JP4872702B2 - Distributed workflow simulation system, method, and program - Google Patents

Description

  The present invention relates to a distributed workflow simulation system, method, and program. More specifically, the present invention relates to a distributed workflow simulation system, method, and program for simulating a distributed workflow system that performs a business process flow in a workflow execution device.

  As a business system, there is a distributed workflow system in which a flow of business processing is performed by a plurality of workflow execution devices. Although the usage status of business systems in business constantly changes, business systems are required to always maintain high performance. In order to avoid unexpected situations and maintain high performance in a business system, it is necessary to correctly predict the performance in the future usage situation, and to control and reconstruct the business system based on the prediction. For this reason, highly accurate performance prediction is required.

  In the distributed workflow system, partial processing within the business, for example, billing processing and ordering processing are provided by “services” operating on the server machine. The workflow execution apparatus realizes the entire business by linking services according to the flow of business processing. A plurality of distributed workflow execution devices may cooperate to realize one business. The workflow execution device and the server machine (service) are each connected to a network and can access each other. FIG. 15 shows an example of a business system. As shown in the figure, a plurality of services may operate on one server machine. Communication between the service and the workflow execution apparatus is performed using messages, and the service and the workflow execution apparatus perform parallel processing of messages.

  An example of a conventional workflow simulation system is disclosed in Patent Document 1. FIG. 16 shows the configuration of the distributed workflow simulation system described in Patent Document 1. A conventional distributed workflow simulation system 20 includes a network simulation model creating means 21, a diagram creating 22 concerning a business process, a traffic foundation model creating means 23, a database device 24, a business process traffic model creating means 25, and a network simulation system. Execution means 26.

  The conventional distributed workflow simulation system having the above configuration operates as follows. That is, the traffic foundation model creating unit 23 creates a diagram of the interaction between the node devices corresponding to the business process object. The business process traffic model creation means 25 associates traffic information with each link of the diagram created by the traffic foundation model creation means 23 and creates a network including the traffic information in the business process. Traffic information generated on the network system is set in the network created by the business process traffic model creation means 25, and this is simulated by the network simulation system execution means 26. In Patent Document 1, it is assumed that a highly accurate distributed workflow simulation can be realized by modeling the entire traffic flowing on the distributed workflow system.

Japanese Patent Laid-Open No. 2003-178103 (FIG. 1)

  Here, as a simulation method, there is a method for predicting performance by abstracting a message flow in a workflow using a model such as Petri Net and estimating a message behavior using this model. FIG. 17 shows a specific example of the simulation model. In FIG. 17, the portion represented by a rectangle is called a transition and indicates a service. Further, a portion represented by a circle is called a place, and a stay message in the service is shown for each input route. A black circle in the place indicates a message.

  Considering the processing time in the service, since the service performs parallel processing, if the number of staying messages in the service increases, the processing capacity of the server machine is shared correspondingly, and the processing time is extended. In the server machine, there is a correlation between the number of stay messages and the processing time, and high-precision performance prediction is assumed by reproducing the number of stay messages faithfully to the actual operation during simulation. However, the prior art does not consider the timing at which the message processing in the service overlaps, so there is a problem that the prediction accuracy is lowered because the change in the processing time due to the change in the number of staying messages is not considered.

  Further, when the service is processed, even when the number of stay messages is the same, the processing time is not constant, and the fluctuation occurs with a certain fluctuation range. However, in the prior art, only the average value of the service processing time is considered and the fluctuation of the service processing time is not considered, so that the prediction accuracy of the service processing time is low. In addition, since the number of stay messages changes each time a message arrives at the service, it is necessary to calculate the processing time of all stay messages and their fluctuations in order to improve the prediction accuracy. Is also not considered.

  In the business system, processing such as message analysis, destination determination, and path control is performed on the message from the user in the workflow execution device, and then the message is sent from the workflow execution device to the service. At this time, in the workflow execution apparatus, a delay occurs due to message analysis or the like. Since the workflow execution device processes messages in parallel, the delay time in the workflow execution device is considered to be proportional to the number of stay messages in the workflow execution device. However, the prior art does not assume the change in the delay time, and there is a problem that the prediction accuracy is lowered due to the delay time change in the workflow execution apparatus.

  In a business system, multiple services may be operated on the same server machine. In such a case, the multiple services share the processing capacity of the server machine. When the number of stay messages in the shared service is large, the processing time is extended accordingly. However, in the prior art, the processing time of a service is predicted based on the number of stay messages for the service, and the influence of server machine sharing is not added to the processing time prediction. Therefore, when the same server machine provides a plurality of services, there is a problem that the prediction accuracy of the service processing time is lowered.

  The present invention provides a distributed workflow simulation system, method, and program capable of solving the above-described problems of the prior art and realizing a simulation considering the influence of resource sharing.

  Another object of the present invention is to provide a distributed workflow simulation system, method, and program capable of reproducing the workflow behavior of the timing at which message processing in a service overlaps.

  Another object of the present invention is to provide a distributed workflow simulation system, method, and program capable of realizing simulation in consideration of the influence of delay time variation in the workflow execution apparatus.

  To achieve the above object, a distributed workflow simulation system of the present invention is a distributed workflow simulation system that simulates a business system including a workflow execution device that controls message delivery and a server machine that executes at least one service according to a received message. In the workflow simulation system, for a service that acquires the number of stay messages in the service and the message processing time of the service and shares the server machine with other services, the number of stay messages in the service, the service, and the server machine Service processing time prediction formula generating means for formulating a relationship between the number of stay messages in another service sharing the service and the processing time of the message in the service, and generating the service processing time prediction formula; A simulation model generating means for generating a simulation model based on a workflow definition for defining a message delivery route in the system, and the service processing time prediction formula; and a message in the business system based on the simulation model. For the simulation means that sequentially reproduces the behavior and the service that shares the server machine with the other service, the number of stay messages in the service reproduced by the simulation means, and the service and the server machine are shared Service processing time calculation means for calculating the service processing time based on the number of stay messages in the other service and the service processing time prediction formula.

  In the distributed workflow simulation system of the present invention, when a plurality of services are executed on the server, the number of stay messages in the service collected from the business system and the stay messages in other services sharing the service and the server machine The service processing time prediction formula is generated based on the number and the service processing time, and the simulation model is generated based on the workflow definition and the service processing time prediction formula to simulate the behavior of the message in the business system. To do. At that time, for services that share the server machine with other services, the generated service processing time prediction formula, the number of stay messages in the service in the simulation, and the number of stay messages in the service that shares the service and server machine Based on the above, the service processing time is calculated. When a server machine executes multiple services, the service processing time in one service is affected by the processing in other services that share the server machine (resource), so only the number of stay messages in that service. Then it is not found. In the present invention, a service processing time prediction formula is generated from the relationship between the number of stay messages in the service, the number of stay messages in other services sharing the server machine, and the service processing time, and the service processing time is used for this. Therefore, in the simulation, the service processing time can be obtained in a form close to the actual service processing time. Therefore, it is possible to increase the accuracy of the simulation and accurately predict the service processing time.

  In the distributed workflow simulation system of the present invention, the service processing time prediction formula generation means, when one service is executed on one server machine, the number of stay messages in the service and the number of messages in the service Formulating a relationship with the processing time to generate a service processing time prediction formula, the service processing time calculation means includes the number of stay messages in the service reproduced by the simulation means, and the service processing time prediction formula The service processing time can be calculated based on the above. When one service is executed on one server machine, a service processing time prediction formula is obtained from the relationship between the number of staying messages in the service and the service processing time, and simulation is performed using this formula. Thus, the service processing time may be calculated.

  In the distributed workflow simulation system of the present invention, the service processing time prediction formula generation means refers to a message log storage device that stores measurement information about messages that are input to and output from the service, and the number of stay messages in the service The message processing time can be acquired. As the measurement information, for example, a message log with a time stamp including a message ID, a business ID, a related message ID, and the like can be considered.

  In the distributed workflow simulation system of the present invention, the service processing time calculation means, when the number of stay messages in the service changes after the calculation of the service processing time, the number of stay messages after the change and the service processing time A configuration in which the service processing time is recalculated based on the prediction formula can be adopted. In this case, by recalculating the service processing time according to the change in the number of staying messages in the service, the service processing time can be calculated in a more realistic manner in the simulation, and the behavior of the message can be further improved. Can be faithfully reproduced.

  In the distributed workflow simulation system of the present invention, the service processing time prediction formula generating means can employ a configuration for estimating a probability distribution of service processing time. In this case, the service processing time calculation means may employ a configuration that calculates a service processing time considering variation based on the calculated service processing time and the probability distribution of the service processing time. In an actual business system, the service processing time is not uniquely determined from the number of stay messages in the service, but the service processing time often fluctuates even if the number of stay messages is the same. Therefore, it is estimated what probability distribution the actual service processing time is distributed with respect to the service processing time obtained by the service processing time prediction formula, and using this, the service processing time prediction formula is used. To the service processing time obtained. By doing so, in the simulation, the service processing time can be calculated in a more realistic manner, and the behavior of the message can be reproduced more faithfully.

  The distributed workflow simulation system of the present invention acquires the number of stay messages in the workflow execution device and the delay time of the message by the workflow execution device, and the number of stay messages in the workflow execution device and the workflow execution device A delay time prediction formula generating unit that formulates a relationship with a message delay time to generate a workflow delay time prediction formula, a number of stay messages of the workflow execution device reproduced by the simulation unit, and the workflow delay time A configuration further comprising workflow delay time calculation means for calculating the delay time of the message by the workflow execution device based on a prediction formula can be adopted. When message delivery control is performed in the workflow execution device, the delay time from when the message is input to the workflow execution device until the message delivery destination is determined and the message is output to the next service or workflow execution device Is considered to change according to the number of stay messages in the workflow execution device. Therefore, a workflow delay time prediction formula is generated from the relationship between the number of stay messages in the workflow execution device and the delay time by the workflow execution device. In the simulation, the delay time by the workflow execution device is generated using the workflow delay time prediction formula. Calculate the impact of By doing so, it is possible to reproduce the behavior of the message in a more realistic form, and to increase the accuracy of the simulation.

  In the distributed workflow simulation system of the present invention, the simulation model generation means is based on a workflow definition that defines a message delivery route in the business system, the service processing time prediction formula, and the workflow delay time prediction formula. A configuration for generating a simulation model can be employed. In the simulation, the service processing time is calculated using the service processing time prediction formula, and the workflow delay time is calculated using the workflow delay time prediction formula, whereby the behavior of the message can be reproduced in a more realistic manner.

  In the distributed workflow simulation system of the present invention, the delay time prediction formula generation means refers to a message log storage device that stores measurement information about messages input to and output from the workflow execution device, and the stay message of the workflow execution device A configuration for acquiring the number and the delay time of the message can be employed.

  In the distributed workflow simulation system of the present invention, when the number of staying messages in the workflow execution device changes after the calculation of the delay time of the message, the workflow delay time calculating means calculates the number of staying messages after the change, and the workflow A configuration in which the delay time of the message is recalculated based on the delay time prediction formula can be adopted. In this case, by recalculating the workflow delay time according to the change in the number of staying messages in the workflow execution device, the workflow delay time can be calculated in a more realistic manner in the simulation, and the message behavior Can be reproduced more faithfully.

  In the distributed workflow simulation system of the present invention, the simulating means can adopt a configuration that reproduces the behavior of the message in the business system based on the calculated service processing time and message delay time.

  A business system management apparatus according to the present invention is a business system management apparatus that manages a business system including a workflow execution device that performs message delivery control and a server machine that executes at least one service in accordance with a received message. For a service that acquires the number of stay messages in the service and the message processing time of the service and shares the server machine with other services, the number of stay messages in the service and other services that share the service and server machines A service processing time prediction formula generating means for formulating a relationship between the number of stay messages in the service and a message processing time in the service to generate a service processing time prediction formula, and defining a delivery route of the message in the business system Workflow definition and the service process Based on the time prediction expression, characterized in that it comprises a simulation model generating means for generating a simulation model.

  In the business system management apparatus according to the present invention, the service processing time prediction formula generation means, when one service is executed on one server machine, the number of stay messages in the service and the number of messages in the service A configuration that formulates a relationship with the processing time and generates a service processing time prediction formula can be employed.

  The business system management device of the present invention acquires the number of stay messages in the workflow execution device and the delay time of the message by the workflow execution device, and the number of stay messages in the workflow execution device and the workflow execution device It is possible to employ a configuration further including a delay time prediction expression generation unit that generates a workflow delay time prediction expression by formulating the relationship with the message delay time.

  In the business system management device of the present invention, the simulation model generation means is based on a workflow definition that defines a message delivery route in the business system, the service processing time prediction formula, and the workflow delay time prediction formula, A configuration for generating a simulation model can be employed.

  A simulation apparatus according to the present invention is a simulation apparatus for simulating a business system including a workflow execution apparatus that performs message delivery control and a server machine that executes at least one service according to a received message, For the service sharing the server machine, formulate the relationship between the number of stay messages in the service, the number of stay messages in other services sharing the service and the server machine, and the message processing time in the service. Based on the simulation model generated based on the service processing time prediction formula generated in this way and the workflow definition that defines the delivery route of the message in the business system, the behavior of the message in the business system is sequentially determined. In For the present simulating means and the service sharing the server machine with the other service, the number of stay messages in the service reproduced by the simulating means, and the other service sharing the server machine with the service Service processing time calculation means for calculating the service processing time based on the number of stay messages and the service processing time prediction formula.

  In the simulation apparatus of the present invention, the service processing time prediction formula is obtained by calculating the number of stay messages in the service and the message processing time in the service when one service is executed on one server machine. The service processing time calculation means is generated by formulating a relationship, and the service processing time calculation means is based on the number of stay messages in the service reproduced by the simulation means and the service processing time prediction expression. A configuration for calculating the processing time can be employed.

  In the simulation apparatus according to the present invention, when the number of stay messages in the service changes after the calculation of the service processing time, the service processing time calculation means calculates the number of stay messages after the change and the service processing time prediction formula Based on the above, a configuration for recalculating the service processing time can be adopted.

  In the simulation apparatus of the present invention, the service processing time calculation means adopts a configuration that calculates a service processing time considering variation based on the calculated service processing time and a probability distribution of the estimated service processing time. it can.

  In the simulation apparatus of the present invention, the simulation model is a workflow delay time prediction formula generated by formulating the relationship between the number of staying messages in the workflow execution device and the delay time of the message by the workflow execution device. And the workflow delay time prediction formula and the workflow definition, and based on the workflow delay time prediction formula and the workflow delay time prediction formula reproduced by the simulation means and the workflow execution device Further, it is possible to adopt a configuration further comprising a workflow delay time calculation means for calculating the delay time of the message by the workflow execution device.

  In the simulation apparatus of the present invention, when the number of stay messages in the workflow execution apparatus changes after calculating the delay time of the message, the workflow delay time calculation means calculates the number of stay messages after the change and the workflow delay time. A configuration in which the delay time of the message is recalculated based on the prediction formula can be adopted.

  The distributed workflow simulation method of the present invention is a distributed workflow simulation method for simulating a business system including a workflow execution device that performs message delivery control using a computer and a server machine that executes at least one service according to a received message. In the service, the computer acquires the number of stay messages in the service and the message processing time of the service, and for the service that shares the server machine with other services, the number of stay messages in the service, the service and the server Formulating a relationship between the number of stay messages in another service sharing a machine and the message processing time in the service to generate a service processing time prediction formula; and A step of generating a simulation model based on a workflow definition for defining a message delivery route in the business system and the service processing time prediction formula; and the computer in the business system based on the simulation model. Sequentially reproducing the message behavior of the message, and for the service in which the computer shares the server machine with the other service, the number of stay messages in the service reproduced by reproducing the message behavior, and And calculating the service processing time based on the number of stay messages in another service sharing the service and the server machine and the service processing time prediction formula.

  The distributed workflow simulation method of the present invention formulates a relationship between the number of stay messages in the service and the message processing time in the service when the computer executes one service on one server machine. Generating a service processing time prediction formula and the number of stay messages in the service reproduced by reproducing the message behavior when the computer executes one service on one server machine And a step of calculating the service processing time based on the service processing time prediction formula can be adopted.

  In the distributed workflow simulation method of the present invention, in the step of obtaining the number of staying messages in the service and the processing time of the message, the computer stores a message log storage device that stores measurement information about a message that inputs and outputs the service Referring to FIG. 4, it is possible to adopt a configuration for acquiring the number of staying messages in the service and the processing time of the message.

  In the distributed workflow simulation method of the present invention, when the number of stay messages in the service changes after the calculation of the service processing time, the computer uses the number of stay messages after the change and the service processing time prediction formula. Based on this, it is possible to adopt a configuration for recalculating the service processing time.

  The distributed workflow simulation method of the present invention may employ a configuration in which the computer further includes a step of estimating a probability distribution of service processing time.

  In the distributed workflow simulation method according to the present invention, in the step of calculating the service processing time, the computer is configured to consider a variation based on the calculated service processing time and the probability distribution of the service processing time. The structure which calculates can be employ | adopted.

  In the distributed workflow simulation method of the present invention, the computer obtains the number of stay messages in the workflow execution device and the delay time of the message by the workflow execution device, the number of stay messages in the workflow execution device, and the workflow Formulating a relationship with the delay time of the message by the execution device to generate a workflow delay time prediction formula, and the computer, the number of stay messages of the workflow execution device reproduced by reproducing the message behavior, It is possible to adopt a configuration further comprising a step of calculating a delay time of the message by the workflow execution device based on the workflow delay time prediction formula.

  In the distributed workflow simulation method of the present invention, in the step of generating the simulation model, the computer defines a workflow definition that defines a message delivery route in the business system, the service processing time prediction formula, and the workflow delay time prediction. Based on the formula, it is possible to employ a configuration for generating a simulation model related to workflow delay.

  In the distributed workflow simulation method of the present invention, in the step of acquiring the number of stay messages of the workflow execution device and the delay time of the message, the computer stores measurement information about messages input to and output from the workflow execution device. A configuration in which the number of staying messages of the workflow execution device and the delay time of the message is acquired with reference to a log storage device can be adopted.

  In the distributed workflow simulation method of the present invention, when the number of stay messages in the workflow execution device changes after the delay time of the message is calculated, the computer determines the number of stay messages after the change and the workflow delay time prediction formula. Based on the above, it is possible to adopt a configuration for recalculating the delay time of the message.

  In the distributed workflow simulation method of the present invention, in the step of reproducing the behavior of the message, the computer reproduces the behavior of the message in the business system based on the calculated service processing time and message delay time. Configuration can be adopted.

  The program of the present invention is a program for causing a computer to execute a process for managing a business system including a workflow execution device that performs message delivery control and a server machine that executes at least one service according to a received message. For the service that acquires the number of stay messages in the service and the message processing time of the service in the computer and shares the server machine with other services, the number of stay messages in the service, the service and the server machine A process for formulating the relationship between the number of staying messages in another service to be shared and the message processing time in the service and generating a service processing time prediction formula, and a workflow for defining a message delivery route in the business system Definition and the service Based on the scan processing time prediction equation, wherein the executing the process of generating a simulation model.

  In the process of generating the service processing time prediction formula, the program of the present invention is configured to process the number of stay messages in the service and the message processing in the service when one service is executed on one server machine. It is possible to adopt a configuration that formulates a relationship with time and generates a service processing time prediction formula.

  The program of the present invention acquires the number of stay messages in the workflow execution device and the delay time of the message by the workflow execution device to the computer, and the number of stay messages in the workflow execution device and the workflow execution device It is possible to employ a configuration in which a process for formulating a relationship with the delay time of the message and generating a workflow delay time prediction formula is further executed.

  In the process of generating the simulation model, the program according to the present invention performs a simulation based on a workflow definition that defines a message delivery route in the business system, the service processing time prediction formula, and the workflow delay time prediction formula. A configuration for generating a model can be employed.

  The program of the present invention is a program for causing a computer to execute a process of simulating a business system including a workflow execution device that performs message delivery control and a server machine that executes at least one service according to a received message. For the service that shares the server machine with another service, the number of stay messages in the service, the number of stay messages in the other service that shares the service and the server machine, and the messages in the service Based on the simulation model generated based on the service processing time prediction formula generated by formulating the relationship with the processing time and the workflow definition defining the delivery route of the message in the business system, the business In the system For the process of sequentially reproducing the behavior of the message and the service sharing the server machine with the other service, the number of stay messages in the service reproduced by the process of reproducing the behavior of the message, and the service A process of calculating the service processing time based on the number of stay messages in another service sharing the server machine and the service processing time prediction formula is performed.

  In the program of the present invention, the service processing time prediction formula is the relationship between the number of stay messages in the service and the message processing time in the service when one service is executed on one server machine. In the process of calculating the service processing time, the number of stay messages in the service reproduced in the process of reproducing the behavior of the message, the service processing time prediction formula, The service processing time can be calculated based on the above.

  In the process of calculating the service processing time, when the number of stay messages in the service changes after the calculation of the service processing time, the program of the present invention calculates the number of stay messages after the change and the service processing time prediction. A configuration for recalculating the service processing time based on the formula can be adopted.

  The program of the present invention has a configuration in which, in the process of calculating the service processing time, the service processing time is calculated in consideration of fluctuations based on the calculated service processing time and the probability distribution of the estimated service processing time. Can be adopted.

  In the program of the present invention, the simulation model includes a workflow delay time prediction formula generated by formulating a relationship between the number of staying messages in the workflow execution device and the delay time of the message by the workflow execution device; The number of stay messages of the workflow execution apparatus, which is generated based on the workflow delay time prediction formula and the workflow definition, and is reproduced in the computer by the process of reproducing the behavior of the message, and the workflow It is possible to adopt a configuration that further executes a process of calculating the delay time of the message by the workflow execution device based on the delay time prediction formula.

  In the process of calculating the workflow delay time, the program of the present invention, when the number of stay messages in the workflow execution apparatus changes after the calculation of the delay time of the message, the number of stay messages after the change, and the workflow delay A configuration in which the delay time of the message is recalculated based on the time prediction formula can be adopted.

  In the distributed workflow simulation system, method, and program of the present invention, for services that share a server machine with other services, the number of stay messages in the service collected from the business system, and other services that share the service and server machine A service processing time prediction formula is generated based on the number of stay messages in the service and the service processing time. Further, a simulation model is generated based on the workflow definition and the service processing time prediction formula, and the behavior of the message in the business system is simulated. In the simulation, for the service sharing the server machine with other services, the generated service processing time prediction formula, the number of stay messages in the service in the simulation, and the number of stay messages in the service sharing the service and the server machine By calculating the service processing time based on the above, the service processing time can be obtained in a form close to the actual service processing time, and the service processing time can be accurately predicted.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. FIG. 1 shows the configuration of a distributed workflow simulation system according to an embodiment of the present invention. The distributed workflow simulation system 10 includes a server machine 100, a workflow execution device 200, a business system management device 400, and a simulation device 500. The server machine 100 and the workflow execution apparatus 200 constitute a distributed workflow system (business system). FIG. 2 shows the connection status of each device constituting the distributed workflow simulation system. As shown in FIG. 2, each device is connected to a network 700 and can communicate with each other via the network 700.

  The server machine 100 provides hardware equipment for executing services. The workflow execution apparatus 200 performs message delivery control. The business system management apparatus 400 generates a simulation model reflecting the operation status of the business system based on the message processing status in the server machine 100 and the workflow execution apparatus 200. The simulation apparatus 500 performs a simulation using the simulation model generated by the business system management apparatus 400 and outputs a result of reproducing the performance of the business system.

  Referring to FIG. 1, the server machine 100 includes a service providing unit 101 and a service monitoring unit 102. The service providing unit 101 processes the service according to the received message and returns a response message. This process is performed in parallel. The service monitoring unit 102 monitors the service providing unit 101 and acquires service measurement information such as a time-stamped message log and error log based on the message processing status in the service providing unit 101. The measurement information acquired by the service monitoring unit 102 is stored in the message log storage device 302. Further, the service monitoring unit 102 stores information indicating the correspondence between the service and the server machine in the server machine / service correspondence information storage device 303.

  The workflow execution device 200 includes a workflow execution unit 201 and a workflow monitoring unit 202. The workflow execution unit 201 performs message delivery path control according to the workflow definition stored in the workflow definition storage device 301 in accordance with the received message. The workflow monitoring unit 202 monitors the workflow execution unit 201 and acquires measurement information such as a time-stamped message log and error log based on the message processing status in the workflow execution unit 201. The measurement information acquired by the workflow monitoring unit 202 is stored in the message log storage device 302.

  The business system management apparatus 400 includes a message log collection unit 401, a service processing time prediction formula generation unit 402, a workflow delay prediction formula generation unit 403, and a simulation model generation unit 404. The message log collection unit 401 receives measurement information related to service processing in the server machine 100 and measurement information related to workflow processing in the workflow execution device 200 (message log) from the message log storage device 302 in the business system to be managed. To collect.

  The service processing time prediction formula generation unit 402 determines the number of staying messages in the service and the message processing time based on the measurement information related to the service processing, formulates these relationships, and generates a service processing time prediction formula. Further, the service processing time prediction formula generation unit 402 estimates a probability distribution for the service processing time. More specifically, for example, it is estimated according to what distribution the service processing time that is an actual sample point exists with respect to the service processing time obtained by the service processing time prediction formula.

  When generating the processing time prediction formula, the service processing time prediction formula generation unit 402 refers to the server machine / service correspondence information storage device 303 and confirms the correspondence between the service and the server machine. When one service is operating on one server machine 100, a service processing time prediction formula is generated based on the number of message stays in the service and the service processing time. When a plurality of services are operating on one server machine 100, the number of stay messages in other services that supply server machines (resources) is added as a parameter, and the number of stay messages in the service The service processing time prediction formula is generated based on the number of stay messages in the other service requesting the resource and the service processing time.

  The workflow delay prediction formula generation unit 403 determines the number of stay messages in the workflow execution device and the message processing time (delay time) based on the measurement information related to the workflow, formulates the relationship between them, and the workflow execution device 200 Generate a prediction formula for the delay time. The simulation model generation unit 404 includes a service processing time prediction formula generated by the service processing time prediction formula generation unit 402, a workflow delay time prediction formula generated by the workflow delay prediction formula generation unit 403, and a workflow in the workflow definition storage device 301. Integrate definitions and build a simulation model. The constructed simulation model is stored in the simulation model storage device 304.

  The simulation apparatus 500 includes a simulation unit 501, a workflow delay calculation unit 502, and a service processing time calculation unit 510. The simulation unit 501 retrieves the simulation model from the simulation model storage device 304, and sequentially reproduces the behavior of the message in the business system based on this and the simulation setting from the input device / property 601. The simulation setting includes, for example, a message arrival frequency setting and a message total number setting at the time of simulation.

  The simulating unit 501 calculates a workflow delay time using the workflow delay calculating unit 502, and calculates a service processing time using the service processing time calculating unit 510. The workflow delay calculation unit 502 receives the number of messages in the workflow execution device in the simulator as an input, and calculates a delay time between the workflow execution device and the service or between the two workflow execution devices. The service processing time calculation means 510 receives the number of stay messages in the service in the simulator and calculates the processing time in the service.

  The service processing time calculation unit 510 includes a service processing time distribution calculation unit 511 and a server machine shared service processing time calculation unit 512. The service processing time calculation unit 511 calculates an average processing time in the service using a service processing time prediction formula. At this time, when a plurality of services are executed on one server machine, the server machine shared service processing time calculation unit 512 uses the number of stay messages of other services sharing the resource, Calculate the average prediction time for.

  The service processing time distribution calculation means 511 gives a variation to the average service processing prediction time calculated using the service processing time prediction formula using the probability distribution of the service processing time. The service processing time of the message in the simulation is determined based on the average prediction processing time calculated using the service processing time prediction formula and its fluctuation. During the simulation, the number of stay messages in the service may change within the service processing waiting time. In that case, the calculation of the average prediction processing time and the calculation of the fluctuation are performed again, and the recalculated service processing time is set in the simulation unit 501.

  The workflow delay time calculated by the workflow delay calculation unit 502 and the service processing time considering the variation calculated by the service processing time calculation unit 510 are obtained when the simulation unit 501 sequentially reproduces the behavior of the message. Used for. The result of the simulation by the simulation unit 501 is output to the output device / GUI 602. Further, it is stored in the simulation result storage device 305.

  Derivation of the workflow delay prediction formula will be described. FIG. 3 shows an operation procedure for generating a workflow delay prediction formula. The workflow monitoring unit 202 acquires measurement information such as a message log with a time stamp and an error log from the workflow execution unit 201 and stores the measurement information in the message log storage device 302. The message log collection unit 401 of the business system management apparatus 400 extracts a message log related to the workflow from the message log storage unit 302 and passes it to the workflow delay prediction formula generation unit 403.

  The workflow delay prediction formula generation unit 403 takes out the workflow definition from the workflow definition storage device 301 and decomposes the workflow into partial workflows (step A1). Workflows can be classified into three categories: (1) from message reception to service call, (2) from service call to next service call, and (3) from service call to message output. Each decomposed partial workflow belongs to one of these three categories.

  The total number of stay messages in the same workflow execution device 200 affects the delay time in various partial workflows executed by the workflow execution device 200. The workflow delay prediction formula generation unit 403 includes a combination of (i) a delay time in each partial workflow and (ii) an average number of staying messages in the workflow execution apparatus 200 in the time range in the log of messages that have passed through the workflow. Is calculated (step A2). This process is performed for all messages. In calculating the average stay message count, for example, if the time from when a message is received by the workflow execution apparatus 200 until it is output to another workflow execution apparatus 200 is 60 seconds, the average stay message count is 60 seconds. The total number of stay messages of the workflow execution device 200 is measured at intervals of 1 second, and a value obtained by dividing the total number of stay messages by 60 seconds is defined as the average number of stay messages.

For each partial workflow, the workflow delay prediction formula generation unit 403 determines the relationship between the delay time in each partial workflow and the number of stay messages in the workflow based on the set of (i) and (ii) above. Formulate using regression analysis to generate a workflow delay prediction formula (step A3). Specifically, D is a partial workflow delay time, M is the number of staying messages, α and β are coefficients and constants by formulation,
D = αM + β
This is used as a workflow delay time prediction formula. As a formulation technique, a neural network or the like can be used. With this prediction formula, the delay time in each partial workflow can be calculated from the number of stay messages in the workflow. When the workflow delay prediction expression generation unit 403 generates the workflow delay prediction expression, the workflow delay prediction expression generation unit 403 passes the set of generated workflow delay prediction expressions to the simulation model generation unit 404 (step A4).

  Next, derivation of the service processing time prediction formula will be described. FIG. 4 shows a procedure for generating a service processing time prediction formula. The service monitoring unit 102 acquires measurement information such as a message log with a time stamp and an error log from the service providing unit 101 and stores the measurement information in the message log storage device 302. The message log collection unit 401 of the business system management apparatus 400 takes out the message log about the service from the message log storage unit 302 and passes it to the service processing time prediction formula generation unit 402.

  The service processing time prediction formula generation unit 402 starts a service processing time prediction formula generation process (step B1). In this process, first, for each service message log, a set of (p) service processing time and (q) average stay message count in the service in the time range is calculated (step B2). Next, with reference to the server machine / service correspondence information storage device 303, the correspondence relationship between the server machine and the service is checked, and it is determined whether or not the service shares the server machine 100 with other services (step B3). . When sharing with other services, the average number of stay messages in the shared service is calculated (step B4), as in (q) above.

The service processing time prediction formula generation unit 402 calculates the relationship between the processing in the service and the number of staying messages in the service based on the set of the processing time in the service calculated in step B2 and the average number of staying messages in the service. For example, it formulates using regression analysis and produces | generates a service processing time prediction formula (step B5). In step B5, specifically, T is the service processing time, M is the number of staying messages, γ and δ are the coefficients and constants from the formulation,
T = γM + δ
This is used as a service processing time prediction formula.

In step B5, if there is another service sharing the resource, add the average stay message count of other services sharing the resource calculated in step B4 to the average stay message count in the service, The relationship between the processing in the service and the number of stay messages is formulated using, for example, multiple regression analysis. Specifically, when two services of service 1 and service 2 are operating on one server machine 100, M ₁ is the number of stay messages in service 1, and M ₂ is a stay message in service 2. As a number
T = γ ₁ M ₁ + γ ₁ M ₁ + δ
This is used as a service processing time prediction formula.

  The service processing time prediction formula generation unit 402 obtains a probability distribution of service processing time when formulating the relationship between processing in the service and the number of staying messages in the service (step B6). In this probability distribution estimation, for example, how much each sample point is deviated with respect to the service processing time obtained using the service processing time prediction formula, the histogram is taken, and the distribution approximates the normal distribution. Whether the distribution is an approximate distribution or a Poisson distribution. From step B2 to step B6, a processing time prediction formula is obtained for one service. This process is performed for each service (step B7), and a processing time prediction formula is generated for each service. The service processing time prediction formula generation unit 402 passes the generated set of service processing time prediction formulas for each service to the simulation model generation unit 404 (step B8).

  The simulation model generation unit 404 extracts the workflow definition from the workflow definition storage device 301, and configures a simulation model by combining the workflow definition in the business system with the workflow delay prediction formula and the service processing time prediction formula. An example of the simulation model is shown in FIG. In the configuration of the simulation model, a service existing in the business system is determined from the workflow definition, and a service processing time prediction formula corresponding to the service is embedded in the simulation model. Further, the message path of the business system is determined from the workflow definition, and a delay time prediction formula for the partial workflow delay of the workflow execution apparatus 200 corresponding to the path is embedded in the simulation model.

  Next, calculation of workflow delay time in simulation will be described. The simulating means 501 takes out the simulation model from the simulation model storage device 304, and based on this simulation model and simulation settings such as message arrival frequency setting and message total number setting input from the input device / property 601. Then, the behavior of messages in the business system is reproduced sequentially. By reproducing this message behavior, the number of stay messages in the workflow and the number of stay messages in each service are obtained.

  FIG. 6 shows a procedure for calculating the workflow delay time. The simulation means 501 performs a determination process for the number of stays of messages in the workflow when a message arrives in the workflow or at regular time intervals in the simulation (step C1). In this process, the simulation unit 501 determines whether there is a change in the number of staying messages in the workflow (step C2). When there is a change in the number of stay messages, the workflow delay calculation means 502 recalculates the delay time for each partial workflow based on the new number of stay messages and the predicted partial workflow delay value (step C3). The delay time for each partial workflow recalculated by the workflow delay calculation means 502 is set in the simulation means 501 and used for the simulation (step C4).

  The simulating unit 501 rotates a loop when a message arrives in the workflow or at regular time intervals (step C5), and when there is a change in the number of staying messages, a delay time is calculated accordingly. By doing so, the delay time of the workflow execution device in the simulation changes according to the change in the number of stay messages in the workflow. Each time a message is input to the workflow execution device in the simulation, the simulation unit 501 delays the message by a set delay time and passes the message to the next server machine or workflow execution device, so that the behavior of the message is reached. To reproduce.

  FIG. 7 shows a procedure for calculating the service processing time. The simulation unit 501 performs a process for determining the number of stays in the service when the message arrives at the service or when the stay message in the service is finished (step D1). The simulating means 501 determines whether or not there is a change in the number of stay messages in the service (step D2). If there is a change, it is determined whether the service shares the server machine 100 with other services (step D3). When the service does not share the server machine 100 with other services, the service processing time calculation unit 510 resets the average processing time of the service based on the number of stay messages after the change and the service processing time prediction formula. Calculate (step D6).

  If a service shares a server machine with another service, the change in the number of stay messages is also notified to the other shared service, and the processing time of the other service is also changed (step) D4). Also, the number of stay messages in other shared services is acquired (step D5). Thereafter, the process proceeds to step D6, where the server machine shared service processing time calculation means 512 is used to determine the number of stay messages in the service, the number of stay messages in other services requesting resources, and the service processing time. Based on this, the service average processing time is recalculated.

  The service processing time calculated in step D6 is an average value of the service processing time when the processing time variation is not taken into consideration. When the service processing time calculation unit 510 recalculates the service average processing time, the service processing time variance calculation unit 511 uses the probability distribution of the service processing time obtained by the service processing time prediction formula generation unit 402 to consider variation. The processing time of the service is calculated (step D7). The service processing time considering the variation calculated by the service processing time distribution calculating unit 511 is set in the simulating unit 501 and used for the simulation (step D8).

  The simulation unit 501 rotates a loop when a message arrives in the workflow or every time a stay message in the service is processed (step D9). When there is a change in the number of stay messages, a process corresponding thereto is performed. Perform time calculation. By doing in this way, the service processing time calculated by simulation changes according to the change in the number of stay messages in the service or in the service and the service that supplies the service and the server machine. Each time a message is input to the workflow execution device in the simulation, the simulation unit 501 delays the message by a set delay time and passes the message to the next server machine or workflow execution device, so that the behavior of the message is reached. To reproduce.

  When the workflow delay time or service processing time (message stay time) changes during the message stay, the simulating means 501 changes the time from when the message is input until the calculated message stay time changes. The degree of progress of the process is calculated from the relationship with the previous message stay time, and the remaining message stay time is calculated from the message stay time after the change and the remaining ratio of the process. For example, when the message stay time is initially calculated as “20 seconds” and the message stay time is changed to “30 seconds” as the number of stay messages increases 10 seconds after the message is input, 10 seconds / 20 seconds = 50 It is judged that processing is progressing by%. In this case, in the simulation, the remaining message stay time is 30 seconds × 50% = 15 seconds after the change.

  Hereinafter, a specific example will be described. FIG. 8 shows the configuration of the business system. This business system is a supply chain business system, and is composed of a retail service (Retailer), a wholesale service (Warehouse A, B), and a manufacturing service (Manufacture 1, 2). FIG. 9 shows an apparatus configuration for realizing the business system of FIG. In this example, the business system includes two workflow execution devices (200A, 200B) and four server machines (100A to 100D). It is assumed that messages in XML format are transmitted and received between services using a protocol such as SOAP, and the service and the workflow execution device can perform parallel processing of messages.

  The workflow execution apparatus 200A receives a message from the business system user and passes it to the server machine 100A. The server machine 100A performs processing by the service providing unit 101A corresponding to the Retailer service, and then returns the processing result to the workflow execution apparatus 200A. The workflow execution apparatus 200A passes the processing result in the server machine 100A to the server machine 100B and the workflow execution apparatus 200B. The server machine 100B performs processing by the service providing unit 101B corresponding to the Warehouse A service, and then returns the processing result to the workflow execution apparatus 200A. The workflow execution apparatus 200A passes the processing result in the server machine 100B to the workflow execution apparatus 200B.

  The workflow execution device 200B passes the processing result in the server machine 100A received from the workflow execution device 200A to the server machine 100C. The server machine 100C performs processing by the service providing unit 101C corresponding to the WarehouseB service, and then returns the processing result to the workflow execution apparatus 200B. The workflow execution device 200B passes the processing result returned from the server machine 100C and the processing result in the server machine 100B received from the workflow execution device 200A to the server machine 100D. The server machine 100D performs processing with the service providing unit 101D corresponding to Manufacture1 and the service providing unit 101E corresponding to Manufacture2, and returns the processing result to the workflow execution apparatus 200B. The workflow is completed when the workflow execution apparatus 200B outputs the processing result received from the server machine 100D.

  The workflow monitoring unit 202 (FIG. 1) of the workflow execution device 200 </ b> A acquires measurement information from the workflow execution unit 201 and stores it in the message log storage device 302. Similarly, for the workflow execution device 200 </ b> B, the workflow monitoring unit 202 acquires measurement information from the workflow execution unit 201 and stores it in the message log storage device 302. FIG. 10 shows a specific example of measurement information. In this example, the measurement information includes a message ID, a transmission source URL, a transmission destination URL, a related message ID, a Request / Response flag, a business ID, and a time stamp.

  The workflow delay prediction formula generation unit 403 in the business system management apparatus 400 takes out the workflow definition from the workflow definition storage device 301 and decomposes the workflow into partial workflows. For example, for the workflow that the workflow execution apparatus 200A is in charge of, the workflow is (1) from message input to Retailer service call, (2) From Retailer service response to WarehouseA service call, and (3) From WarehouseA service response to workflow execution apparatus 200B. Is broken down into three partial workflows:

  The workflow delay prediction formula generation unit 403 refers to the message log storage device 302 and calculates, for each partial workflow, the delay time of the partial workflow and the average number of staying messages in the workflow execution device within the delay time range. In calculating the delay time, referring to the measurement information stored in the message log storage device 302 (FIG. 10), the messages entering and exiting each workflow execution device 200 are tracked, and a message corresponding to the input message is output. Calculate the time until. In calculating the average number of stay messages, the stay messages in each workflow execution device 200 from when a message is input to when it is output are examined, and the average value is calculated.

  FIG. 11 shows a set of the calculated partial workflow delay time and the average stay message count. The workflow delay prediction formula generation unit 403 uses the set of the calculated partial workflow delay time and the average stay message number as a sample value, and formulates the relationship between the partial workflow delay time and the average stay message number by regression analysis. . By this formulation, a workflow delay prediction formula for estimating the delay time of the partial workflow is obtained using the number of message stay messages in the workflow execution apparatus 200 as a parameter.

  On the other hand, the service monitoring unit 102 (FIG. 1) of each server machine 100 acquires the measurement information (FIG. 10) from the service providing unit 101 and stores it in the message log storage device 302. The service processing time prediction formula generation unit 402 in the business system management device 400 refers to the message log storage device 302 and calculates the processing time and the average number of stay messages in the processing time range for each service. In the calculation of the processing time, the measurement information stored in the message log storage device 302 is referred to, the message that enters and exits the server machine 100 is traced, and the message corresponding to the input message is output for each service. Calculate time. In calculating the average number of stay messages, the stay messages in the service from when a message is input until it is output are examined, and the average value is calculated.

  FIG. 12 shows a set of calculated service processing time and average stay message count. The service processing time prediction formula generation unit 402 uses the set of the calculated service processing time and the average stay message number as a sample, and formulates the relationship between the service processing time and the average stay message number by regression analysis. With this formulation, a service processing time prediction formula for estimating the service processing time is obtained using the number of stay messages in the service as a parameter. Further, the service processing time prediction formula generation unit 402 applies Bayesian estimation or the like to the service processing time set to estimate the probability distribution of the service processing time.

  As for the server machine 100D, since two services of Manufacture1 and Manufacture2 are performed on the same server machine, not only the average number of stay messages in its own service but also other servers that share the server machine A service processing time prediction formula is generated by adding the average number of stay messages in the service. Specifically, when generating the service processing time prediction formula for Manufacture1, as shown in FIG. 13, the service processing time of Manufacture1 (service ID: WS4) and the average stay message count of Manufacture1 are combined. , Manufacture2 (service ID: WS5) average stay message count is added. Then, using this as a sample, the relationship between the service processing time and the average stay message count is formulated by multiple regression analysis. By this formulation, a service processing time prediction formula for estimating the service processing time of Manufacture 1 is obtained using two parameters, the average number of stay messages in Manufacture 1 and the average number of stay messages in Manufacture 2.

  FIG. 14 shows a business system configuration in the simulator. In the figure, a rectangle is called a transition and shows a service, and a circle is called a place and shows the number of stay messages in the service for each input route. The delay due to the workflow execution unit 201 (FIG. 1) can be expressed as the time required from when the workflow execution unit 201 receives a message until the message (corresponding to the black circle in FIG. 17) is input to the place. The service processing time can be expressed as a processing time in the transition. The message in the service providing means 101 at the time of simulation can be grasped as a place (circular shape) within the workflow engine range.

For example, the delay time of the partial workflow in the workflow can be estimated by calculating the average number of stay messages in the workflow in one minute period in the simulator time. The delay time prediction formula in the partial workflow from the message input by the business system user to the call to the Retailer service, which is handled by the workflow execution apparatus 200A,
D = 0.8M + 2.4
Suppose that In this case, if the average stay message count of the workflow execution apparatus 200A is “2”, the delay time in the partial workflow is calculated as 4 seconds.

  In the above description, the delay time is updated at a cycle of 1 minute. Instead of the cycle, the delay time can be calculated by calculating the number of staying messages in the workflow using message input / output for the workflow as an event. The simulation unit 501 delays the message by the calculated delay time every time a message is input to the workflow execution apparatus according to the calculated delay time, and sends the message to the next stage service or workflow execution apparatus. By passing the message, the behavior of the message in the business system is reproduced.

In calculating the service processing time, the number of staying messages is calculated every time a message arrives at the service, and the processing time in the server machine is estimated. In the estimation of the service processing time, first, the average processing time of the service is calculated. For example, for the server machine 100D, the processing time prediction formula in the Manufacture1 service is
T = 7.2 × M ₁ + 1.3 × M ₂ +8.4
Suppose that If the average stay message number M ₁ of the Manufacture 1 service is “2” and the average stay message number M ₂ of the Manufacture 2 service is “1”, the service average processing time of the Manufacture 1 service is calculated as 24.1 seconds. Next, using the calculated service average processing time and the probability distribution estimated by the service processing time prediction formula generation unit 402, a service processing time considering variation is obtained.

Specifically, the service processing time in consideration of fluctuation is calculated by calculating T ′ as processing time in consideration of fluctuation, R as a random number based on probability distribution (0 ≦ x ≦ 1), K as standard deviation, and T as calculated. The average service processing time is calculated by the following formula.
T ′ = (R−0.5) × 2 × K + T
For the random number R based on the probability distribution in the above formula, for example, a random number generation function corresponding to the probability distribution is a normal distribution, a random number generation function corresponding to the Poisson distribution is prepared, and the probability distribution of the service processing time is Which type of distribution is judged and a random number generation function to be used is selected and generated by the selected random number generation function. The simulation unit 501 reproduces the behavior of the message in the business system according to the service processing time considering the fluctuation calculated in this way.

  In the present embodiment, when one server machine 100 executes a plurality of services, the number of stay messages in the service and the service processing time including the number of stay messages in other services sharing the server machine 100 are included. The service processing time prediction formula is generated by formulating the relationship between the service processing time, the service processing time is calculated using the service processing time prediction formula generated in the simulation, and the workflow processing time is predicted. For this reason, even when there is a service that shares resources, the accuracy of workflow processing time prediction does not drop. Further, in the present embodiment, the influence by the workflow execution device 200 is formulated to generate a delay time prediction formula, and the delay time by the workflow execution device 200 is predicted using the generated delay time prediction formula in the simulation. For this reason, even when the workflow execution device 200 exists in the business system, the prediction accuracy of the workflow processing time does not drop.

  In this embodiment, in the simulation, a change in the number of stay messages in the server machine and the workflow execution device is detected, and the service processing time and workflow delay time used for reproducing the message behavior are recalculated. By doing in this way, the behavior of the message in the workflow can be reproduced in a form close to the actual operation situation, and the actual operation situation can be reproduced faithfully. Accordingly, it is possible to improve the prediction accuracy of the service processing time and the workflow delay time. In addition, a probability distribution for the service processing time is obtained, and in consideration thereof, the service processing time calculated using the service processing time prediction formula is changed. As a result, the workflow processing time can be predicted in a more realistic manner.

  Although the present invention has been described based on the preferred embodiments, the distributed workflow simulation system, method, and program of the present invention are not limited to the above embodiments, and the configuration of the above embodiments. Various modifications and changes are also included in the scope of the present invention.

The block diagram which shows the distributed workflow simulation system of one Embodiment of this invention. The block diagram which shows the connection condition of each apparatus which comprises a distributed workflow simulation system. The flowchart which shows the operation | movement procedure of the production | generation of a workflow delay prediction formula. The flowchart which shows the procedure of the production | generation of a service processing time prediction formula. The figure which shows the specific example of a simulation model. The flowchart which shows the procedure of workflow delay time calculation. The flowchart which shows the procedure of service processing time calculation. The block diagram which shows the structure of the business system of an Example. The block diagram which shows the apparatus structure which implement | achieves the business system of an Example. The figure which shows the specific example of measurement information. The figure which shows the group of partial workflow delay time and the average stay message number. The figure which shows the group of service processing time and the average number of stay messages. The figure which shows the group of the service processing time and the average number of stay messages about the service which shares a server machine with another service. The model figure which shows the business system in a simulator. The block diagram which shows an example of a business system. 1 is a block diagram showing a distributed workflow simulation system described in Patent Document 1. FIG. The model figure which shows the specific example of a simulation model.

Explanation of symbols

10: distributed workflow simulation system 100: server machine 101: service providing unit 102: service monitoring unit 200: workflow execution unit 201: workflow execution unit 202: workflow monitoring unit 301: workflow definition storage unit 302: message log storage unit 303: server Machine service correspondence information storage device 304: Simulation model storage device 305: Simulation result storage device 400: Business system management device 401: Message log collection unit 402: Service processing time prediction formula generation unit 403: Workflow delay prediction formula generation unit 404 : Simulation model generation means 500: Simulation device 501: Simulation means 502: Workflow delay calculation means 510: Service processing time calculation means 5 1: Service Processing time dispersion calculating means 512: the server machine sharing service processing time calculation unit 601: input device / Property 602: Output device / GUI
700: Network

Claims (42)

In a distributed workflow simulation system that simulates a business system including a workflow execution device that performs message delivery control and a server machine that executes at least one service according to a received message,
The number of stay messages in the service and the message processing time of the service are acquired, and for services that share the server machine with other services, the number of stay messages in the service and other services that share the server machine with the service A service processing time prediction formula generating means for formulating a relationship between the number of stay messages in the service and the processing time of the message in the service, and generating a service processing time prediction formula;
A simulation model generating means for generating a simulation model based on a workflow definition for defining a delivery route of a message in the business system and the service processing time prediction formula;
Based on the simulation model, simulating means for sequentially reproducing the behavior of messages in the business system;
For the service sharing the server machine with the other service, the number of stay messages in the service reproduced by the simulating means, the number of stay messages in the other service sharing the service and the server machine, A distributed workflow simulation system comprising service processing time calculation means for calculating the service processing time based on a service processing time prediction formula.   The service processing time prediction formula generation unit formulates the relationship between the number of stay messages in the service and the message processing time in the service when one service is executed on one server machine. A service processing time prediction formula is generated, and the service processing time calculation means calculates the service processing time based on the number of stay messages in the service reproduced by the simulation means and the service processing time prediction formula. The distributed workflow simulation system according to claim 1, wherein the calculation is performed.   The service processing time prediction formula generating means refers to a message log storage device that stores measurement information about messages that are input to and output from the service, and acquires the number of stay messages in the service and the processing time of the message The distributed workflow simulation system according to claim 1.   When the service processing time calculation means changes the number of stay messages in the service after the service processing time is calculated, the service processing time calculation means calculates the service processing time based on the service processing time prediction formula based on the number of stay messages after the change. The distributed work flow simulation system according to claim 1, wherein the processing time is recalculated.   The distributed workflow simulation system according to claim 1, wherein the service processing time prediction formula generation unit estimates a probability distribution of service processing time.   The distributed workflow simulation system according to claim 5, wherein the service processing time calculation unit calculates a service processing time in consideration of variation based on the calculated service processing time and a probability distribution of the service processing time. The number of stay messages in the workflow execution device and the delay time of the message by the workflow execution device are acquired, and the relationship between the number of stay messages in the workflow execution device and the delay time of the message by the workflow execution device is formulated. A delay time prediction expression generating means for generating a workflow delay time prediction expression
Workflow delay time calculation means for calculating the delay time of the message by the workflow execution apparatus based on the number of stay messages of the workflow execution apparatus reproduced by the simulation means and the workflow delay time prediction formula; The distributed workflow simulation system according to claim 1, comprising:   8. The simulation model generation means generates a simulation model based on a workflow definition that defines a message delivery route in the business system, the service processing time prediction formula, and the workflow delay time prediction formula. Distributed workflow simulation system described in 1.   The delay time prediction formula generation means refers to a message log storage device that stores measurement information about messages input to and output from the workflow execution device, and the number of stay messages of the workflow execution device, the delay time of the message, The distributed workflow simulation system according to claim 7, wherein:   When the number of stay messages in the workflow execution device changes after calculating the delay time of the message, the workflow delay time calculation means, based on the number of stay messages after the change and the workflow delay time prediction formula, The distributed workflow simulation system according to claim 7, wherein the delay time of the message is recalculated.   8. The distributed workflow simulation system according to claim 7, wherein the simulating means reproduces the behavior of a message in the business system based on the calculated service processing time and message delay time. A business system management device that manages a business system including a workflow execution device that performs message delivery control and a server machine that executes at least one service according to a received message,
The number of stay messages in the service and the message processing time of the service are acquired, and for services that share the server machine with other services, the number of stay messages in the service and other services that share the server machine with the service A service processing time prediction formula generating means for formulating a relationship between the number of stay messages in the service and the processing time of the message in the service, and generating a service processing time prediction formula;
A business system management apparatus comprising: a workflow definition that defines a message delivery route in the business system; and a simulation model generation unit that generates a simulation model based on the service processing time prediction formula.   The service processing time prediction formula generation unit formulates the relationship between the number of stay messages in the service and the message processing time in the service when one service is executed on one server machine. The business system management apparatus according to claim 12, which generates a service processing time prediction formula.   The number of stay messages in the workflow execution device and the delay time of the message by the workflow execution device are acquired, and the relationship between the number of stay messages in the workflow execution device and the delay time of the message by the workflow execution device is formulated. The business system management apparatus according to claim 12, further comprising: a delay time prediction expression generation unit that generates a workflow delay time prediction expression.   15. The simulation model generation unit generates a simulation model based on a workflow definition that defines a message delivery route in the business system, the service processing time prediction formula, and the workflow delay time prediction formula. The business system management device described in 1. A simulation device for simulating a business system comprising a workflow execution device that performs message delivery control and a server machine that executes at least one service according to a received message,
For services that share the server machine with other services, the relationship between the number of stay messages in the service, the number of stay messages in other services that share the service and server machine, and the message processing time in the service Of the message in the business system based on a simulation model generated based on the service processing time prediction formula generated by formulating and the workflow definition that defines the delivery route of the message in the business system. A simulation means to reproduce the behavior sequentially,
For the service sharing the server machine with the other service, the number of stay messages in the service reproduced by the simulating means, the number of stay messages in the other service sharing the service and the server machine, A distributed workflow simulation apparatus comprising service processing time calculation means for calculating the service processing time based on a service processing time prediction formula.   The service processing time prediction formula formulates the relationship between the number of stay messages in the service and the message processing time in the service when one service is executed on one server machine. And the service processing time calculation means calculates the service processing time based on the number of stay messages in the service reproduced by the simulation means and the service processing time prediction formula. Item 17. The distributed workflow simulation apparatus according to Item 16.   When the service processing time calculation means changes the number of stay messages in the service after the service processing time is calculated, the service processing time calculation means calculates the service processing time based on the service processing time prediction formula based on the number of stay messages after the change. The distributed work flow simulation apparatus according to claim 16, wherein the processing time is recalculated.   The distributed workflow simulation according to claim 16, wherein the service processing time calculation unit calculates a service processing time considering variation based on the calculated service processing time and a probability distribution of the estimated service processing time. apparatus. The simulation model is a workflow delay time prediction formula generated by formulating a relationship between the number of stay messages in the workflow execution device and the delay time of the message by the workflow execution device, and the workflow delay time prediction. Generated based on the formula and the workflow definition,
Workflow delay time calculating means for calculating the delay time of the message by the workflow execution device based on the number of stay messages of the workflow execution device reproduced by the simulation means and the workflow delay time prediction formula is further provided. The distributed workflow simulation apparatus according to claim 16.   When the number of stay messages in the workflow execution device changes after calculating the delay time of the message, the workflow delay time calculation means, based on the number of stay messages after the change and the workflow delay time prediction formula, The distributed workflow simulation apparatus according to claim 20, wherein the delay time of the message is recalculated. In a distributed workflow simulation system for simulating a business system including a workflow execution device that performs message delivery control using a computer and a server machine that executes at least one service according to a received message,
For the service in which the computer acquires the number of stay messages in the service and the message processing time of the service and shares the server machine with other services, the number of stay messages in the service, the service and the server machine Formulating a relationship between the number of staying messages in another service to be shared and the message processing time in the service and generating a service processing time prediction formula;
The computer generates a simulation model based on a workflow definition that defines a message delivery route in the business system and the service processing time prediction formula;
The computer sequentially reproducing the behavior of messages in the business system based on the simulation model;
For a service in which the computer shares a server machine with the other service, the number of stay messages in the service reproduced by reproducing the message behavior, and a stay in another service sharing the server machine with the service A distributed workflow simulation method comprising: calculating the service processing time based on the number of messages and the service processing time prediction formula. When the computer executes one service on one server machine, a service processing time prediction formula is generated by formulating the relationship between the number of staying messages in the service and the message processing time in the service. And steps to
When the computer executes one service on one server machine, based on the number of stay messages in the service reproduced by reproducing the message behavior and the service processing time prediction formula, The distributed workflow simulation method according to claim 22, further comprising a step of calculating a service processing time.   In the step of obtaining the number of stay messages in the service and the processing time of the message, the computer refers to a message log storage device that stores measurement information about messages that are input to and output from the service. The distributed workflow simulation method according to claim 22, wherein the number of stay messages and the processing time of the messages are acquired.   When the number of stay messages in the service changes after the calculation of the service processing time, the computer resets the service processing time based on the number of stay messages after the change and the service processing time prediction formula. The distributed work flow simulation method according to claim 22, wherein the calculation is performed.   The distributed workflow simulation method according to claim 22, further comprising the step of estimating a probability distribution of service processing time by the computer.   27. In the step of calculating the service processing time, the computer calculates a service processing time considering variation based on the calculated service processing time and a probability distribution of the service processing time. Distributed workflow simulation method. The computer acquires the number of stay messages in the workflow execution device and the delay time of the message by the workflow execution device, and the number of stay messages in the workflow execution device and the delay time of the message by the workflow execution device Formulating the relationship between and generating a workflow delay time prediction formula,
The computer calculating the delay time of the message by the workflow execution device based on the number of stay messages of the workflow execution device reproduced by reproducing the message behavior and the workflow delay time prediction formula; The distributed workflow simulation method according to claim 22, further comprising:   In the step of generating the simulation model, the computer relates to a workflow delay based on a workflow definition that defines a message delivery route in the business system, the service processing time prediction formula, and the workflow delay time prediction formula. The distributed workflow simulation method according to claim 28, wherein a simulation model is generated.   In the step of acquiring the number of stay messages of the workflow execution device and the delay time of the message, the computer refers to a message log storage device that stores measurement information about messages input to and output from the workflow execution device, and 29. The distributed workflow simulation method according to claim 28, wherein the stay message count of the workflow execution device and the delay time of the message are acquired.   When the number of stay messages in the workflow execution device changes after calculating the message delay time, the computer delays the message based on the number of stay messages after the change and the workflow delay time prediction formula. 29. The distributed workflow simulation method according to claim 28, wherein the time is recalculated.   29. The distributed workflow according to claim 28, wherein in the step of reproducing the message behavior, the computer reproduces the message behavior in the business system based on the calculated service processing time and message delay time. Simulation method. A program for causing a computer to execute a process for managing a business system including a workflow execution device that performs message delivery control and a server machine that executes at least one service in accordance with a received message.
The number of stay messages in the service and the message processing time of the service are acquired, and for services that share the server machine with other services, the number of stay messages in the service and other services that share the server machine with the service A process of formulating a relationship between the number of stay messages in the service and the message processing time in the service to generate a service processing time prediction formula;
A program for executing a process for generating a simulation model based on a workflow definition for defining a message delivery route in the business system and the service processing time prediction formula.   In the process of generating the service processing time prediction formula, when one service is executed on one server machine, the relationship between the number of staying messages in the service and the message processing time in the service is formulated. 34. The program according to claim 33, wherein the program generates a service processing time prediction formula.   The computer acquires the number of stay messages in the workflow execution device and the delay time of the message by the workflow execution device, and the number of stay messages in the workflow execution device and the delay time of the message by the workflow execution device 34. The program according to claim 33, further comprising executing a process of formulating the relationship of generating a workflow delay time prediction formula.   In the process of generating the simulation model, the simulation model is generated based on a workflow definition that defines a message delivery route in the business system, the service processing time prediction formula, and the workflow delay time prediction formula. Item 36. The program according to Item 35. A program for causing a computer to execute a process of simulating a business system including a workflow execution device that performs message delivery control and a server machine that executes at least one service in accordance with a received message.
For services that share the server machine with other services, the relationship between the number of stay messages in the service, the number of stay messages in other services that share the service and server machine, and the message processing time in the service Of the message in the business system based on a simulation model generated based on the service processing time prediction formula generated by formulating and the workflow definition that defines the delivery route of the message in the business system. Processing to reproduce the behavior sequentially,
Regarding the service sharing the server machine with the other service, the number of stay messages in the service reproduced in the process of reproducing the behavior of the message, and the stay message in the other service sharing the service and the server machine And a process for calculating the service processing time based on the number and the service processing time prediction formula.   The service processing time prediction formula formulates the relationship between the number of stay messages in the service and the message processing time in the service when one service is executed on one server machine. In the process of calculating the service processing time that has been generated, the service processing based on the number of stay messages in the service reproduced in the process of reproducing the behavior of the message and the service processing time prediction formula The program according to claim 37, wherein the time is calculated.   In the process of calculating the service processing time, when the number of stay messages in the service changes after the calculation of the service processing time, based on the number of stay messages after the change and the service processing time prediction formula, The program according to claim 37, wherein the service processing time is recalculated.   38. The program according to claim 37, wherein, in the process of calculating the service processing time, a service processing time considering variation is calculated based on the calculated service processing time and the estimated probability distribution of the service processing time. . The simulation model is a workflow delay time prediction formula generated by formulating a relationship between the number of stay messages in the workflow execution device and the delay time of the message by the workflow execution device, and the workflow delay time prediction. Generated based on the formula and the workflow definition,
The delay time of the message by the workflow execution device is calculated on the computer based on the number of stay messages of the workflow execution device reproduced by the process of reproducing the behavior of the message and the workflow delay time prediction formula. The program according to claim 37, further causing the process to be executed.   In the process of calculating the workflow delay time, when the number of stay messages in the workflow execution apparatus changes after the delay time of the message is calculated, the number of stay messages after the change and the workflow delay time prediction formula are used. 42. The program according to claim 41, wherein the delay time of the message is recalculated.

Images