WO2014134757A1 - Rule engine of parallel business and implementation method therefor - Google Patents

Abstract

Provided are a rule engine of parallel business and an implementation method therefor, comprising: declaring four selection modes of selection, unification, competition and simultaneousness in a parallel executor of an engine; generating and finding a branch by the parallel executor in a mapping mode; then executing the branch; and returning a result. In the present invention, a parallel function is added on the basis of the existing rule engine; moreover, parallel computation can be quickly configured by an engine user; and the needs of backup, system integration, high-performance computation, etc. of the user are met. The technology is also applied to a parallel workflow engine.

Description

 Parallel business rule bow I engine and its implementation method

Technical field

 [0001] The present invention relates to the field of computer software development, and is applicable to system integration, software system configuration, fault tolerance, and high performance computing. Specifically, the design scheme provides an effective support for parallel computing by adding a parallel execution body in the rule engine, declaring the way the branch is selected in the parallel execution body, and generating and searching modes.

Background technique

 [0002] A regular rule refers to a rule in a business process that is represented by an easy-to-understand script. The script does not need to be compiled. It is read and interpreted by the rule engine framework during the runtime of the program. The application invokes the rule engine interface to trigger. Execute the corresponding rules to achieve separation of business logic and business programming implementation.

 [0003] The rule engine is different from the rule discovery engine, and the rule engine belongs to the program configuration system. The rule discovery engine refers to a certain rule in the business process through pattern matching, which belongs to the pattern recognition category.

 [0004] The rules engine includes the following parts: rule scripts (no compilation required), rules engine framework, applications that invoke the rules engine. The general rule script includes: an executable body and a rule body, and the rule body includes a trigger condition and a call to the execution body and the method to be executed.

 [0005] Workflow refers to a series of steps that are connected in succession. The workflow engine uses an easy-to-understand script or the like to represent the relationship between these steps. The workflow engine framework reads and interprets these steps. When used, the application invokes the corresponding workflow steps through the workflow engine interface. And get the result. The workflow engine separates the workflow and specific programming.

 [0006] Parallel computing (operation) in the present invention means that since data is distributed at different logical addresses, it is required to operate separately through the same or different subroutine modules, and the combination of these different subroutines or different logical addresses is called Branch, an operation (calculation) of data, needs to be selected from one or more possible branches, operations (or calculations) are spatially parallel, so called parallel computing.

 [0007] The business rule engine and the workflow engine basically realize the effective configuration of the variable parts of the system, but at present, there is no better solution engine that can conveniently and quickly realize the effective support for parallel computing such as system data backup and joint query. . Summary of the invention

 [0008] To fill the gap in the parallel computing of the business rules engine, the parallel business rules engine uses parallel executables in the configuration file. A parallel operation of data can be achieved by using different subroutines or using different logical addresses, and the applicant of the present invention refers to these different subroutines or a combination of different logical addresses as a branch.

 [0009] Parallel execution has two modes of operation on data: data read, and data write (including modification and deletion). Data reading includes branching, selection, competition, etc. Branch selection: Joint reading is to read data from a series of branches, and finally assemble into a result set; the choice is to assume that each branch data is the same, the execution body is from the branch (random Select a branch to read the result set; the competition is to assume that the data of each branch is the same, the executable reads from each branch, but only takes the fastest one, and the rest is discarded. Data writing includes joint and simultaneous branch selection: Union is to split the data set into rules according to rules, each branch gets incomplete data, but the sum of all branch data is complete data; The data is not broken up and is completely saved to each branch.

 [0010] The branch search and generation rules are as follows: The branch address or subroutine name is obtained by mapping the parameters passed in the engine, and the parameter value of the participating mapping may be "" or null. The mapping rule is that there are i variables, {3⁄4} is the set of these variables, the branch address or subroutine name is F({Xi}), then find a F({Xi}), through F(li) Obtain some or all of the required parts in {Xi} by inverse mapping. The rule engine obtains support for multiple data reading and writing modes by using variable mapping rules instead of selecting branches by conditional judgment, which is a feature of the present invention.

 [0011] Parallel computing is implemented by parallel execution. There are three types of parallel execution parameters: constants, variables M that are passed in by the application when the rule engine framework is called, and parameters that are mapped from constant M and other parameters. There are two ways to implement a method or a body reference: a constant method, a method that maps from constants and other parameters. The second type of parallel execution body maps the logical address branch, and the second reference method of the parallel execution body method or parallel execution body maps the subroutine branch.

 [0012] Since the parallel execution body is through the mapping instead of the conditional decision selection, SIlH, the scheme of the present invention can be applied to the workflow engine.

Correction page (Article 91) Specifically: the parallel execution body is an execution node of the workflow engine; I: the previous node of the flow engine or the MJ of the flow engine is passed in the corresponding parameter, and the node is called; Each branch, then return the final result to eleven nodes; if this is the last node of the engine, then return to the Tuning engine. It is necessary to explain that: JBPM and other workflow engines use forked nodes and merged nodes, and AND, OR, X0R means "simultaneous satisfaction", "only one" and "satisfy at least one". However, JBPM and the present invention belong to different technologies. Specifically, there are three points: 1. Data reading and data input include three steps of bifurcation, calculation and merging of branches, and JBPM's bifurcation node only represents points. Fork, the merged node of JBPM only represents the station, and the concept of branching in JBPM is different. 2. The AND, OR, and X0R of JBPM only indicate whether the branch node is executed or not, and does not represent the data. Whether the result of reading is to merge or take a single complete result, etc., does not mean that the written data is taken apart or completely saved; 3. The fork node and the merge node used by the workflow engine such as JBPM can only pass The path between the forked node and the merged node is configured with |*|, a path between the forked node and the merged node represents a branch, and a 0 to the running period cannot increase or decrease these branches. Ming parallel execution body is generated dynamically by mapping mode and (to) is configured to any number of branches, it is very flexible. DRAWINGS

 [0013] FIG. 1 is an explanation of the structure of a parallel business rules engine. The parallel business rules engine needs to configure two components: the executable body and the rule body are owned by the general business rules engine, and the parallel rule body is unique to the solution. Parallel rule body parameters and "method into executable body reference" can be mapped separately. By this combination of branch mapping, the parallel business rule engine can prepare for good scalability.

 [0014] FIG. 2 is an illustration of the application of a parallel rule body in a 1:1 flow engine. I: In the streaming engine, the previous node of the parallel node (such as 3⁄4) is used by the application to call the parallel point through the engine framework: the parallel execution body implicitly performs the branching by mapping; then, the parallel execution body obtains the parallel operation The end result; and the final result and control to the next node (if any), if there is no next node, then the result is passed to the HJ program that invokes the engine through the engine framework. detailed description

 [0015] The implementation of the business rules engine 冇 the following steps:

 1. The rules engine needs to be interpreted by the rules engine framework, and the program is read into the rules engine framework.

 2. Configure the business rules in the configuration file, the application reads in the configuration files, and interprets the rules through the framework.

 3. The application passes the call to the framework SPI in the corresponding place, and passes the corresponding parameters, executes the rules, and finally obtains the result.

4. If the program changes the requirements in the corresponding place, and the parameters are the same, the programmer can achieve the effect by re-distributing the S business rules.

 [0016] Parallel Business Rule Engine and Parallel Service I: The configuration process of the flow engine is as follows:

 1. First define the method of the program that needs to be called the branch of MJ. If the method of the branch program is different, the method is further encapsulated as an executable.

 2. Define a parallel execution body, parallel execution body Luming for reading or writing types, and Luming joint, select, compete, and simultaneously.

3. Configure the incoming parameters, references to "methods and executables" in the parallel execution body. The type of the parameter and the "method into the execution body" can be generated by the mapping, and the result of the mapping can be used as a parameter; or as the name of the execution body defined in step 1, the parameter is further obtained by the execution body; the implementation is defined as the execution of step 1. The name of the body is quoted by "method or executive body".

 4. The application adjusts the ffl engine framework, executes the parallel execution body, and performs parallel operations.

 [0017] The present invention will be further clarified below in conjunction with the application scenario of the body. The present invention can be applied to T but not limited to the following situations:

 [0018] It is assumed that the log system needs to generate one log file per large, and can log the log of the day.

 Developers can follow the steps to develop and configure:

1- Compile a method for passing HJ in the program, and use the Ding Day to generate a log file if there is no corresponding diary file. 2. Configure parallel execution of the body 1, D - parallel operation of the log file. The "Method into Executable Reference" portion of the parallel execution body is configured to direct the HJ of the method of step 1, and the incoming parameters are consistent with the parameters in step 1.

 3. Parallel Execution 1 is a write type and is declared as a simultaneous mode to accommodate the need for parallel writes.

 4. The log file name parameter in parallel execution body 1 is configured in a mapping manner. The mapping formula is F (X) = Log | draw DD, the brain is month, DD is day; MM and DD are respectively passed as a given parameter, indicating The current time, therefore, only one identical branch is executed for each write operation of parallel execution body 1.

 5. Compile a generic method in the program and read the log with Ding.

 6. Configure parallel execution body 2 to perform parallel operations on log files. The "Method or Execution Reference" section of the parallel execution body is configured as a reference to the method of step 1, and the incoming parameters are consistent with the parameters in step 1.

 7. The log file name parameter in parallel execution body 2 is configured in a mapping manner. The mapping formula is F (X) = Log | 丽 DD, MM is month, and DD is 曰.

 8. Parallel Execution 2 is a read type and is declared as a union to accommodate the need to jointly query data from each branch.

[0019] Assume that the application system needs to access two databases. The two databases have the same data, one for backup and one for business access: the two data sequences have different addresses.

 Developers can follow the steps to develop and configure:

 1- In the engine script, first configure two executables, returning the addresses of the two data ports respectively. Now assume that the two executables are named DB0001 and DB0002;

 2. Compile a generic method in the program, and use D to read the two databases. The parameters of the method should include the database address.

 3. Configure parallel execution body 1 for parallel read operations on both databases. The method execution part of the parallel execution body is configured as a reference to the method of step 2, and the incoming parameters are consistent with the parameters in step 2.

 4. The database address parameter in parallel rule body 1 is configured as a reference to the executable DB000X, and the referenced executable name is derived from the mapping F (X) = DB | 000X (as in step 1).

 5. Parallel Execution 1 Select to read the class and declare it as a selection method to accommodate the need to read a branch of data from the backup.

6. Compile a pass ffl method in the program to write to the two data files. The parameters of the method should contain the database address.

 7. Configure parallel execution 2 to perform parallel write operations on both databases. The "Method into Execution" section of the parallel execution body is configured to reference the method of step 2, and the parameters in the incoming parameter step 6 are consistent.

 8. The data/ΐ address parameter in the parallel rule body 2 is configured to be the source of the executable DB000X, and the referenced executable name ^ is derived from the mapping F (X) = DB | 000X (as in step 1).

 9. The parallel execution body 2 is selected as the write type, and Lu Ming is the simultaneous mode to accommodate the need to read a branch data from the backup.

[0020] The ticketing system buys tickets from different companies respectively. Assume that N companies, each company system purchase interface interface parameter format and return data are the same (or can be converted into the same), but the specific implementation method is different, The ticket-sending system needs to get the first ticket purchased. If one ticket is grabbed, the remaining tickets will be discarded.

 Developers can follow the steps to develop and configure:

 1. Configure N executables in the rule engine script. The parameters are the interface parameters of each company system. The implementation methods are the implementation methods of each company system. The executable names are ZXT1 to ZXTN.

2. Configure the parallel execution, J Τ · the parallel purchase operation of this family. The parameter formats of D-these companies are the same. In this case, the parameters of the parallel execution body are the same as those of step 1. The "method or executable reference" part of the executable is configured as a reference to the N executables. The name of the executable of the HJ is derived from the mapping F(X)=ZXT|N (see step 1).

Parallel Execution Body Selection Rules Lu Ming is a competitive reading method to adapt to competition from various branches? Oh, so you get the fastest ticket demand.

Claims

Claims  [oooi] The rule engine refers to an easy-to-follow code that represents the rules of a business. It does not need to be compiled and read and executed during the running of the program. The existing business rules engine only solves the requirement of "what under what conditions". The characteristics of the parallel business rules engine and its scheme are: Parallel business rules engine according to the needs of the business, the MJ parallel execution body obtains all by mapping Branching, and performing selective operations on the data in the branch, selecting methods, combining, selecting, competing, and simultaneously, so as to conveniently meet the needs of various parallel computing. [0002] A parallel business rule engine and scheme according to claim [0001], characterized in that _{: the} parallel business rule engine uses a variable mapping method to find and generate different branches, which is different from the traditional pass condition. The judgment statement is a conditional triggering statement to find and generate a branch; the mapping rule is as follows, there are i variables, {Xi} is the set of these changes, the branch address is the sub-order name {Y}, then the mountain is a {Y} Mapping of =F({Xi}), mapping {Xi} to F({Xi}) by computing the map.  [0003] According to the claim [0001], the business rule engine and scheme are characterized in that: and the execution body includes two types: 卖 sell and ^, through Lu Ming simultaneous, joint, competition Selecting and selecting the selection method to adapt to the parallel operation requirements under different scenarios. [0004] The parallel business rule engine and scheme according to claim [0001], characterized in that: the scheme can be used in the parallel parallel I: flow engine, and the engine parallelizes the operation of the parallel node For parallel operation, it is convenient to obtain good performance; work. Workflow refers to a series of successive steps; I: The workflow engine makes HJ easy scripts represent these steps The relationship between I, the flow engine framework to read and unseed these steps, so that HJ, the program through the I: flow engine interface to adjust the corresponding I: flow steps, and get the results; I: Program programming for processes and processes.