JP4932749B2 - Resource state prediction apparatus, resource state prediction method, and program - Google Patents

Description

  The present invention predicts the utilization rate of each resource after a predetermined time by using the utilization rate of each resource collected from a control target system having a plurality of computer resources. The present invention relates to a technique for calculating an additional amount when predicted, and calculating a return amount when surplus is predicted.

  With the progress of communication network technology, there are many integrated systems that operate in parallel by linking multiple computer systems through a network. By integrating the systems, there are merits such that the processing load can be distributed, and the surplus and shortage of resources can be suppressed by accommodating computer resources among the systems. In order to reliably enjoy these merits, it is necessary to accurately predict dynamic needs for computer resources.

  This is because if this prediction is not accurate, the ratio of distributing the processing load among the systems may be wrong, resulting in a shortage of resources somewhere, which may trigger a system down. . If computer systems are linked via a network, the system down may be chained and the expected benefits of integration may even turn into major damage. Therefore, predicting the dynamic needs of computer resources is a central issue in network integration systems.

  As a technique for dealing with such a problem, for example, in Patent Document 1 below, a load state after a predetermined period is predicted from an operation history of a past computer resource and an operation state of a current computer resource. .

JP 11-261702 A

  By the way, when the system to be monitored becomes large, the number of computer resources becomes enormous, and the number of data collected from the system to be monitored becomes enormous. In the above-mentioned patent document 1, since each computer resource is predicted using an operation history measured a plurality of times in the past, the total data amount is enormous.

  Further, in Patent Document 1 described above, the state of one computer resource is predicted. However, considering the entire system, it may be necessary to predict the state after a predetermined period for a large number of computer resources. In this case, in Patent Document 1 described above, a state after a predetermined period is predicted for each computer resource, and the processing load may increase. If the system to be monitored becomes large, the number of computer resources may become enormous and the processing load may increase.

  The present invention has been made in view of the above circumstances, and an object of the present invention is to reduce the processing load when predicting the dynamic needs of computer resources and to reduce the amount of data required for predictive calculation. There is.

  In order to solve the above-mentioned problem, the present invention discretizes the measured value of each resource at two consecutive time points into one of a plurality of states according to a predetermined threshold value for each resource. For each measured value at the time of, the state of each resource is regarded as one cell, and a causal relationship model of a cellular automaton is introduced. Backward calculation of the transition rule from the state of the predetermined number of cells to the state of one cell in the predetermined number of consecutive cells adjacent at a later time, and after a predetermined time using the calculated transition rule Predict the state of each resource.

For example, the present invention is a resource state prediction device that predicts a state after each resource based on the current state of each resource in the system to be monitored, and outputs a prediction result.
A measurement value collecting means for collecting measurement values of each resource from the monitored system and discretizing each measurement value into one of a plurality of states using a predetermined threshold for each resource;
Resource state holding means for holding the state of each resource collected and discretized by the measurement value collecting means in association with the time when the resource was measured;
The state of each resource at two consecutive times is sequentially read from the resource state holding means and arranged in one column at each time, each resource is regarded as one cell in the cellular automaton, and each resource at the previous time is In a state, for each predetermined number of adjacent cells including the target cell, the state of each of the predetermined number of cells is set as an input state, and becomes the target in the state of each resource at a later time A transition rule extracting means for extracting a transition rule having a cell state as an output state for each of all cells;
Of the transition rules extracted by the transition rule extracting means, the ratio of different output states is calculated for each transition rule having the same input state, and the highest output state is output in the transition rule. A transition function creating means for creating a transition function composed of a plurality of transition rules including the adopted output state,
Accepting means for accepting input of time from the current time to the time at which the state of each resource should be predicted;
A measurement number calculation means for calculating a measurement number, which is the number of times a measurement value of each resource is measured, from the current time until the time input via the reception unit elapses;
The current state of each resource is sequentially read out from the measurement value holding means and arranged in one column, each resource is regarded as one cell in the cellular automaton, and a predetermined number of cells adjacent to each other including the target cell Each time, a transition rule having the respective states of the predetermined number of cells as input states is extracted from the transition function created by the transition function creating means, and the output state included in the extracted transition rule is A next state predicting means for executing processing for calculating the state at the next time of the target cell as a target for each of all cells;
The next state predicting means further includes:
With reference to the transition function, the process of calculating the next state of each resource from the calculated next state of each resource is repeated by the number of times of measurement calculated by the number-of-measurement calculating means minus one, and finally There is provided a resource state prediction apparatus that outputs the calculated state of each resource to the outside as the state of each resource after a time input via the receiving unit has elapsed.

  According to the resource state prediction apparatus of the present invention, it is possible to reduce the processing load when predicting the dynamic needs of computer resources and to reduce the amount of data necessary for the prediction calculation.

  First, a first embodiment of the present invention will be described.

  FIG. 1 is a system configuration diagram showing a configuration of a control system 10 according to an embodiment of the present invention. The control system 10 includes an autonomous control device 13, a monitoring target system 14, a resource change amount calculation device 20, and a resource state prediction device 30. The monitored system 14 provides a predetermined service to a plurality of client computers 11 via a communication line 12 such as the Internet. The autonomous control device 13, the monitoring target system 14, the resource change amount calculation device 20, and the resource state prediction device 30 are connected via a management communication line.

  The monitoring target system 14 is a system managed by the autonomous control device 13, and is configured by interconnecting a plurality of computer resources (for example, the server 140). Each server 140 includes a plurality of resources 143 or resources 144 having a calculation function such as a CPU (Central Processing Unit), a storage function such as a memory or a hard disk, or a communication function. A sufficient amount of computer resources for realizing a service provided to the client computer 11 via the communication line 12 is set and arranged and connected to the monitoring target system 14 in advance by a system designer or the like.

  Each server 140 has a resource pool 142 in addition to the operating resource 141, and in response to an instruction from the autonomous control device 13, the resource 144 in the resource pool 142 is added to the operating resource 141. The resource 143 in the resource 141 in operation is returned to the resource pool 142. As a result, the monitoring target system 14 maintains service provision to the client computer 11 and effectively utilizes resources in the monitoring target system 14. In addition to the active server 140, the inactive server 140 is provided in the monitoring target system 14, and the server 140 is operated or withdrawn according to an instruction from the autonomous control device 13. May be.

  The autonomous control device 13 acquires the usage rate of each resource in the monitoring target system 14 as a measurement value, and instructs the monitoring target system 14 to add / return each resource according to a predetermined control policy. In addition, when the autonomous control device 13 is notified of the amount of change of each resource based on the prediction result of the state of each resource after a predetermined time and the likelihood indicating the certainty of the prediction result, the autonomous control device 13 uses the likelihood. If it is determined whether to control the monitoring target system 14 using the change amount, and it is determined to control the monitoring target system 14 using the change amount, the monitoring target system 14 is based on the change amount. To control.

  The resource state prediction device 30 acquires the usage rate of each resource in the monitoring target system 14 as a measured value, and predicts the state of each resource after the time specified by the user has elapsed based on the acquired measured value. To do. Then, the resource state prediction device 30 sends the prediction result to the resource change amount calculation device 20 together with the likelihood of the transition function used for the prediction result.

  The resource change amount calculation device 20 calculates the change amount of each resource necessary to maintain the current state based on the prediction result received from the resource state prediction device 30, and uses the calculated change amount as the resource state prediction device 30. Are sent to the autonomous control device 13 together with the likelihood of the transition function received from.

  FIG. 2 is a block diagram illustrating an example of a functional configuration of the resource state prediction device 30 according to the first embodiment. The resource state prediction apparatus 30 includes a display unit 300, a resource state holding unit 301, a transition rule extraction unit 302, a transition function creation unit 303, a next state prediction unit 304, a measured value collection unit 305, a threshold information holding unit 306, and a reception unit 307. And a measurement number calculation unit 308.

  For example, as illustrated in FIG. 3, the threshold information holding unit 306 stores a threshold 3061 to be applied to the measurement value of the corresponding resource for each resource ID 3060 for identifying each resource. The measurement value collection unit 305 acquires the measurement value and the time when the measurement value is measured for each resource from the monitoring target system 14. In the present embodiment, the measurement value acquired by the measurement value collection unit 305 from the monitoring target system 14 is the utilization rate of each resource, and each measurement value is measured at predetermined time intervals (for example, every minute).

  The measurement value collection unit 305 refers to the threshold information holding unit 306 for each resource, discretizes the acquired measurement value into one of a plurality of states, and uses the discretized state of each resource as a source of the state. The measured value is stored in the resource state holding unit 301 in association with information indicating the time when the measured value is measured. In the present embodiment, the measurement value collection unit 305 sets the state of the resource to 1 when the measurement value acquired from the monitoring target system 14 is greater than or equal to the threshold stored in the threshold information holding unit 306 for each resource. If it is less than the threshold, the resource state is discretized as 0.

For example, as illustrated in FIG. 4, the resource state holding unit 301 includes a value indicating the state of the resource corresponding to each resource ID 3010 for each identifier 3011 indicating the time when the state of the resource is collected by the measurement value collection unit 305. Is stored. In the example illustrated in FIG. 4, “T ₀ ” indicates the time when the current state of each resource is collected, and “T ₋₁ ” is measured before “T ₀ ”. Indicates the time.

When the measurement value collection unit 305 discretizes the measurement value of each resource corresponding to the new time, the measurement value collection unit 305 deletes the state of “T ₋₄ ” and _shifts “T ₀ ” to “T ₋₃ ” one to the left. The data in the resource state holding unit 301 is updated by shifting the state one by one and storing the state of each resource corresponding to the new time in “T ₀ ”. In this embodiment, the resource state holding unit 301 stores the state of each resource at the past five points in time, but the resource state holding unit 301 stores the state of each of the past two or more resources. Just do it.

  The accepting unit 307 receives, via the input device 16 such as a keyboard or a mouse, a prediction instruction including a prediction time indicating a time until prediction and a generation time indicating a time interval for generating a transition function, as a resource state When received from the user of the prediction device 30, the prediction time is sent to the measurement number calculation unit 308, and a transition rule collection start instruction including the creation time is sent to the transition rule extraction unit 302.

  When receiving the predicted time from the accepting unit 307, the measurement number calculation unit 308 calculates a measurement number that is the number of times the measured value of each resource is measured within the prediction time, and the calculated measurement number is Send to the prediction unit 304. In this embodiment, the monitoring target system 14 measures the measurement value of each resource every minute, and the measurement count calculation unit 308 measures the quotient obtained by dividing the predicted time received from the reception unit 307 by one minute. Calculate as the number of times.

When the transition rule extraction unit 302 receives a transition rule collection start instruction from the reception unit 307, the transition rule extraction unit 302 displays the state of each resource at two consecutive times for each creation time included in the collection start instruction. Read sequentially from the holding unit 301. In the present embodiment, the transition rule extraction unit 302 reads the state of each resource at time T ₀ (ie, the current state of each resource) and the state of each resource at time T ₋₁ from the resource state holding unit 301.

  Then, as illustrated in FIG. 5, the transition rule extraction unit 302 regards the state of each resource as one cell in the cellular automaton, and arranges information indicating the state of each resource in one column for each time. In the example shown in FIG. 5, a plain cell indicates a 0 state, and a shaded cell indicates a 1 state.

  Then, the transition rule extraction unit 302 inputs the state of each of the predetermined number of cells for each predetermined number of consecutive cells including the target cell in the state of each resource at the previous time. In addition, in the state of each resource at a later time, a transition rule whose output state is the state of the target cell is extracted.

  In the present embodiment, the transition rule extraction unit 302 sets the state of each of the three cells as an input state for each of three adjacent cells in the state of each resource at the previous time. The transition rule extraction unit 302 sets the target cell as the center cell of the three cells, and sets the state of the target cell as an output state in the state of each resource at a later time. Transition rules 40 extracted for all cells by the transition rule extraction unit 302 are as shown in FIG. 6, for example. The transition rule extraction unit 302 sends the transition rules extracted for all cells to the transition function creation unit 303.

  When a transition rule is extracted for resource 1, transition rule extraction unit 302 uses resource n, resource 1, and resource 2 as the state of each resource at the previous time. When extracting a transition rule for resource n, transition rule extraction unit 302 uses resource n-1, resource n, and resource 1 as the state of each resource at the time.

  Further, when the transition function extraction unit 302 is instructed to re-extract the transition rule from the transition function creation unit 303, the transition rule extraction unit 302 displays the state of each resource at the previous time from the resource state holding unit 301 as shown in FIG. Sequential readout, transition rules for each cell are extracted again from the state of each resource at the previous time and the state of each resource at the earlier time, and a transition function is created for the extracted transition rule again Send to section 303.

  When the transition function creation unit 303 receives the transition rule from the transition rule extraction unit 302, for example, as illustrated in FIG. 8, the transition function creation unit 303 calculates the ratio of the output state for each transition rule having the same input state. . In the example shown in FIG. 8, in the previous state 42, when the left cell 420 is “0”, the central cell 421 is “0”, and the right cell 422 is “0”, the state after the central cell 43 are all shown to be “0”. In the previous state 42, the left cell 420 is “0”, the center cell 421 is “1”, and the right cell 422 is “1”. In the state 43 after the central cell, the ratio of “0” is 70%, and the ratio of “1” is 30%.

  The transition function creating unit 303 adopts the highest output state in each transition rule as the output state in the transition rule, and creates a transition function using a plurality of transition rules including the adopted output state. . The transition function 44 created by the transition function creation unit 303 is, for example, as shown in FIG. Then, the transition function creating unit 303 created a ratio of transition rules in which the ratio of the output state is equal to or higher than a first threshold (for example, 80%) preset by the administrator of the resource state prediction device 30 or the like. Calculated as a likelihood indicating the likelihood of the transition function. In the example shown in FIG. 8, when the first threshold is 80%, the transition function creation unit 303 calculates the likelihood of the transition function 44 shown in FIG. 9 as 7 ÷ 8 × 100 = 87.5%. To do.

  Then, the transition function creation unit 303 determines whether or not the calculated likelihood of the transition function is equal to or greater than a second threshold (for example, 70%) preset by the administrator of the resource state prediction device 30 or the like. To do. When the likelihood of the calculated transition function is equal to or greater than the second threshold, the transition function creating unit 303 sends the calculated transition function to the next state prediction unit 304 together with the calculated likelihood.

  On the other hand, when the calculated likelihood of the transition function is less than the second threshold, the transition function creation unit 303 instructs the transition rule extraction unit 302 to re-extract the transition rule. After that, when a new transition rule is received from the transition rule extraction unit 302, the transition function creation unit 303 integrates the received transition rule with the transition rule previously received from the transition rule extraction unit 302, and the input state is the same. For each transition rule, the ratio of each output state is calculated. As described above, since the extraction of the transition rule is repeated until the likelihood of the transition function becomes a predetermined value or more, the resource state prediction device 30 can improve the accuracy of the calculated transition function.

  When the next state prediction unit 304 receives the measurement number from the measurement number calculation unit 308, the next state prediction unit 304 sequentially reads the current state of each resource from the resource state holding unit 301. Then, the next state prediction unit 304 regards each resource as one cell in the cellular automaton, and each state of the predetermined number of cells for each predetermined number of consecutive cells including the target cell. Are extracted from the transition function received from the transition function creation unit 303.

  Then, the next state prediction unit 304 calculates the output state included in the extracted transition rule as the state at the next time of the target cell. As shown in FIG. 10, the next state prediction unit 304 targets all the cells, sets the current state of each resource as an initial state, and uses the transition function received from the transition function creation unit 303 to By repeating the processing for obtaining the next state for the number of times of measurement received from the number of times of measurement calculation unit 308, the state of each resource at the time point instructed by the user is predicted.

  Then, the next state prediction unit 304 creates a prediction result 50 as shown in FIG. 11, for example, and sends the created prediction result 50 to the resource change amount calculation device 20 and the display unit 300. In the prediction result 50, for each resource ID 51, the current state 52 of the corresponding resource and the prediction state 53 of the corresponding resource are stored together with the likelihood 54 of the transition function used to calculate the prediction state 53. . The next state predicting unit 304 uses the transition function previously calculated by the transition function creating unit 303 until the new transition function is calculated by the transition function creating unit 303. Predict resource status.

  As described above, the resource state prediction apparatus 30 according to the present embodiment needs only one transition function to predict the state of each resource. Therefore, it is not necessary to perform a process for obtaining an approximate expression representing the behavior of the measured value, and the processing load can be reduced. In addition, in order to obtain the transition function, it is only necessary to have the measured values of each resource measured at at least two different times. Thus, as in the past, dozens or hundreds of measured value histories are stored for each resource. There is no need to keep it. Therefore, the required memory capacity can be reduced.

  The display unit 300 displays the state of each resource predicted by the next state prediction unit 304 and the actual state of each resource collected by the measurement value collection unit 305 as a prediction result screen 60 as shown in FIG. It is displayed on the display device 15. In the prediction result screen 60 shown in FIG. 12, the resource ID of each resource is displayed in the area 61, the state of each resource is displayed in the area 62, and the state predicted in each resource is displayed in the area 63. A predicted success rate indicating the rate at which the actual state matches is displayed.

  When the button 64 or the button 65 is operated on the prediction result screen 60, the display unit 300 displays the older state of each resource or the newer state of each resource on the display device 15. When the button 66 or the button 67 is operated, the display unit 300 displays the state of other resources not displayed on the screen on the display device 15.

  By browsing the prediction result screen 60, the user of the resource state prediction device 30 can visually recognize the accuracy of prediction by the resource state prediction device 30. Further, the user of the resource state prediction device 30 can find a resource that is likely to fail in prediction by browsing the prediction result screen 60.

  FIG. 13 is a block diagram illustrating an example of a functional configuration of the resource change amount calculation device 20. The resource change amount calculation device 20 includes a change amount acquisition unit 21, a resource change amount calculation unit 22, and a resource information holding unit 23.

  For example, as illustrated in FIG. 14, the resource information holding unit 23 includes, for each resource ID 230, a resource type 231 corresponding to the resource ID 230, a control 232 indicating whether or not the resource can be changed, and the resource information The current total amount 233 is stored. For resources that can be added or returned, a circle is stored in the control 232 column, and for resources that cannot be added or returned, a cross is stored in the control 232 column.

  When the change amount acquisition unit 21 receives the actually changed resource amount from the autonomous control device 13, the change amount acquisition unit 21 updates the total amount 233 in the resource information holding unit 23 for the corresponding resource.

  When the resource change amount calculation unit 22 receives the prediction result 50 shown in FIG. 11 from the resource state prediction device 30, the resource change amount calculation unit 22 compares the current state with the prediction state, and the resource ID of the resource whose state is changing Is identified. Then, the resource change amount calculation unit 22 refers to the change resource extraction unit 23, and when the resource corresponding to the identified resource ID is a resource that can be added or returned, the current state is “1”, If the predicted state is “0”, it is determined that the resource should be returned, and if the current state is “0” and the predicted state is “1”, it is determined that the resource should be added. .

  Next, the resource change amount calculation unit 22 refers to the change resource extraction unit 23 and calculates the amount of the resource to be added or returned for each resource whose state has changed. In the present embodiment, the resource change amount calculation unit 22 calculates an amount corresponding to a predetermined ratio (for example, 10%) as the amount of resources to be added or returned. For example, in FIG. 14, the resource change amount calculation unit 22 calculates 5 G bits × 10% = 500 M bits as the amount of resources for the resource with the resource ID “0002”.

  Then, the resource change amount calculation unit 22 calculates, from the resource state prediction device 30, the amount of the resource to be added or returned, calculated for each resource whose state has been changed and to which the resource can be added or returned. It is sent to the autonomous control device 13 together with the likelihood included in the received prediction result 50. The autonomous control device 13 that has received this is based on the amount of resources to be added or returned received from the resource change amount calculation device 20 when the likelihood is equal to or greater than a predetermined value (for example, 90%). To control the monitoring target system 14.

  Here, in the autonomous control in the present embodiment, the monitoring target system 14 is in a desirable load state (the utilization rate of each resource is not too high and can cope with a slight increase in load, and the use of each resource) When the system is operating in a state where the rate is not too low and the utilization efficiency of each resource is high), control is performed so as to maintain the load state. That is, the user of the resource state prediction device 30 causes the resource state prediction device 30 to predict the load state of each resource after a predetermined time when the monitored system 14 is in a desired load state, and the resource change amount calculation device 20 From the prediction result of the resource state prediction device 30, a resource change amount for returning the load state of the monitored system 14 to the desired load state is provided to the autonomous control device 13.

  FIG. 15 is a flowchart illustrating an example of the operation of the control system 10 in the first embodiment. When a prediction instruction including a prediction time indicating a time until a time point at which each resource state is to be predicted and a generation time indicating a time interval for generating a transition function is received from the user, the control system 10 performs this flowchart. The operation shown in is started. The processing related to the collection of measurement values by the measurement value collection unit 305 of the resource state prediction device 30 and the acquisition of resource change amounts by the change amount acquisition unit 21 of the resource change amount calculation device 20 is executed as needed. And

  First, the reception unit 307 sends the prediction time to the measurement number calculation unit 308 and also sends a transition rule collection start instruction including the creation time to the transition rule extraction unit 302. Then, the transition rule extraction unit 302 resets a timer for counting the creation time (S100).

  Next, the measurement number calculation unit 308 calculates a measurement number that is the number of times the measured value of each resource is measured within the prediction time received from the reception unit 307, and sends the calculated measurement number to the next state prediction unit 304. Send (S101). Then, the resource state prediction device 30 executes a prediction process for predicting the state of each resource after the prediction time has elapsed, and outputs a prediction result to the resource change amount calculation device 20 (S200). Details of the prediction process will be described later.

  Next, the resource change amount calculation unit 22 compares the current state included in the prediction result received from the resource state prediction device 30 with the prediction state, and identifies the resource ID of the resource whose state has changed. Then, with respect to the identified resource ID, the changed resource extraction unit 23 is referred to, and the amount of the resource to be added or returned is calculated for each resource whose state has changed (S102). Then, the resource change amount calculation unit 22 calculates the amount of the resource to be added or returned, calculated for each resource whose state has changed, in the prediction result 50 received from the resource state prediction device 30. To the autonomous control device 13 together with the degree.

  Next, the autonomous control device 13 determines whether or not to adopt the resource change amount calculated by the resource change amount calculation device 20, for example, by determining whether or not the likelihood is equal to or greater than a predetermined value. Determine (S103). When the resource change amount calculated by the resource change amount calculation device 20 is adopted (S103: Yes), the autonomous control device 13 in the monitoring target system 14 in accordance with the resource change amount calculated by the resource change amount calculation device 20 The amount of each resource is changed (S104), and the resource state prediction device 30 executes the process shown in step S200 again.

  On the other hand, when the resource change amount calculated by the resource change amount calculation device 20 is not employed (S103: No), the autonomous control device 13 monitors the system to be monitored according to the control policy set in advance by the operation manager of the control system 10. 14 is controlled (S105), and the resource state prediction device 30 executes the process shown in step S200 again.

  FIG. 16 is a flowchart illustrating an example of the prediction process (S200) in the first embodiment.

  First, the transition rule extraction unit 302 refers to a timer for counting the creation time and determines whether or not it is time to create a transition function (S201). When it is time to create a transition function (S201: Yes), the transition rule extraction unit 302 resets the timer (S202). Then, the transition rule extraction unit 302 sequentially reads out the state of each resource at each of two consecutive times from the resource state holding unit 301 (S203).

  Next, as described with reference to FIG. 5, the transition rule extraction unit 302 regards each resource as one cell in the cellular automaton, arranges information indicating the state of each resource at each time in one column, For each predetermined number of adjacent cells including the target cell in the state of each resource at the time, the respective states of the predetermined number of cells are set as input states, and each resource at a later time In this state, a transition rule whose output state is the state of the target cell is extracted. Then, the transition rule extraction unit 302 sends the transition rules (see FIG. 6) extracted for all cells to the transition function creation unit 303 (S204).

  Next, as described with reference to FIG. 8, the transition function creating unit 303 calculates the ratio of each output state for each transition rule having the same input state. Then, the transition function creation unit 303 adopts the highest output state in each transition rule as the output state in the transition rule, and uses the plurality of transition rules including the adopted output state to generate the transition function. Create (S205).

  Next, the transition function creation unit 303 creates a ratio of transition rules whose output state ratio is equal to or higher than a first threshold (for example, 80%) preset by the administrator of the resource state prediction apparatus 30 or the like. It is calculated as a likelihood indicating the certainty of the transition function. Then, the transition function creation unit 303 determines whether or not the calculated likelihood of the transition function is equal to or greater than a second threshold (for example, 70%) preset by the administrator of the resource state prediction device 30 or the like. (S206).

  When the calculated likelihood of the transition function is greater than or equal to the second threshold (S206: Yes), the transition function creation unit 303 sends the calculated transition function together with the calculated likelihood to the next state prediction unit 304. The next state prediction unit 304 regards each resource as one cell in the cellular automaton, and inputs each state of the predetermined number of cells for each predetermined number of cells adjacent to each other including the target cell. A transition rule to be a state is extracted from the transition function received from the transition function creation unit 303.

  Then, the next state prediction unit 304 calculates the output state included in the extracted transition rule as the state at the next time of the target cell. As described with reference to FIG. 10, the next state prediction unit 304 targets all of the cells, sets the current state of each resource as an initial state, and uses each transition function received from the transition function creation unit 303. The state of each resource at the time point instructed by the user is predicted by repeating the processing for obtaining the next state for the number of times of measurement received from the number of times of measurement calculation unit 308 (S207).

  Next, the next state prediction unit 304 creates the prediction result 50 illustrated in FIG. 11, and sends the created prediction result 50 to the resource change amount calculation device 20 and the display unit 300 (S208). The display unit 300 displays the state of each resource predicted by the next state prediction unit 304 and the actual state of each resource collected by the measurement value collection unit 305 as a prediction result screen 60 as shown in FIG. The information is displayed on the display device 15 (S209), and the resource state prediction device 30 ends the prediction process (S200) shown in this flowchart.

  If it is not the transition function creation timing in step S201 (S201: No), the next state prediction unit 304 uses the transition function created last time by the transition function creation unit 303 at the time point instructed by the user. The state of each resource is predicted (S213), and the next state prediction unit 304 executes the process shown in step S208.

  In step S206, when the calculated likelihood of the transition function is less than the second threshold (S206: No), the transition function creation unit 303 instructs the transition rule extraction unit 302 to re-extract the transition rule. Then, the transition rule extraction unit 302 determines whether or not the measured value of each resource at the previous time exists in the resource state holding unit 301 (S210).

  Further, when the measured value of each resource at the previous time is present in the resource state holding unit 301 (S210: Yes), the transition rule extracting unit 302, as described in FIG. The state is sequentially read from the resource state holding unit 301 (S212). Then, in step S204, the transition rule extraction unit 302 extracts again the transition rule for each cell from the state of each resource at the previous time and the state of each resource at the previous time.

  On the other hand, when the measured value of each resource at the previous time does not exist in the resource state holding unit 301 (S210: No), the transition rule extracting unit 302 notifies the user of an error via the display device 15 or the like ( (S211), the resource state prediction apparatus 30 ends the prediction process (S200) shown in this flowchart. In addition, when the measured value of each resource at the previous time does not exist in the resource state holding unit 301 (S210: No), the transition function is not created in this processing, but the next state prediction unit 304 is created last time. The state of each resource after a predetermined time can be predicted using the transition function continuously.

  The first embodiment of the present invention has been described above.

  As is clear from the above description, according to the resource state prediction device 30 of the present embodiment, the processing load when predicting the dynamic needs of computer resources is kept low, and the amount of data required for prediction calculation is reduced. can do.

  Next, a second embodiment of the present invention will be described.

  FIG. 17 is a block diagram illustrating an example of a functional configuration of the resource state prediction device 30 according to the second embodiment. The resource state prediction apparatus 30 includes a display unit 300, a resource state holding unit 301, a transition rule extraction unit 302, a transition function creation unit 303, a next state prediction unit 304, a measured value collection unit 305, a threshold information holding unit 306, and a reception unit 307. A measurement number calculation unit 308 and a device configuration information holding unit 309.

  Except for the points described below, in FIG. 17, the components denoted by the same reference numerals as those in FIG. 2 have the same or similar functions as those in FIG. The resource state prediction apparatus 30 of the present embodiment is different from the resource state prediction apparatus 30 of the first embodiment in that the device state information holding unit 309 is provided.

  For example, as illustrated in FIG. 18, the device configuration information holding unit 309 includes, for each resource ID 3090, a resource name 3091 corresponding to the resource ID 3090 and an adjacent resource ID that indicates the ID of another resource adjacent to the resource. A resource ID 3092 is stored. Data in the device configuration information holding unit 309 is set in advance by a designer of the monitoring target system 14 or the like.

  Here, the term “adjacent” means not only a state in which the CPU and the memory are arranged on the same substrate, but also a state where they are physically separated, as well as a state where they are physically separated. This also means that one state has a great influence on the other state, such as servers that communicate with each other via a communication line. For example, as shown in FIG. 19, the adjacency relationship between resources is represented by a node 70 indicating each resource and a link 71 indicating that the resources are adjacent to each other. In addition, the figure which shows the adjacency relationship between the resources shown in FIG. 19 is predetermined by the designer of the monitoring target system 14.

  When the transition rule extraction unit 302 receives a transition rule collection start instruction from the reception unit 307, the transition rule extraction unit 302 refers to the device configuration information holding unit 309 and determines that the distance between the resources is the minimum when the resources are arranged in one column. The position of each resource is exchanged so that For example, when the relationship of resources shown in FIG. 19 is made with one link 71 as the distance 1 and the matrix 72 is created, for example, FIG. 20 is obtained. In order to minimize the distance between resources when the resources A to F are arranged in a line, the component on the right side of the diagonal component having only the distance 0 (the dotted line component in FIG. 20). The positions of the resources may be switched so that the sum of the distances in () is minimized.

  In the example of FIG. 20, the resource B and the resource E are exchanged, then the resource B and the resource D are exchanged, and then the resource F and the resource D are exchanged. Can be minimized. FIG. 21 shows, for example, a matrix 73 in which the total of the distances of the components on the right side of the diagonal component is minimized by such an operation.

  The transition rule extraction unit 302 replaces the positions of the resources so that the distance between the resources is minimized, and then, for each creation time included in the transition rule collection start instruction, the transition rule extraction unit 302 The state is sequentially read from the resource state holding unit 301 to extract a transition rule. Subsequent processing of the resource state prediction device 30 is the same as in the first embodiment.

  In this way, state transitions closer to the behavior of an actual system can be reproduced by applying cellular automata by placing highly relevant resources at close positions. Thereby, the precision of the created transition function can be increased.

  FIG. 22 is a flowchart showing an example of the operation of the control system 10 in the second embodiment. Except for the points described below, in FIG. 22, the processes denoted by the same reference numerals as those in FIG. 15 are the same as or similar to the processes in FIG.

  In step S101, after the number-of-measurements calculation unit 308 calculates the number of measurements, the transition rule extraction unit 302 refers to the device configuration information holding unit 309, and the distance between the resources is minimum when the resources are arranged in one column. The position of each resource is changed so that (S110), and the resource state prediction device 30 executes a prediction process (S200).

  The second embodiment of the present invention has been described above.

  Next, a third embodiment of the present invention will be described.

  FIG. 23 is a block diagram illustrating an example of a functional configuration of the resource state prediction device 30 according to the third embodiment. The resource state prediction apparatus 30 includes a display unit 300, a resource state holding unit 301, a transition rule extraction unit 302, a transition function creation unit 303, a next state prediction unit 304, a measured value collection unit 305, a threshold information holding unit 306, and a reception unit 307. , A measurement number calculation unit 308, a risk value calculation unit 310, and a cost information holding unit 311.

  Except for the points described below, in FIG. 23, the components denoted by the same reference numerals as those in FIG. 2 have the same or similar functions as those in FIG. The resource state prediction apparatus 30 of the present embodiment is different from the resource state prediction apparatus 30 of the first embodiment in that a risk value calculation unit 310 and a cost information holding unit 311 are provided.

  In the cost information holding unit 311, for example, as illustrated in FIG. 24, a cost value 3111 is stored for each state change 3110 indicating whether or not there is a resource state change. The cost value is the cost that is incurred when control is performed to correct the change when there is a change in the resource state. If the resource state does not change, the state is maintained. This is the cost incurred.

  The next state prediction unit 304 sends the created prediction result (see FIG. 11) to the resource change amount calculation device 20, the display unit 300, and the risk value calculation unit 310. When the risk value calculation unit 310 receives the prediction result from the next state prediction unit 304, the risk value calculation unit 310 refers to the cost information holding unit 311 and calculates the cost value from the current state and the prediction state for each resource. Then, the risk value calculation unit 310 calculates the risk value by summing the calculated cost values for all resources, and sends the calculated risk value to the resource change amount calculation device 20.

  The resource change amount calculation unit 22 of the resource change amount calculation device 20 calculates the risk value received from the next state prediction unit 304 for each resource whose state has changed, the amount of resources to be added or returned, and Then, it is sent to the autonomous control device 13 together with the likelihood of the transition function received from the next state prediction unit 304.

  The autonomous control device 13 that has received this calculates the parameter k using, for example, the following equation (1), and if the calculated parameter k is a predetermined value (for example, 30) or more, The monitoring target system 14 is controlled based on the amount of resources to be added or returned received from the change amount calculation device 20.

例えば、予測結果に従ってリソースの変更を行ったとしても、遷移関数の尤度が低ければ、予測が外れて無駄な制御を行ってしまう場合がある。また、遷移関数の尤度がそれほど低くない場合であっても、リスク値が高ければ、予測結果が誤っていた場合の損失が大きくなる場合がある。リソース状態予測装置３０は、予測結果に従って監視対象システム１４のリソースの量を変更するか否かを判定するための判断材料として、遷移関数の尤度だけでなく、リスク値も算出して自律制御装置１３に提供する。これにより自律制御装置１３は、より多くの観点で、リソース変更量算出装置２０からのリソースの変更量を採用するか否かを判定することができる。

For example, even if the resource is changed according to the prediction result, if the likelihood of the transition function is low, the prediction may be missed and useless control may be performed. Even if the likelihood of the transition function is not so low, if the risk value is high, the loss when the prediction result is incorrect may increase. The resource state prediction device 30 calculates not only the likelihood of the transition function but also the risk value as a judgment material for determining whether or not to change the resource amount of the monitoring target system 14 according to the prediction result, and autonomous control. Provided to the device 13. Thereby, the autonomous control device 13 can determine whether or not to adopt the resource change amount from the resource change amount calculation device 20 from more viewpoints.

  FIG. 25 is a flowchart illustrating an example of the prediction process (S200) in the third embodiment. Except for the points described below, in FIG. 25, the processes denoted by the same reference numerals as those in FIG. 16 are the same as or similar to the processes in FIG.

  After the process shown in step S207 or S213 is executed, the risk value calculation unit 310 refers to the cost information holding unit 311 and calculates a cost value from the current state and the predicted state for each resource (S220). ). Then, the next state prediction unit 304 sends the created prediction result to the resource change amount calculation device 20 and the display unit 300, and the risk value calculation unit 310 sends the calculated cost value to the resource change amount calculation device 20 (S221). . Thereafter, the display unit 300 executes the process shown in step S209.

  Heretofore, the third embodiment of the present invention has been described.

  Next, a fourth embodiment of the present invention will be described.

  The resource state prediction apparatus 30 in the fourth embodiment has the same configuration as the resource state prediction apparatus 30 in the second embodiment described with reference to FIG. However, the contents of the data stored in the device configuration information holding unit 309 are partially different, and accordingly, the processing of the transition rule extracting unit 302, the transition function creating unit 303, and the next state predicting unit 304 is partially different. ing.

  For example, as illustrated in FIG. 26, the device configuration information holding unit 309 stores, for each resource ID 3090, a resource name 3091 corresponding to the resource ID 3090 and a group ID 3093 that identifies the group to which the resource belongs. . A group is, for example, a resource group incorporated in one device such as a server, a resource group assigned to the same business process, or a resource group assigned to the execution of the same service. . Data in the device configuration information holding unit 309 is set in advance by a designer of the monitoring target system 14 or the like.

  When the transition rule extraction unit 302 receives a transition rule collection start instruction from the reception unit 307, the transition rule extraction unit 302 refers to the device configuration information holding unit 309 for each creation time included in the transition rule collection start instruction. For each resource associated with the same group ID, the state of each resource at two consecutive times is sequentially read from the resource state holding unit 301 to extract a transition rule. Then, the transition rule extraction unit 302 sends the extracted transition rule to the transition function creation unit 303 for each resource associated with the same group ID.

  The transition function creation unit 303 creates a transition function for each resource associated with the same group ID, and sends the created transition function to the next state prediction unit 304. The next state prediction unit 304 uses the transition function created by the transition function creation unit 303 for each resource associated with the same group ID to predict the state of each resource at the time point instructed by the user. .

  By making highly related resources the same group, it is possible to reproduce a state transition closer to the behavior of an actual system. Thereby, the precision of the created transition function can be increased.

  FIG. 27 is a flowchart illustrating an example of the prediction process (S200) in the fourth embodiment. Except for the points described below, in FIG. 27, the processes denoted by the same reference numerals as those in FIG. 16 are the same as or similar to the processes in FIG.

  After resetting the timer in step S202, the transition rule extraction unit 302 refers to the device configuration information holding unit 309, selects one unselected group (S230), and continues for resources belonging to the selected group. The state of each resource at each of the two times is sequentially read from the resource state holding unit 301 (S203).

  In step S206, when the likelihood of the transition function calculated by the transition function creation unit 303 is equal to or greater than the second threshold (S206: Yes), the transition function creation unit 303 corresponds to the created transition function. The information is sent to the next state prediction unit 304 together with the group ID of the group and the calculated likelihood. The next state prediction unit 304 refers to the resource state holding unit 301 and uses the transition function received from the transition function creation unit 303 for the resource associated with the group ID received from the transition function creation unit 303, and the user The state of each resource at the time point instructed from is predicted (S231).

  In step S231, after the state of the resource belonging to the selected group is predicted, or in step S211, an error is notified by the transition rule extraction unit 302, the transition rule extraction unit 302 includes the resource state holding unit 301. , It is determined whether all groups have been selected (S232).

  When all the groups have been selected (S232: Yes), the next state prediction unit 304 creates a prediction result including the state of each resource and the likelihood of the transition function at the time point instructed by the user (S208). When any group is not selected (S232: No), the transition rule extraction unit 302 executes the process shown in step S230 again.

  In step S201, when it is not the creation timing of the transition function (S201: No), the next state prediction unit 304 refers to the resource state holding unit 301 and is created last time by the transition function creation unit 303 for each group. Using the transition function, the state of each resource at the time point instructed by the user is predicted (S233), and the process shown in step S208 is executed.

  Heretofore, the fourth embodiment of the present invention has been described.

  Note that the resource state prediction apparatus 30 in each of the embodiments described above is realized by a computer 80 having a configuration as shown in FIG. 28, for example. FIG. 28 is a hardware configuration diagram illustrating an example of a computer 80 that implements the functions of the resource state prediction device 30. The computer 80 includes a central processing unit (CPU) 81, a random access memory (RAM) 82, a read only memory (ROM) 83, a hard disk drive (HDD) 84, a communication interface (I / F) 85, an input / output interface (I). / F) 86 and a media interface (I / F) 87.

  The CPU 81 operates based on a program stored in the ROM 83 or the HDD 84 and controls each unit. The ROM 83 stores a boot program executed by the CPU 81 when the computer 80 is started up, a program depending on the hardware of the computer 80, and the like. The HDD 84 stores a program executed by the CPU 81. The communication interface 85 receives data from other devices via a communication line and sends the data to the CPU 81, and transmits data generated by the CPU 81 to other devices via the communication line.

  The CPU 81 controls the display device 15 and the input device 16 via the input / output interface 86. The CPU 81 acquires data from the input device 16 via the input / output interface 86. Further, the CPU 81 outputs the generated data to the display device 15 via the input / output interface 86.

  The media interface 87 reads a program or data stored in the recording medium 88 and provides it to the CPU 81 via the RAM 82. The CPU 81 loads the program from the recording medium 88 onto the RAM 82 via the media interface 87, and executes the loaded program. The recording medium 88 is, for example, an optical recording medium such as a DVD (Digital Versatile Disk) or PD (Phase change rewritable disk), a magneto-optical recording medium such as an MO (Magneto-Optical disk), a tape medium, a magnetic recording medium, or a semiconductor memory. Etc.

  When the computer 80 functions as the resource state prediction apparatus 30 of the first embodiment, the CPU 81 of the computer 80 executes a program loaded on the RAM 82, thereby causing the display unit 300, the resource state holding unit 301, the transition rule. The functions of the extraction unit 302, the transition function creation unit 303, the next state prediction unit 304, the measurement value collection unit 305, the threshold information holding unit 306, the reception unit 307, and the measurement number calculation unit 308 are realized. Further, the RAM 82 or the HDD 84 stores data in the resource state holding unit 301 and the threshold information holding unit 306.

  Further, when the computer 80 functions as the resource state prediction device 30 of the second or fourth embodiment, the CPU 81 of the computer 80 executes the program loaded on the RAM 82, thereby causing the display unit 300 to hold the resource state. Unit 301, transition rule extraction unit 302, transition function creation unit 303, next state prediction unit 304, measurement value collection unit 305, threshold information storage unit 306, reception unit 307, measurement frequency calculation unit 308, and device configuration information storage unit 309 Each function is realized. The RAM 82 or the HDD 84 stores data in the resource state holding unit 301, the threshold information holding unit 306, and the device configuration information holding unit 309.

  Further, when the computer 80 functions as the resource state prediction apparatus 30 of the third embodiment, the CPU 81 of the computer 80 executes a program loaded on the RAM 82, thereby causing the display unit 300, the resource state holding unit 301, Transition rule extraction unit 302, transition function creation unit 303, next state prediction unit 304, measurement value collection unit 305, threshold information holding unit 306, reception unit 307, measurement frequency calculation unit 308, risk value calculation unit 310, and cost information holding Each function of the unit 311 is realized. The RAM 82 or the HDD 84 stores data in the resource state holding unit 301, the threshold information holding unit 306, and the cost information holding unit 311.

  The computer 80 reads these programs from the recording medium 88 and executes them, but as another example, these programs may be acquired from other devices via a communication medium. The communication medium refers to a communication line or a digital signal or a carrier wave that propagates through the communication line.

  In addition, this invention is not limited to above-described embodiment, Many deformation | transformation are possible within the range of the summary.

  For example, in each of the above-described embodiments, the measurement value collection unit 305 discretizes each resource into two states, but the present invention is not limited to this and may be discretized into three or more states. Further, in each of the above-described embodiments, the resource state prediction device 30 uses the three cells that are consecutively adjacent in the state of each resource at the time of each resource at two times. However, the present invention is not limited to this, and in each previous resource state, using 2n + 1 (n ≧ 2) or more cells that are consecutively adjacent to each other, The state after the center cell of the 2n + 1 cells may be predicted.

  In each embodiment described above, when the likelihood of the transition function calculated by the transition function creation unit 303 is less than the second threshold, the transition rule extraction unit 302 further determines the measured value of each resource at the previous time. Although the transition rule is further extracted using, the present invention is not limited to this. For example, the transition rule extraction unit 302 increases the number of transition rule extraction targets by increasing the number of transition rule extraction targets in addition to increasing the number of transition rule extraction targets by tracing back the past state of each resource. You may do it.

It is a system configuration figure showing the composition of control system 10 concerning one embodiment of the present invention. It is a block diagram which shows an example of a function structure of the resource state prediction apparatus 30 in 1st Embodiment. 6 is a diagram illustrating an example of the structure of data stored in a threshold information holding unit 306. FIG. FIG. 4 is a diagram illustrating an example of the structure of data stored in a resource state holding unit 301. It is a conceptual diagram which shows an example of the collection of the cell handled as a cellular automaton. It is a figure which shows an example of the transition rule collected for every resource by the transition rule extraction part 302. FIG. It is a conceptual diagram which shows an example of the collection of the cell in the case of extracting a transition rule again. It is a figure which shows the ratio of the subsequent state in a center cell about the transition rule with the same state of the previous cell. It is a figure which shows an example of the transition function calculated by the transition function preparation part 303. FIG. It is a conceptual diagram which shows the process in which the state of each resource after predetermined time is estimated using the transition function calculated by the transition function preparation part 303. FIG. It is a figure which shows an example of the structure of the data output from the next state estimation part. It is a conceptual diagram which shows an example of the prediction result screen 60 displayed on the display apparatus 15 by the display part 300. FIG. 3 is a block diagram illustrating an example of a functional configuration of a resource change amount calculation device 20. FIG. It is a figure which shows an example of the structure of the data stored in the change resource extraction part. It is a flowchart which shows an example of operation | movement of the control system 10 in 1st Embodiment. It is a flowchart which shows an example of the prediction process (S200) in 1st Embodiment. It is a block diagram which shows an example of a function structure of the resource state prediction apparatus 30 in 2nd Embodiment. It is a figure which shows an example of the structure of the data stored in the apparatus structure information holding | maintenance part 309 in 2nd Embodiment. It is a conceptual diagram which shows an example of the connection relation of each resource. It is a figure which shows an example of the matrix 72 which shows the distance between each resource. It is a figure which shows an example of the matrix 73 at the time of arrange | positioning each resource so that the distance between adjacent resources may become the minimum. It is a flowchart which shows an example of operation | movement of the control system 10 in 2nd Embodiment. It is a block diagram which shows an example of a function structure of the resource state prediction apparatus 30 in 3rd Embodiment. An example of the structure of data stored in the cost information holding unit 311 is shown. It is a flowchart which shows an example of the prediction process (S200) in 3rd Embodiment. It is a figure which shows an example of the structure of the data stored in the apparatus structure information holding | maintenance part 309 in 4th Embodiment. It is a flowchart which shows an example of the prediction process (S200) in 4th Embodiment. FIG. 3 is a hardware configuration diagram illustrating an example of a computer 80 that realizes the function of the resource state prediction device 30.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 ... Control system, 11 ... Client computer, 12 ... Communication line, 13 ... Autonomous control device, 14 ... Monitored system, 15 ... Display device, 16 ... Input device , 20 ... Resource change amount calculation device, 21 ... Change amount acquisition unit, 22 ... Resource change amount calculation unit, 23 ... Resource information holding unit, 30 ... Resource state prediction device, 300 Display unit 301 Resource state holding unit 302 Transition rule extraction unit 303 Transition function creation unit 304 Next state prediction unit 305 Measurement value collection unit 306: Threshold value holding unit, 307 ... Accepting unit, 308 ... Measurement number calculating unit, 309 ... Device configuration information holding unit, 310 ... Risk value calculating unit, 311 ... Cost information Holding unit, 40... Transition , 42 ... previous state, 43 ... subsequent state, 44 ... transition function, 50 ... prediction result, 51 ... resource ID, 52 ... current state, ..Prediction state, 54 ... Likelihood, 60 ... Prediction result screen, 80 ... Computer, 81 ... CPU, 82 ... RAM, 83 ... ROM, 84 ... HDD, 85 ... Communication interface, 86 ... Input / output interface, 87 ... Media interface, 88 ... Recording medium

Claims (9)

A resource state prediction device that predicts a state after each resource based on the current state of each resource in the monitored system and outputs a prediction result,
A measurement value collecting means for collecting measurement values of each resource from the monitored system and discretizing each measurement value into one of a plurality of states using a predetermined threshold for each resource;
Resource state holding means for holding the state of each resource collected and discretized by the measurement value collecting means in association with the time when the resource was measured;
The state of each resource at two consecutive times is sequentially read from the resource state holding means and arranged in one column at each time, each resource is regarded as one cell in the cellular automaton, and each resource at the previous time is In a state, for each predetermined number of adjacent cells including the target cell, the state of each of the predetermined number of cells is set as an input state, and becomes the target in the state of each resource at a later time A transition rule extracting means for extracting a transition rule having a cell state as an output state for each of all cells;
Of the transition rules extracted by the transition rule extracting means, the ratio of different output states is calculated for each transition rule having the same input state, and the highest output state is output in the transition rule. A transition function creating means for creating a transition function composed of a plurality of transition rules including the adopted output state,
Accepting means for accepting input of time from the current time to the time at which the state of each resource should be predicted;
A measurement number calculation means for calculating a measurement number, which is the number of times a measurement value of each resource is measured, from the current time until the time input via the reception unit elapses;
The current state of each resource is sequentially read out from the measurement value holding means and arranged in one column, each resource is regarded as one cell in the cellular automaton, and a predetermined number of cells adjacent to each other including the target cell Each time, a transition rule having the respective states of the predetermined number of cells as input states is extracted from the transition function created by the transition function creating means, and the output state included in the extracted transition rule is A next state predicting means for executing processing for calculating the state at the next time of the target cell as a target for each of all cells;
The next state predicting means further includes:
With reference to the transition function, the process of calculating the next state of each resource from the calculated next state of each resource is repeated by the number of times of measurement calculated by the number-of-measurement calculating means minus one, and finally A resource state prediction apparatus that outputs the calculated state of each resource to the outside as the state of each resource after a time input via the receiving unit has elapsed. The resource state prediction apparatus according to claim 1,
The measurement value collecting means includes
The measurement value of each resource collected every predetermined time is discretized into one of two states using a threshold value predetermined for each resource,
The transition rule extraction means includes
In the state of each resource at the previous time, for each of three consecutive cells centered on the target cell, the state of each of the three cells is set to the input state, and In the state, a transition rule whose output state is the state of the target cell is extracted for each of all the cells,
The next state prediction means includes:
Extraction is made from the transition function created by the transition function creating means for each of three consecutive cells centered on the target cell, with the state of each of the three cells as the input state. And performing a process for calculating the output state included in the extracted transition rule as the state at the next time of the target cell for each of all the cells. apparatus. The resource state prediction apparatus according to claim 1,
The transition function creating means includes
Referring to the ratio of output states adopted for each transition rule, the ratio of transition rules including output states that are equal to or higher than a predetermined ratio among the plurality of transition rules constituting the transition function Is further calculated as a likelihood indicating the certainty and output together with the transition function,
The next state prediction means includes:
A resource state characterized in that the likelihood of the transition function calculated by the transition function creating means is output to the outside together with information indicating the state of each resource after the time input via the accepting means has elapsed. Prediction device. The resource state prediction apparatus according to claim 3,
The transition rule extraction means includes
When the likelihood of the transition function is less than a predetermined value, the previous state of each resource is read out from the resource state holding unit, and the previous time in the state of each resource at the two consecutive times The state of each resource is set to the state of each resource at a later time, the state of each read resource is set to the state of each resource at the previous time, and further transition rules are extracted.
The transition function creating means includes
When the likelihood of the transition function is less than a predetermined value, the transition rule further extracted by the transition rule extracting unit is further used to change the output state different for each transition rule having the same input state. A resource state prediction apparatus characterized by calculating a ratio of The resource state prediction apparatus according to claim 1,
Each resource further comprises distance information holding means for holding information indicating the distance to other resources,
The transition rule extraction means and the next state prediction means are:
When sequentially reading out the state of each resource from the resource state holding unit and arranging them in a line, the state of each resource is arranged so that the distance between adjacent resources is minimized with reference to the distance information holding unit. A resource state prediction apparatus characterized by The resource state prediction apparatus according to claim 1,
Cost information holding means for holding a cost value when the resource state has changed and a cost value when the resource state has not changed;
For each resource, the cost information holding means is calculated from the current state of each resource and the state of each resource after the time input via the accepting means calculated by the next state predicting means. The system further comprises a risk value output means for calculating a cost value for each resource, calculating a risk value by summing the calculated cost values, and outputting the calculated risk value to the outside. Resource state prediction device. The resource state prediction apparatus according to claim 1,
For each resource, it further comprises group information holding means for holding a group ID for identifying a group to which the resource belongs,
The transition rule extraction means includes
With reference to the group information holding means, extract a transition rule for each of a plurality of resources belonging to the same group,
The transition function creating means includes
Refer to the group information holding means, create a transition function for each of a plurality of resources belonging to the same group,
The next state prediction means includes:
Resource state prediction characterized by outputting the state of each resource after the time input via the accepting unit has elapsed for each of a plurality of resources belonging to the same group with reference to the group information holding unit apparatus. A resource state prediction method in a resource state prediction device that predicts a state after each resource based on the current state of each resource in a monitored system and outputs a prediction result,
The resource state prediction apparatus
Collect measurement values of each resource from the monitored system, discretize each measurement value into one of a plurality of states using a predetermined threshold for each resource, and at the time when the resource is measured A measurement value collecting step for storing in the resource state holding means in association;
The state of each resource at two consecutive times is sequentially read from the resource state holding means and arranged in one column at each time, each resource is regarded as one cell in the cellular automaton, and each resource at the previous time is In a state, for each predetermined number of adjacent cells including the target cell, the state of each of the predetermined number of cells is set as an input state, and becomes the target in the state of each resource at a later time A transition rule extracting step for extracting a transition rule having a cell state as an output state for each of all cells;
Of the transition rules extracted in the transition rule extraction step, the ratio of different output states is calculated for each transition rule that has the same input state, and the highest output state is determined as the output state in the transition rule. A transition function creating step that creates a transition function composed of a plurality of transition rules including the adopted output state,
A reception step for receiving an input of time from the current time to a time at which the state of each resource should be predicted;
A measurement count calculation step for calculating a measurement count that is the number of times the measurement value of each resource is measured before the time received in the reception step elapses from the current time;
The current state of each resource is sequentially read out from the measurement value holding means and arranged in one column, each resource is regarded as one cell in the cellular automaton, and a predetermined number of cells adjacent to each other including the target cell Each time, a transition rule having each state of the predetermined number of cells as an input state is extracted from the transition function created in the transition function creating step, and the output state included in the extracted transition rule is taken as the target A process of calculating the state at the next time of the cell to be executed as a next state prediction step for executing all of the cells as targets,
In the resource state prediction apparatus, in the next state prediction step,
The process of calculating the next state of each resource from the calculated next state of each resource with reference to the transition function is repeated by the number of times of measurement calculated in the measurement number calculation step—one time and finally calculated. A resource state prediction method characterized in that the state of each resource is output to the outside as the state of each resource after the time received in the receiving step has elapsed. A program for causing a computer to function as a resource state prediction device that predicts a state after each resource based on the current state of each resource in the system to be monitored and outputs a prediction result,
In the computer,
A measurement value collection function that collects measurement values of each resource from the monitored system and discretizes each measurement value into one of a plurality of states using a predetermined threshold for each resource;
A resource state holding function for holding the state of each resource collected and discretized by the measurement value collecting function in association with the time when the resource is measured;
The state of each resource at two consecutive times is sequentially read out from the resource state holding function and arranged in one column at each time, each resource is regarded as one cell in the cellular automaton, and each resource at the previous time is In a state, for each predetermined number of adjacent cells including the target cell, the state of each of the predetermined number of cells is set as an input state, and becomes the target in the state of each resource at a later time Transition rule extraction function that extracts the transition rule with the cell state as the output state for each of all cells,
Of the transition rules extracted by the transition rule extraction function, the ratio of different output states is calculated for each transition rule having the same input state, and the highest output state is output in the transition rule. Transition function creation function that creates a transition function composed of multiple transition rules including the adopted output state,
A reception function that accepts input of time from the current time to the time when the state of each resource should be predicted,
A measurement count calculation function for calculating a measurement count, which is the number of times a measurement value of each resource is measured, from the current time until the time input via the reception function elapses; and
The current state of each resource is sequentially read out from the measurement value holding function and arranged in one column, each resource is regarded as one cell in the cellular automaton, and a predetermined number of adjacent cells including the target cell Each time, a transition rule having each state of the predetermined number of cells as an input state is extracted from the transition function created by the transition function creation function, and the output state included in the extracted transition rule is Realize the next state prediction function that executes the process of calculating the state at the next time of the target cell for each of all cells,
The next state prediction function further includes:
The process of calculating the next state of each resource from the calculated next state of each resource with reference to the transition function is repeated by the number of measurement times calculated by the measurement number calculation function minus one, and finally A program which outputs the calculated state of each resource to the outside as the state of each resource after the time input via the reception function has elapsed.

Images