecting the distribution node for distributing said database information to other nodes and the join nodes for executing a database operation for the database information distribution by said distribution node;

deciding in said decision management node, the processing procedure to be executed by the selected distribution node and join node among said plurality of processing procedure candidates;

assigning a process to be executed in each of said distribution node and said join node to said distribution node, and said join node according to the decided processing procedure; and

obtaining a process result for said query by executing said assigned process in said distribution node and said join node, wherein said plurality of nodes process independently and said join nodes input the information distributed from said distribution node successively and process each inputted information.

12. The query processing method according to claim 11, wherein said selection step comprises the steps of:

selecting said distribution node on the basis of the information indicating the storing destination of the database information to be processed for said query;

calculating the expected processing time in said distribution node;

deciding the number of nodes to be selected as said join node on the basis of said expected processing time; and

selecting said join nodes on the basis of said decided number.

13. The query processing method according to claim 12, wherein said step for deciding the number of nodes decides said number of nodes so that the time required for the process which is to be executed in said join node in parallel with the process to be executed in said distribution node is equal to said expected processing time.

14. The query processing method according to claim 13, wherein said step for selecting join nodes selects nodes more than said decided number of nodes as said join nodes.

15. The query processing method according to claim 11, wherein said step for deciding the processing procedure contains a step for generating data distribution information for distributing data equally to said join nodes on the basis of said information indicating the data storage status.

16. The query processing method according to claim 11, wherein said step for assigning the process comprises the steps of:

judging whether there is a scattering in the expected processing time in said distribution node; and

assigning a part of the process of said database operation to a distribution node in which said expected processing time is short on the basis of said judgement result.