JP2015005191A - Batch performance predicting and performance optimization supporting method, and system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To reduce the number of man-hours spent on responses for executing efficient performance optimization through accuracy enhancement by accepting user correction of batch performance prediction and limiting the batches subject to the performance optimization.SOLUTION: A computer is caused to calculate unit prices of performances and the past average of processing time lengths on the basis of actual past records of batch operation, a screen interface that executes processing time prediction using such statistical indicators as correlation coefficient and variation coefficient on the basis of the DB-derived number of cases of processing as of the previous day, and makes user correction based on the result of prediction, is prepared, and thereby efficient execution of performance optimization centering on batches taking long times to predict processing is achieved.

Description

  The present invention relates to a performance prediction method and a performance countermeasure method in a system that executes batch application processing, and in particular, efficiently predicts performance in a system that needs to execute a large amount of batch application processing in a predetermined time zone such as at night. And a method for providing countermeasure support.

  In the application system, a task that normally requires real-time performance is realized by online processing, and a task that processes a large amount of data, such as a task that is not so, or a batch processing of a day, is realized by batch processing. Due to its characteristics, batch processing is often required to be completed in a limited time zone (hereinafter referred to as a batch executable time zone) excluding an online operation time zone and a maintenance time zone. For this reason, it is required to grasp in advance whether or not the batch processing is completed in the batch executable time zone, and to implement performance measures for completion if not completed.

  In Patent Literature 1, load information in a time zone in which batch processing is executed is acquired from the load information DB, an execution date that approximates load information is identified from past execution dates, and the execution time of the execution date is determined. A technique for predicting a processing time is disclosed.

JP 2012-252422 A

  Although the performance prediction of batch processing is performed by the above-described known technology or the like, this known technology does not mention application of measures based on the performance prediction result. Normally, batch processing is executed every day at a fixed time, and even if the number of batch applications increases, it is required to perform performance prediction to countermeasure implementation in daily operation. It is necessary to operate efficiently up to performance measures. This invention makes it a subject to solve this point.

  The present invention is a system that executes a plurality of batch application processes with order in a predetermined batch executable time zone, and the batch performance prediction and countermeasure support server includes a copy of DB information of the previous day. -Obtain the number of processings of each batch application as of the previous day from the DB server by SQL, and multiply each by the performance unit price (processing time per processing item) calculated from the past batch operation results (number of processing times x The processing time for all batches is calculated.

  The batch performance prediction and countermeasure support server calculates a correlation coefficient between the number of processing cases and processing time (processing time of batch processing) from the past batch operation results.

Here, the correlation coefficient in the present invention refers to the number of processes x, the processing time required to execute x processes (total processing time) as y, and the relationship between x and y as y = f ( x) = it is that the index k when expressed as Cx ^k. In general, when the number of processing increases, the load on the computer increases when the processing is executed. Therefore, f (x) is an upwardly convex graph, k takes a value of 0 <k <1, and the load is small. k is substantially equal to 1 (the overall processing time is proportional to the number of processing cases), and the value of k decreases as the load increases (the overall processing time gradually reaches its peak as the number of processing cases increases). The constant C is a coefficient determined by the processing mode of batch processing.

If Y = logy, X = logx, and C ′ = logC, Y = kX + C ′. Therefore, the processing time when the number of processing cases is changed is measured, and the logarithms X and Y of these values are used as variables. The logarithm C ′ of the coefficient k and the constant C (that is, the constant C) is obtained by the least square method. Note that the coefficient k determined by the least squares method, because it contains the correlation coefficient _{k xy} data strings _{{X i (= logx i)} } and _{{Y i (= logy i)} } ( i.e., k = (σ _y / σ _x ) k _xy , where σ _x and σ _y are standard deviations of X _i (= logx _i ) and Y _i (= logy _i ), respectively, in the present invention, The index k indicating the relationship of the processing time y is referred to as a correlation coefficient.

If the correlation coefficient is greater than or equal to the threshold value, the calculated processing time is adopted as a predicted value. Otherwise, the average processing time for a specific day of the week or day in the past is used as the predicted value from the batch operation results. Four axes: next (run on the same day in another week), monthly (run on the same day in another month), and year (run on the same day in another year) The average value of the processing times calculated with the axis having the variation coefficient closest to 0 among them is adopted as the predicted value. Here, the variation coefficient is the ratio of the standard deviation σ _x of the variable x to the average value <x> of a certain variable x, that is, σ _x / <x>, and is the relative ratio of the error to the average value.

  Batch performance prediction and countermeasure support based on the adopted processing time prediction value, taking into account correction information on the number of processes input by the user using the operation terminal, correction information on the processing number dependency ratio and processing time performance dependency ratio The server recalculates the processing time, and uses the batch execution definition defined in the batch execution management server to calculate the end time of all the batches for the current day. If the calculated predicted value of the batch end time is not within the batch executable time zone, the execution of performance countermeasures is instructed.

  In order to execute performance measures, the batch performance prediction and countermeasure support server sets the performance measures for all batches set in advance for each batch, the predicted processing time when the performance measures are applied to the batches targeted for performance measures, and the The end time is calculated from the batch operation results and displayed on the operation terminal screen. Based on this display content, the user selects and confirms the performance countermeasure target batch and the performance countermeasure contents of the performance countermeasure target batch, and the batch performance prediction and countermeasure support server instructs the performance countermeasure set based on this information.

  After the batch execution, the processing time, the number of processing, and the results of performance measures implemented for each batch are saved as the batch operation results for that day. The correlation coefficient and performance unit price are recalculated.

  According to the present invention, it is possible to avoid an event in which batch processing exceeds the batch processable time zone. In addition, it is possible to reduce the labor of the operator for performance measures from performance prediction.

It is a figure which shows the example of whole structure of the computer network system 1 which concerns on this embodiment. 2 is a diagram illustrating a hardware configuration example of a DB server 10. FIG. It is a figure which shows the hardware structural example of the HUB-DB server 11. It is a figure which shows the hardware structural example of the batch server 12. FIG. 2 is a diagram illustrating an example of a hardware configuration of a batch execution management server 13. FIG. It is a figure which shows the hardware structural example of a batch performance prediction and countermeasure assistance server. 2 is a diagram illustrating a hardware configuration example of an operation terminal 15. FIG. It is a figure which shows the example of the batch execution definition 503 which the batch execution management server 13 comprises. It is a figure which shows the data structural example of the master table 619 which stores the performance countermeasure method of each batch with which the batch performance prediction and countermeasure support server 14 comprises. It is a figure which shows the example of the table 620 showing the processing number of each batch which the batch performance prediction and countermeasure support server 14 acquires from various business tables 301 and 302 of the HUB-DB server 11 and is provided. It is a figure which shows the data structural example of the table 621 which stores the performance unit price (processing time per process number) of each batch which the batch performance prediction and countermeasure support server 14 comprises. It is a figure which shows the data structural example of the table 622 which stores the process performance of each batch which the batch performance prediction and countermeasure assistance server 14 comprises, processing time, and the past results of performance countermeasure implementation content for every day. It is a figure which shows the data structural example of the work table 623 which stores the progress of the prediction processing time of each batch which the batch performance prediction and countermeasure support server 14 comprises. It is a figure which shows the data structural example of the table 624 which stores the correlation coefficient of the processing number of each batch which the batch performance prediction and countermeasure assistance server 14 comprises, and processing time. It is a figure which shows the data structural example of the table 625 which stores the average processing time when the processing time of each batch with which the batch performance prediction and countermeasure support server 14 is seen at a daily timing, and the variation coefficient of processing time. Data structure of table 626 for storing average processing time and coefficient of variation of processing time when processing time of each batch provided in batch performance prediction and countermeasure support server 14 is viewed at weekly timing (same day of week) It is a figure which shows an example. Data structure of a table 627 for storing the average processing time and the coefficient of variation of the processing time when the processing time of each batch provided in the batch performance prediction and countermeasure support server 14 is viewed at monthly timing (same date with different months) It is a figure which shows an example. Data structure of table 628 storing average processing time and coefficient of variation of processing time when processing time of each batch provided in batch performance prediction and countermeasure support server 14 is viewed at an annual timing (same date in different years) It is a figure which shows an example. It is a figure which shows the data structural example of the table 629 which stores the calculation result of the prediction processing time of each batch which the batch performance prediction and countermeasure support server 14 comprises. It is a figure which shows the flowchart which shows the whole process of this invention. It is a figure which shows the detailed flowchart of reflection performance S2001 of the execution performance of the whole flowchart shown in FIG. It is a figure which shows the detailed flowchart of calculation S2005 of the prediction processing time by the tendency of the past performance of the whole flowchart shown in FIG. It is a figure which shows the detailed flowchart of user correction result application S2006 of the prediction processing time of the whole flowchart shown in FIG. It is a figure which shows the process number correction screen example 2400 displayed on the display part 704 of the operation terminal 15. FIG. It is a figure which shows the example of correction | amendment screens 2500 by the process characteristic displayed on the display part 704 of the operation terminal 15. FIG. It is a figure which shows the batch completion scheduled time display screen example 2600 displayed on the display part 704 of the operation terminal 15. FIG. It is a figure which shows the detailed flowchart of execution S2009 of the performance countermeasure assistance process of the whole flowchart shown in FIG. It is a figure which shows the performance countermeasure assistance screen example 2800 displayed on the display part 704 of the operation terminal 15. FIG.

  Hereinafter, the present embodiment will be described in detail with reference to the drawings. Moreover, in all the drawings used in the description of the present embodiment, the same parts are denoted by the same reference symbols in principle, and the repeated description thereof is omitted.

  FIG. 1 is a diagram illustrating an example of the overall configuration of a computer network system 1 according to the present embodiment.

  The batch server 12 that executes the batch application, the batch execution management server 13 that instructs the batch server 12 to execute the batch application based on the definition information of the execution order of the batch application, and the DB table information of the DB server 10 at the end of the previous day. HUB-DB server 11 that is copied and held, batch performance prediction and countermeasure support server 14 for realizing batch performance prediction and performance countermeasure support (subject of the processing operation of this embodiment), operation terminal used by the operator These are composed of 15 connected to the network 16. In the present embodiment, the configuration example of FIG. 1 is used for ease of explanation, but each server configuration can be realized in either a coexisting or independent form. For example, the DB server 10 and the HUB-DB server 11 are configured by one server, the batch server 12 and the batch execution management server 13 are configured by one server, and further, the DB server 10, the HUB-DB server 11, The batch server 12 and the batch execution management server 13 may be configured as a single server.

  FIG. 2 is a hardware configuration example of the DB server 10. The DB server 10 includes a number of DB tables 201 and 202 in the storage unit 200. These business tables are used by a batch application or the like provided in the batch server 12, but the number of business tables and their contents are not limited to the present embodiment. The network interface unit 204 is connected to the network 16. The control unit 203 is a CPU.

  FIG. 3 is a hardware configuration example of the HUB-DB server 11. The HUB-DB server 11 includes DB tables 301 and 302 in the storage unit 300. The DB tables 301 and 302 include DB tables 201 and 202 included in the DB server 10 and information obtained by copying all data registered in the tables. Although the present embodiment is not limited to the timing of the replication process, it will be described as being replicated at the end of the previous day's batch processing. The network interface unit 304 is connected to the network 16. The control unit 303 is a CPU.

  FIG. 4 is a hardware configuration example of the batch server 12. The storage unit 400 of the batch server 12 includes batch execution management middleware 402 and batch applications (programs) 403 and 404. There are a plurality of batch applications 403 and 404, and each batch application is executed based on the batch execution definition 503 provided in the storage unit 500 of the batch execution management server 13. The batch applications 403 and 404 are executed only in the memory 408 of the batch server 12 or refer to / update the DB tables 201 and 202 provided in the storage unit 200 of the DB server 10 from the network interface unit 404 through the network 16. Although the process of creating / deleting / deleting is executed, the present embodiment is not limited to the contents and number of batch applications 403 and 404. The control unit 406 is a CPU.

  FIG. 5 is a hardware configuration example of the batch execution management server 13. The batch execution management server 13 includes a batch execution management middleware 502 in the storage unit 500, and a batch execution definition 503 in the batch execution middleware 502. An image of the batch execution definition 503 is shown in FIG. The batch execution definition 503 is information that determines the execution order and execution timing of the batch applications 403 and 404 included in the batch server 12. Execution order refers to, for example, batch B and batch C being executed after completion of batch A (800) as shown in mm month 1 day batch execution definition 800 in FIG. Some batches, such as batch N and batch X, have no execution order (803). Execution timing includes daily, weekly (801, processing Z is performed after processing S and Y is completed), monthly, and annual execution dates, and execution such as batch Z is executed at 2:00. There is something that shows time. However, for example, even a batch application executed at the same monthly timing, a batch executed on the first day of each month (Batch B), a batch executed on the third day of every month (Batch M), etc. (802), There are cases where the execution date is different. All batch applications belong to one of the four types of execution timings, but there are batches for which execution time is not specified. The batch execution definition 503 is defined on the batch execution management middleware 502. Based on this information, an execution instruction is sent from the batch execution management middleware 502 to the batch applications 403 and 404 of the batch server 12 through the network 16 via the network interface unit 505. I do. The control unit 504 is a CPU.

  FIG. 6 is a hardware configuration example of the batch performance prediction and countermeasure support server 14. In the storage unit 600 of the batch performance prediction and countermeasure support server 14, as a program 601, a processing number count process 602, a processing time calculation process 603 based on the number of processing cases and a performance unit price, a processing time calculation process 604 based on a past performance trend, a prediction processing time User correction processing 605 (including processing number correction 606 and processing characteristic correction 607), scheduled batch end time calculation processing 608, threshold time determination processing 609, performance countermeasure support processing 610 (measure candidate batch selection 611, countermeasure content selection) 612, countermeasure instruction 613), batch execution result reflection processing 614 (processing number and processing time extraction 615, performance unit price and correlation coefficient recalculation 616, implementation countermeasure storage 617). The DB table 618 of the storage unit 600 includes a performance countermeasure master 619, a processing number table 620, a performance unit price table 621, an operation result table 622, a processing time prediction work table 623, a correlation coefficient table 624, a variation coefficient (daily). ) Table 625, variation coefficient (weekly) table 626, variation coefficient (monthly) table 627, variation coefficient (yearly) table 628, and processing time prediction result table 629. Further, it is connected to the network 16 through the network interface unit 631. The control unit 630 is a CPU.

  FIG. 7 is a hardware configuration example of the operation terminal 15. The storage unit 700 of the operation terminal 15 includes browser software 701 and displays a screen program included in the batch performance prediction and countermeasure support server 14 through the display unit 704. The network interface unit 703 is connected to the network 16. The control unit 702 is a CPU.

  FIG. 9 is a diagram illustrating a data configuration example of the performance countermeasure master 619. The performance countermeasure master 619 stores the batch application name 900, countermeasure candidate A901, countermeasure candidate B902, and countermeasure candidate C903 in association with each other. The batch application name 900 stores the names of all batch applications, and the countermeasure candidates A901, countermeasure candidates B902, and countermeasure candidates C903 store performance countermeasures that can be implemented in each batch for all batches. There are three types of performance measures: split execution (separate batch processing according to the number of processes), execution SKIP (does not perform specific batch processing), and scale-up (improves computer processing performance). Accordingly, priority is set in the order of countermeasure candidate A, countermeasure candidate B, and countermeasure candidate C. Items that cannot be selected as countermeasure candidates are not set. The values stored in each column of the performance countermeasure master 619 are registered in advance and are not updated in the present embodiment.

  FIG. 10 is a diagram illustrating a data configuration example of the processing number table 620. The processing number table 620 stores the batch application name 1000 and the processing number 1001 in association with each other. All batch application names are stored in the batch application name 1000, and the batch performance prediction and the processing support count server 14 count processing for the DB tables 301 and 302 provided in the HUB-DB server 11 for the processing count 1001. The number of processes acquired using 602 is stored. The stored processing number 1001 is referred to in the processing time calculation process 603 based on the processing number and the performance unit price.

  FIG. 11 is a diagram illustrating a data configuration example of the performance unit price table 621. The performance unit price table 621 stores batch application names 1100 and performance unit prices 1101 in association with each other. The batch application name 1100 stores the names of all batch applications, and the performance unit price 1101 is calculated using the batch unit performance prediction and the correlation coefficient recalculation 616 of the batch execution result reflection process 614 of the countermeasure support server 14. The processing time (unit: millisecond / case) per processing number of each batch to be stored is stored. The stored performance unit price 1101 is referred to in the processing time calculation process 603 based on the number of processing cases and the performance unit price.

  FIG. 12 is a diagram illustrating a data configuration example of the operation result table 622. The operation result table 622 stores the date 1200, the batch application name 1201, the processing number 1202, the processing time (seconds) 1203, and the performance countermeasure implementation content 1204 in association with each other. The date 1200 stores the number of batch execution results reflecting processing 614 provided in the batch performance prediction and countermeasure support server 14 and the past batch execution date acquired using the processing time extraction 615, and the batch application The name 1201 stores the names of all batch applications executed on the batch execution date 1200, and the processing performance number 1202 includes the batch performance prediction and countermeasure support server 14 based on the application log 405 included in the batch server 12. The processing number of each batch application acquired by using the processing number counting process 602 is stored, and the processing time (seconds) 1203 is a batch performance prediction and countermeasure based on the application log 405 provided in the batch server 12 Reflecting the batch execution results of the support server 14 Processing time for each batch application that retrieves using the processing number and the processing time extracting 615 sense 614 is stored. The stored value of each item is referred to in processing time calculation processing 604 based on past performance trends, processing number correction 606 in user correction processing 605 for predicted processing time, and countermeasure candidate batch selection 611 in performance countermeasure support processing 610.

  FIG. 13 is a diagram illustrating a data configuration example of the processing time prediction work table 623. The processing time prediction work table 623 stores the batch application name 1300 and the predicted processing time (seconds) 1301 in association with each other. The batch application name 1300 stores the names of all batch applications, and the estimated processing time (seconds) 1301 is a processing time calculation process 603 based on the number of processes and the performance unit price provided in the batch performance prediction and countermeasure support server 14. The predicted processing time to be calculated is stored. The stored prediction processing time 1301 is referred to by the correction 606 by the processing characteristic of the user correction processing 605 of the prediction processing time.

  FIG. 14 is a diagram illustrating a data configuration example of the correlation coefficient table 624. The correlation coefficient table 624 stores the batch application name 1400 and the correlation coefficient 1401 in association with each other. The batch application name 1400 stores the names of all batch applications, and the correlation coefficient 1401 recalculates the performance unit price and correlation coefficient of the batch execution result reflection processing 614 provided in the batch performance prediction and countermeasure support server 14. The correlation coefficient between the processing number and processing time calculated based on the processing number 1202 and processing time 1203 of each batch stored in the operation result table 622 using 616 is stored.

Here, the correlation coefficient in this embodiment means that the number of processing cases is x, the processing time (total processing time) required to execute x processings is y, and the relationship between x and y is y = f. It is the index k when expressed as (x) = Cx ^k . In general, when the number of processing increases, the load on the computer increases when the processing is executed. Therefore, f (x) is an upwardly convex graph, k takes a value of 0 <k <1, and the load is small. k is substantially equal to 1 (the overall processing time is proportional to the number of processing cases), and the value of k decreases as the load increases (the overall processing time gradually reaches its peak as the number of processing cases increases). The constant C is a coefficient determined by the processing mode of batch processing.

If Y = logy, X = logx, and C ′ = logC, Y = kX + C ′. Therefore, the processing time when the number of processing cases is changed is measured, and the logarithms X and Y of these values are used as variables. The logarithm C ′ of the coefficient k and the constant C (that is, the constant C) is obtained by the least square method. Note that the coefficient k determined by the least squares method, because it contains the correlation coefficient _{k xy} data strings _{{X i (= logx i)} } and _{{Y i (= logy i)} } ( i.e., k = (σ _y / σ _x ) k _xy , where σ _x and σ _y are standard deviations of X _i (= logx _i ) and Y _i (= logy _i ), respectively, in the present embodiment, the number of processes x The index k indicating the relationship of the processing time y to is referred to as a correlation coefficient.

  The stored correlation coefficient 1401 is referred to in the processing time calculation processing 603 based on the number of processing cases and the performance unit price.

  FIG. 15 is a diagram illustrating a data configuration example of the coefficient of variation (daily) table 625. The variation coefficient (daily) table 625 stores a batch application name 1500, a variation coefficient 1501, and an average processing time (second) 1502 in association with each other. The names of all batch applications are stored in the batch application name 1500, and the number of processing times and the processing time extraction 615 of the batch execution result reflection processing 614 provided in the batch performance prediction and countermeasure support server 14 are used as the variation coefficient 1501. The coefficient of variation calculated based on the processing time 1203 of each batch stored in the operation result table 622 is stored, and the average processing time (seconds) 1502 includes the past that the batch performance prediction and countermeasure support server 14 has. The average value of the processing time of the processing time 1203 of each batch stored in the operation result table 622 using the processing time calculation processing 604 based on the tendency of the results is stored.

Here, the variation coefficient is the ratio of the standard deviation σ _x of the variable x to the average value <x> of a certain variable x, that is, σ _x / <x>, and is the relative ratio of the error to the average value.

  The stored coefficient of variation 1501 and average processing time (seconds) 1502 are referred to in the processing time calculation processing 604 based on the past performance trend provided in the batch performance prediction and countermeasure support server 14.

  FIG. 16 is a diagram illustrating a data configuration example of the coefficient of variation (weekly) table 626. The variation coefficient (weekly) table 626 stores the batch application name 1600, the variation coefficient 1601, and the average processing time (seconds) 1602 in association with each other. The batch application name 1600 stores the names of all the batch applications, and the coefficient of variation 1601 uses the number of processing times and the processing time extraction 615 of the batch execution result reflection processing 614 provided in the batch performance prediction and countermeasure support server 14. The variation coefficient of the processing time only for the same day of the week calculated based on the date 1200 and processing time 1203 of each batch stored in the operation result table 622 is stored, and the average processing time (seconds) 1602 stores the batch performance. The average value of the processing times of the processing times 1203 of each batch stored in the operation result table 622 is stored using the processing time calculation processing 604 based on the past result tendency provided in the prediction and countermeasure support server 14. .

  The stored variation coefficient 1601 and average processing time (seconds) 1602 are referred to in the processing time calculation processing 604 based on the past performance trend provided in the batch performance prediction and countermeasure support server 14.

  FIG. 17 is a diagram illustrating a data configuration example of the coefficient of variation (monthly) table 627. The variation coefficient (monthly) table 627 stores the batch application name 1700, the variation coefficient 1701, and the average processing time (seconds) 1702 in association with each other. The batch application name 1700 stores the names of all batch applications, and the coefficient of variation 1701 uses the number of processing times and the processing time extraction 615 of the batch execution result reflection processing 614 provided in the batch performance prediction and countermeasure support server 14. The variation coefficient of the processing time of only the same day of another month calculated based on the date 1200 and processing time 1203 of each batch stored in the operation result table 622 is stored, and an average processing time (second) 1702 is stored. Is an average value of the processing times of the processing times 1203 of each batch stored in the operation result table 622 using the processing time calculation processing 604 based on the past performance tendency provided in the batch performance prediction and countermeasure support server 14. Has been.

  The stored coefficient of variation 1701 and average processing time (seconds) 1702 are referred to in the processing time calculation processing 604 based on the past performance trend of the batch performance prediction and countermeasure support server 14.

  FIG. 18 is a diagram illustrating a data configuration example of the coefficient of variation (yearly) table 628. The variation coefficient (yearly) table 628 stores the batch application name 1800, the variation coefficient 1801, and the average processing time (seconds) 1802 in association with each other. The batch application name 1800 stores the names of all batch applications, and the coefficient of variation 1801 uses the number of processing times and the processing time extraction 615 of the batch execution result reflection processing 614 provided in the batch performance prediction and countermeasure support server 14. The variation coefficient of the processing time of only the same day of another year calculated based on the date 1200 and processing time 1203 of each batch stored in the operation result table 622 is stored, and the average processing time (seconds) 1802 is stored. Is an average value of the processing times of the processing times 1203 of each batch stored in the operation result table 622 using the processing time calculation processing 604 based on the past performance tendency provided in the batch performance prediction and countermeasure support server 14. Has been.

  The stored coefficient of variation 1801 and average processing time (seconds) 1802 are referred to in the processing time calculation processing 604 based on the past performance trend of the batch performance prediction and countermeasure support server 14.

  FIG. 19 is a diagram illustrating a data configuration example of the processing time prediction result table 629. The processing time prediction result table 629 stores the batch application name 1900 and the prediction time (seconds) 1901 in association with each other. The names of all batch applications are stored in the batch application name 1900, and the estimated time (seconds) 1901 is calculated using the processing time calculation processing 603 based on the number of processing cases and the performance unit price or the processing time calculation processing 604 based on the past performance trend. The calculated value is stored and updated to the value calculated using the processing number correction 606 of the user correction processing 605 of the predicted processing time, the correction 607 based on the processing characteristics, and the countermeasure content selection 612 of the performance countermeasure support processing 610. The stored and updated predicted time (seconds) 1901 is referred to in the processing number correction 606 of the user correction processing 605 of the predicted processing time, the correction 607 based on processing characteristics, and the scheduled batch end time calculation processing 608.

  Next, the overall processing flow of performance prediction and performance countermeasure support will be described in detail with reference to the drawings.

  FIG. 20 is a flowchart showing the entire performance prediction and performance countermeasure support processing. The processing shown in the flowchart is executed once a day, and the timing of the regular activation S2000 is after completion of execution of all batches.

  The execution performance reflection S2001 uses the batch execution performance reflection processing 614 provided in the batch performance prediction and countermeasure support server 14. FIG. 21 shows the detailed processing flow of the execution result reflection S2001.

  The processing number extraction S2101 is processing for all batches executed based on the application log 405 provided in the batch server 12 using the processing number and processing time extraction 615 provided in the batch performance prediction and countermeasure support server 14. The number of cases is extracted and inserted into the number of cases 1202 in the operation result table 622 included in the batch server 12. At that time, the date 1200 and the batch application name 1201 are also inserted.

  In the processing time calculation S2102, the processing for all batches executed based on the application log 405 provided in the batch server 12 using the processing number provided in the batch performance prediction and countermeasure support server 14 and the processing time extraction 615. The start time and the processing end time are extracted, the difference is calculated as the processing time, and the processing time (the processing time of the hit line using the year / month / day 1200 and batch application name 1201 of the operation result table 622 included in the batch server 12 as a key) Second) 1203.

  In the performance unit price calculation S2103, the performance unit price and correlation coefficient recalculation 616 provided in the batch performance prediction and countermeasure support server 14 are used to store the operation performance table 622 provided in the batch performance prediction and countermeasure support server 14. Using the batch application name 1201 as a key, the number of processes 1202 and the processing time 1203 are all extracted to calculate the performance unit price, and the average value is the same batch in the performance unit price table 621 included in the batch performance prediction and countermeasure support server 14. The performance unit price 1101 of the application name 1100 is updated. This process is executed for all batch applications.

  In the correlation coefficient recalculation S 2104, the performance unit table provided in the batch performance prediction and countermeasure support server 14 using the performance unit price and correlation coefficient recalculation 616 included in the batch performance prediction and countermeasure support server 14. The same batch application in the correlation coefficient table 624 provided in the batch performance prediction and countermeasure support server 14 calculates the correlation coefficient by extracting all the processing number 1202 and the processing time 1203 using the batch application name 1201 of 622 as a key. The correlation coefficient 1401 of the name 1400 is updated. This process is executed for all batch applications.

  In the process count S2002, the SQL for extracting the process count of each batch using the process count count process 602 provided in the batch performance prediction and countermeasure support server 14 is stored in the DB tables 301 and 302 of the HUB-DB server 11. The processing number 1001 of the same batch application name 1000 in the processing number table 620 provided in the batch performance prediction and countermeasure support server 14 is updated with the processing number extracted and extracted. This process is executed for all batch applications.

  In step S2003, the batch processing time prediction and countermeasure support server 14 uses the processing time calculation process 603 based on the number of processes and the performance unit price provided in the batch performance prediction and countermeasure support server 14. The processing number 1001 is extracted for all batch application names 1000 in the processing number table 620 provided, and the same batch in the performance unit price table 621 provided in the extracted processing number 1001 and the batch performance prediction and countermeasure support server 14 is extracted. The result of multiplying the performance unit price 1101 of the application name 1100 is stored in the predicted processing time 1301 of the same batch application name 1300 of the processing time prediction work table 623 provided in the batch performance prediction and countermeasure support server 14.

  In step S2004 for determining whether the correlation coefficient is 0.7 or more, the batch performance prediction and countermeasure support server uses the processing time calculation process 603 based on the number of processes and the performance unit price that the batch performance prediction and countermeasure support server 14 has. 14 extracts the correlation coefficient 1401 of the correlation coefficient table 624 provided in 14. This correlation coefficient 1401 is a correlation coefficient between the number of processing cases and the processing time. The search process is executed for the correlation coefficient 1401 of all batch application names 1400. It is determined whether or not the correlation coefficient 1401 extracted in the search process is equal to or greater than a threshold value (0.7 in this example). If the correlation coefficient 1401 is equal to or greater than 0.7, the same batch application name 1900 in the processing time prediction result table 629 is determined. Stored in the predicted row time 1901. If the correlation coefficient 1401 is less than 0.7, the present process is terminated, and a predicted processing time calculation S2005 based on the past performance trend is executed. Note that this embodiment can be realized regardless of the threshold value.

In the present embodiment, as already described, the correlation coefficient in the present embodiment is x, where x is the number of processing cases, y is the processing time (total processing time) required to execute x processing, and x This is an index k when the relationship with y is expressed as y = f (x) = Cx ^k. When determining this index k, this index k includes a correlation coefficient in ordinary mathematical statistical processing. Therefore, the index k is referred to as a correlation coefficient. In mathematical statistical processing, as a criterion for determining whether or not there is a correlation between two variables, the absolute value of the correlation coefficient that is the direction cosine (cos θ) of the angle θ formed by the n-dimensional vector of the data string of the two variables is In many cases, a value of 1 / √ (2) (= 0.707) corresponding to θ = π / 4 is used. The threshold value for determining the correlation coefficient value corresponds to this value.

  The calculation processing time 604 based on the past performance trend provided in the batch performance prediction and countermeasure support server 14 is used in the calculation processing S2005 based on the past performance trend. FIG. 22 shows the detailed processing flow of the calculation processing time S2005 based on the past performance trend.

  Prediction processing time calculation (daily) S2201 is provided in the batch performance prediction and countermeasure support server 14 using the processing time calculation process 604 based on the past performance trend that the batch performance prediction and countermeasure support server 14 has. The processing time 1203 of the operation performance table 622 is extracted for each batch application name 1201, and the coefficient of variation is calculated to be the same as the coefficient of variation (daily) table 625 included in the batch performance prediction and countermeasure support server 14. Stored in the variation coefficient 1501 of the batch application name 1501. Further, the average value of the processing time is stored in the average processing time (seconds) 1502. The above processing is executed for all batch application names 1500.

  Prediction processing time calculation (weekly) S2202 is provided in the batch performance prediction and countermeasure support server 14 using the processing time calculation process 604 based on the past performance trend that the batch performance prediction and countermeasure support server 14 has. The batch performance prediction and countermeasure support server 14 includes all the processing times 1203 in which the date 1200 of the operation result table 622 is the same day of the week as the processing execution date for each batch application name 1201 and calculates the coefficient of variation. Are stored in the variation coefficient 1601 of the same batch application name 1601 in the variation coefficient (weekly) table 626 being used. Further, the average value of the processing times is stored in the average processing time (seconds) 1602. The above process is executed for all batch application names 1600.

  Prediction processing time calculation (monthly) S2203 is provided in the batch performance prediction and countermeasure support server 14 using the processing time calculation process 604 based on the past performance trend that the batch performance prediction and countermeasure support server 14 has. The processing time 1203 in which the year / month / day 1200 of the operation result table 622 is the same day as the processing execution date and another month is extracted for each batch application name 1201, and the coefficient of variation is calculated to predict batch performance and support measures The data is stored in the variation coefficient 1701 of the same batch application name 1701 in the variation coefficient (monthly) table 627 provided in the server 14. Further, the average value of the processing times is stored in the average processing time (seconds) 1702. The above process is executed for all batch application names 1700.

  Prediction processing time calculation (annual) S2204 is provided in the batch performance prediction and countermeasure support server 14 using the processing time calculation process 604 based on the past performance trend that the batch performance prediction and countermeasure support server 14 has. The processing date 1203 in the operation result table 622 that is the same date as the processing execution date and another date is extracted for each batch application name 1201, and the coefficient of variation is calculated to estimate the batch performance. The data is stored in the coefficient of variation 1801 of the same batch application name 1801 in the coefficient of variation (annual) table 628 included in the support server 14. Further, the average value of the processing time is stored in the average processing time (second) 1802. The above processing is executed for all batch application names 1800.

  Adopting the predicted processing time with the smallest variation coefficient S2205 is a type in which the variation coefficient is closest to 0 among the four daily, weekly, monthly, and yearly variation coefficients calculated in S2201 to S2204. Is used as the predicted processing time of the batch, and stored in the predicted time 1901 of the same batch application name 1900 of the processing time prediction result table 629 provided in the batch performance prediction and countermeasure support server 14.

  The user correction result application S2006 for the prediction processing time uses the user correction processing 605 for the prediction processing time that the batch performance prediction and countermeasure support server 14 has. A detailed processing flow of the user correction result application S2006 for the prediction processing time is shown in FIG.

  In the processing number correction S2300, the processing number correction screen 2400 of the processing number correction 606 provided in the batch performance prediction and countermeasure support server 14 is used. An example of the processing number correction screen is shown in FIG. All batch application names 1900 registered in the batch application name 1900 of the processing time prediction result table 629 provided in the batch performance prediction and countermeasure support server 14 are displayed in the batch application name 2401 of the processing number correction screen 2400, and the batch The average value obtained by extracting all past processing counts 1202 from the operation result table 622 using the application name 1201 as a key is displayed in the average processing count 2402, and the processing count 1001 extracted from the processing count table 620 using the batch application name 1000 as a key. The processing number 2403 is displayed, and the predicted processing time 2404 displays the predicted processing time 1901 extracted from the processing time prediction result table 629 using the batch application name 1900 as a key. The number of processed cases 2403 can be corrected by the user. By pressing the recalculation 2405 button after correcting the number of processed cases 2403, the batch application name 1100 is used as a key from the performance unit price table 621. A value obtained by multiplying the extracted performance unit price 1101 is displayed again in the prediction processing time 2404. Further, after the decision 2406 button is pressed, the processing time prediction result table 629 is searched with the batch application name 1900 and stored in the prediction processing time 1901 of the corresponding column.

In the correction S2301 based on the processing characteristics, the correction screen 2500 based on the processing characteristics of the correction 607 based on the processing characteristics provided in the batch performance prediction and countermeasure support server 14 is used. A screen example of the correction screen 2500 based on the processing characteristics is shown in FIG. All batch application names 1900 registered in the batch application name 1900 of the processing time prediction result table 629 provided in the batch performance prediction and countermeasure support server 14 are displayed in the batch name 2501 of the correction screen 2500 according to the processing characteristics, and the processing is performed. In the case number dependency (rate) 2502, when the correlation coefficient 1401 extracted from the correlation coefficient table 624 using the batch application name 1400 as a key is equal to or greater than the threshold (0.7), 100 is displayed, and the threshold (0. 7) If less than 0, 0 is displayed, and the processing time dependency (time) 2503 displays the predicted time 1301 extracted from the processing time prediction work table 623 using the batch application name 1300 as a key, and the processing time actual dependency (rate ) 2504 includes a correlation coefficient table with the batch application name 1400 as a key. 100 is displayed when the correlation coefficient 1401 extracted from the data 624 is less than the threshold (0.7), 0 is displayed when the correlation coefficient 1401 is greater than or equal to the threshold (0.7), and the processing time performance dependence (time) 2505 is displayed. Shows the average processing time 1501 extracted from the variation coefficient (daily) table 625 using the batch application name 1500 as a key. The processing number dependency (rate) 2502 and the processing time performance dependency (rate) 2504 total 100, and this value can be corrected by the user. That is, the initial values of “100” and “0” or “0” and “100” can be corrected. By pressing the recalculate 2508 button after correction, the computer recalculates the predicted processing time 2506 using the following formula.
(Equation 1)
Predicted processing time = number of processing cases (rate) / 100 x number of processing cases (time) + actual processing time dependency (rate) / 100 x actual processing time dependency (time)
However, the processing number dependency (rate) and the processing time performance dependency (rate) are values at which the sum of these values becomes 1.

  By pressing the decision 2509 button, the computer searches the processing time prediction result table 629 provided in the batch performance prediction and countermeasure support server 14 for the predicted processing time 2506 by the batch application name 1900, and the predicted time 1901 of the corresponding column. To store.

  The scheduled batch end time calculation S2007 displays the batch end time display screen 2600 of the scheduled batch end time calculation process 608 provided in the batch performance prediction and countermeasure support server 14. A screen example of the batch end time display screen 2600 is shown in FIG. All batch application names 1900 registered in the batch application name 1900 of the processing time prediction result table 629 provided in the batch performance prediction and countermeasure support server 14 are displayed in the batch application name 2601, and the batch application is displayed in the processing time 2602. The predicted time 1901 extracted from the processing time prediction result table 629 is displayed using the name 1900 as a key, and the processing time of each batch stored in the processing time prediction result table 629 at the scheduled batch processing end time 2603 and the batch execution management server 13 are displayed. Based on the batch execution definition 503 stored in, the time when all the batches are finished is calculated and displayed in consideration of the context and execution time of the batch.

  In S2008 for determining whether or not the scheduled batch end time is after 6:00, the scheduled batch processing end time 2603 displayed when the decision 2605 button on the batch end time display screen 2600 is pressed is a threshold value (in this example, 6:00). If it exceeds the threshold value (6:00), this process is completed and the execution of performance countermeasure support processing S2009 is executed, and the threshold value (6:00) is not exceeded. In this case, this processing is completed and the processing proceeds to S2010 of end. Note that the threshold value may be set to the time at which the batch should be completed in consideration of the system characteristics, and this embodiment can be implemented regardless of the threshold value.

  The performance countermeasure support processing 610 of the batch performance prediction and countermeasure support server 14 is used in the performance countermeasure support processing execution S2009. FIG. 27 shows a detailed flow in the performance measure support process execution S2009.

  In step S2700, a candidate batch for performing performance countermeasures is selected using the candidate batch selection 611 included in the batch performance prediction and countermeasure support server 14. When countermeasures are executed on all batches, user operations are required and inefficiency is required. Therefore, Pareto's law is that batches with 20% of the total number of processing accounts for 80% of the total processing time. In accordance with this idea, the batch with the top 20% of the processing time is selected from the batch application name 1900 and the prediction time 1901 of the processing time prediction result table 629 provided in the batch performance prediction and countermeasure support server 14, and the corresponding batch application The name 1900 is displayed in a state in which the countermeasure candidate 2801 on the performance countermeasure screen 2800 shown in FIG. 28 is checked. In the target batch 2802, all batch application names 1900 in the processing time prediction result table 629 provided in the batch performance prediction and countermeasure support server 14 are displayed.

  The measure content selection S 2701 is a measure candidate A 901, measure candidate B 902 for each batch application name 900 registered in advance in the performance measure master 619 using the measure content selection 612 provided in the batch performance prediction and measure support server 14. The countermeasure candidate C903 is displayed in a pull-down format on the countermeasure contents 2803 of the performance countermeasure support screen 2800. The value initially displayed in the pull-down is the value of the countermeasure candidate A901. When the user can select the performance countermeasure content 2803 from the pull-down menu and press the recalculate 2806 button after the change, the selected performance countermeasure content 2803, the target batch 2802, and the batch performance are checked for the countermeasure object 2801. All the processing time 1203 of the row in which the performance countermeasure implementation content 1204 of the operation result table 622 provided in the prediction and countermeasure support server 14 matches the batch application name 1201 are extracted, and the average value is calculated, and the processing time 2804 is obtained. Redisplayed and stored in the predicted time 1901 of the same batch application name 1900 in the processing time predicted result table 629. In addition, the batch end time calculation processing 608 provided in the batch performance prediction and countermeasure support server 14 is called to calculate the batch end time and display it at the scheduled batch processing end time 2805.

  The countermeasure instruction S2702 is displayed as a countermeasure target 2801 on the performance countermeasure support screen 2800 when a decision 2807 button is pressed on the performance countermeasure support screen 2800 included in the countermeasure instruction 613 included in the batch performance prediction and countermeasure support server 14. The countermeasure content 2803 is executed for the target batch 2802 that is checked. If the countermeasure content 2803 is execution SKIP, the batch execution definition 503 of the batch execution management server 13 is changed to skip execution of the corresponding batch application on the current day. If the measure content 2803 is split execution, the batch execution definition 503 of the batch execution management server 13 is changed to skip execution of the current batch, and the batch execution definition for split execution registered in advance is executed. sign up. If the measure content 2803 is scale-up, an email instructing scale-up of the batch server 12 is sent to the operation manager who uses the operation terminal 15.

11: HUB-DB server, 12: Batch server, 13: Batch execution management server, 14: Batch performance prediction and countermeasure support server, 15: Operation terminal, 16: Network, 300, 400, 500, 600, 700: Storage unit , 301, 302, 618: DB table, 401, 501, 601: Program, 405: Application log, 503: Batch execution definition, 704: Display / input unit

Claims (6)

In a system that executes a plurality of batch application processes with order in a predetermined batch executable time zone,
A means to predict the execution time of each batch application from the number of processing, performance unit price and past performance time using statistical methods,
A means for the computer to recalculate the estimated time using as input the number of processes corrected by the user from the operation terminal and the characteristics of the processing time of each batch application,
A means for calculating the end time of all batch application processes on the current day based on the execution order and execution time of batch applications,
A batch performance prediction and countermeasure support system comprising means for allowing a user to select a performance countermeasure and instructing the execution of the selected performance countermeasure when the end time exceeds a batch executable time zone. In a method of executing a plurality of batch application processes with order in a predetermined batch executable time zone using a computer,
Get the number of processes as of the previous day of each batch application,
Multiply the performance unit price of each batch calculated from the past batch performance and the number of processing cases to calculate the processing time for all batches,
Calculate the correlation coefficient between the number of processing cases and processing time from the past batch operation results,
If the correlation coefficient is greater than or equal to a threshold value, the calculated processing time is adopted as a predicted value, and if not, the variation coefficient is the largest among the average values of processing time on various specific dates in the past from the batch operation results. A batch performance prediction and countermeasure support method characterized in that an average value of processing times when the value of 0 is 0 is adopted as a predicted value. Based on the estimated value of the processing time adopted, recalculate the processing time in consideration of the correction information of the processing number input by the user using the operation terminal, the correction number of the processing number dependency rate and the processing time performance dependency rate. ,
Use the predefined batch execution definition to calculate the end time of all batches for the day,
3. The batch performance prediction and countermeasure support method according to claim 2, wherein when the calculated predicted value of the batch end time does not fall within the batch executable time zone, execution of performance countermeasures is instructed. When executing performance measures, the performance measures for all batches set in advance for each batch, and the predicted processing time and end time of all batches when performance measures are applied to the batches targeted for performance measures are calculated from the batch operation results. And
4. The batch performance prediction and countermeasure support method according to claim 3, wherein the batch performance prediction and countermeasure support method are displayed on a screen of an operation terminal.   5. The performance measure target batch and a performance measure content of the performance measure target batch are selected and confirmed based on the display content, and a performance measure set based on the information is instructed. Batch performance prediction and countermeasure support method.   The processing time of each batch, the number of processing cases, the results of performance countermeasure implementation contents are saved as the batch operation results of the day after the batch execution, and the correlation coefficient and the performance unit price are recalculated. 2. Batch performance prediction and countermeasure support method according to 2.

Images