JP7041607B2 - Resource management support device and resource management support method - Google Patents

Description

The present invention relates to a resource management support device and a resource management support method.

In recent years, the complexity and scale of systems have been accelerating, such as device configurations, multi-layered applications, and cooperation with various external systems.

On the other hand, in such a system, resource management work such as monitoring and analyzing the operating status of various resources, early detection of problems in the system, and planning of reinforcement plans with an eye on future operation is becoming more important.

As a conventional technique related to such system operation analysis and the like, for example, in an access number prediction system that predicts the number of accesses from customers, the access history is recorded by dividing the access according to the access requirements. , When the number of accesses in the prediction target period is predicted using the storage means for storing the access history for each predetermined period and the access history, when there is an increase or decrease in the number of accesses for the access requirements, the increase or decrease is used as the access history. A prediction means for predicting the number of accesses in the prediction target period based on the access history, and an access number prediction system (see Patent Document 1) are proposed.

In addition, it is an access number estimation device that estimates the number of user accesses to services in the communication network, and is a DNS query totaling unit that aggregates DNS (Domain Name System) queries for each domain at regular time intervals and records them in the communication log. And, based on the model formula expressed using the cumulative number of queries for each domain calculated from the communication log, the maximum TTL (Time-To-Live), and the number of potential users, the number of accesses is estimated by estimating the number of accesses for each domain. An access number estimation device (see Patent Document 2) having a unit and an output unit that outputs the number of accesses for each domain has also been proposed.

Japanese Unexamined Patent Publication No. 2009-15415 Japanese Unexamined Patent Publication No. 2018-32965

In general, the system engineer in charge is in charge of each task of system operation analysis and resource management based on the conventional system operation analysis.

In this case, the system engineer, for example, periodically acquires and analyzes the operation information of the relevant system, and based on the result, makes a decision on resource management such as changing various settings.
However, as described above, the larger and more complex the target system is, the more man-hours and costs required for the relevant work are significantly increased.

In addition, it is a work with strong personality, and the above-mentioned resource management judgment accuracy differs depending on the system engineer. In addition, if the criteria for making such a judgment are ambiguous in the first place and the skills and experience of the system engineer are insufficient, the judgment result tends to lead to a corresponding operational risk.

If the system is operated based on such accuracy, it may lead to cases where the performance required for business cannot be achieved at high load, or where the operation cost remains high by setting resources with excessively high functionality for the load. do not have.
Therefore, an object of the present invention is to provide a technique for making system resource management efficient and accurate.

The resource management support device of the present invention that solves the above problems has a storage unit that stores past information of the number of users, data processing amount, resource status, and external environment in the target system, and a predetermined period in the past. By machine learning using past information of the number of users, information on financial products, which is information on the external environment, and information on sales promotion activities related to the system, in a predetermined time in the future from the predetermined period in the past. First process to estimate the expected number of users of financial products and sales promotion activities under a predetermined situation , and input the data processing amount in the past predetermined period, the past information of the number of users, and the calendar characteristics of the period. Based on the second process of estimating the predicted data processing amount for the predicted number of users in the predetermined time in the future from the predetermined period in the past, and the past information of the resource status in the predetermined period in the past. Third, it is determined whether the resource usage status and processing time are equal to or longer than the predetermined standard, and the required allocation amount of the resource indicated by the past information is increased or decreased by a predetermined rate from the current state according to the result of the determination. The predicted data processing at a predetermined time in the future from the predetermined period in the past by machine learning using the processing and the required allocation amount of the resource estimated for the predetermined period in the past and the peak value of the data processing amount in the period as inputs. It is characterized by including a fourth process for estimating a required allocation amount for prediction in terms of quantity, and a calculation unit for executing.

Further, in the resource management support method of the present invention, an information processing device provided with a storage unit that stores past information of each of the number of users, the amount of data processing, the resource status, and the external environment in the system to be managed has been used in the past. By machine learning using past information of the number of users , information on financial products, which is information on the external environment, and information on sales promotion activities related to the system in the predetermined period of the above, a predetermined time in the future from the predetermined period in the past. First processing for estimating the estimated number of users of a predetermined financial product and sales promotion activity under a predetermined situation , and the past information of the data processing amount and the number of users in the past predetermined period and the calendar characteristics of the period. The second process of estimating the predicted data processing amount for the predicted number of users in the predetermined period in the future from the predetermined period in the past by machine learning using the above, and the past information of the resource status in the predetermined period in the past. Based on the above, it is determined whether the resource usage status and processing time are equal to or longer than the predetermined standard, and it is estimated that the required allocation amount of the resource indicated by the past information is increased or decreased by a predetermined rate from the current state according to the result of the determination. The third process is performed, and machine learning is performed by inputting the required allocation amount of the resource estimated for the predetermined period in the past and the peak value of the data processing amount in the period. It is characterized by executing the fourth process of estimating the predicted required allocation amount in the predicted data processing amount.

According to the present invention, the resource management of the system can be made efficient and accurate.

It is a figure which shows the example of the network configuration including the resource management support apparatus of this embodiment. It is a figure which shows the hardware configuration example of the resource management support apparatus of this embodiment. It is a figure which shows the data structure example of the active contractor number information of this embodiment. It is a figure which shows the data structure example of the financial information of this embodiment. It is a figure which shows the data structure example of the sales promotion information of this embodiment. It is a figure which shows the data structure example of the resource-related information of this embodiment. It is a figure which shows the flow example of the resource management support method in this embodiment. It is a figure which shows the structural example 1 of the objective variable and the explanatory variable in this embodiment. It is a figure which shows the structural example 2 of the objective variable and the explanatory variable in this embodiment. It is a figure which shows the structural example 3 of the objective variable and the explanatory variable in this embodiment. It is a figure which shows the structural example 4 of the objective variable and the explanatory variable in this embodiment. It is a figure which shows the flow example 2 of the resource management support method in this embodiment.

－－－ Network configuration －－－

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. FIG. 1 is a diagram showing a configuration example of the resource management support device 100 in the present embodiment. The resource management support device 100 shown in FIG. 1 is a computer device that makes system resource management efficient and accurate.

The resource management support device 100 is, for example, a server device operated by a vendor. This vendor is, for example, responsible for managing the business systems of financial institutions. The business system is the target system 300 in FIG.
As an example of the target system 300, a business system such as an Internet banking service provided by a financial institution can be assumed.
On the other hand, the user who uses the target system 300, that is, the terminal operated by the customer of the above-mentioned financial institution is the user terminal 200.

As a specific example of the user terminal 200, a smartphone, a PC, a tablet terminal, or the like can be assumed. In this case, the customer of the financial institution operates the user terminal 200 and uses the Internet banking service provided by the target system 300.

On the other hand, the resource management support device 100 operated by the vendor efficiently and accurately manages the resources in the target system 300 based on various information such as the number of active contractors and the daily transaction volume in the target system 300. It becomes a device.
--- Hardware configuration ---

Further, the hardware configuration of the resource management support device 100 in this embodiment is as follows. As illustrated in FIG. 2, the resource management support device 100 is an SSD (Solid).
A storage unit 101 composed of an appropriate non-volatile storage element such as a State Drive) or a hard disk drive, a memory 103 composed of a volatile storage element such as RAM, and a program 102 held in the storage unit 101 are read into the memory 103. An arithmetic unit 104 such as a CPU that performs integrated control of the device itself and performs various judgments, operations, and control processes, an input unit 105 that accepts key input and voice input from the user, a display that displays processed data, and the like. It includes an output unit 106 and a communication unit 107 that is connected to the network 10 and is responsible for data communication with other devices such as the target system 300 described above.
When the resource management support device 100 is a stand-alone machine, the above-mentioned communication unit 107 is unnecessary.

Further, in the storage device 101, in addition to the program 102 for implementing the function required as the resource management support device of the present embodiment, the number of active contractors information 125, financial information 126, sales promotion information 127, and resource-related information. Information 128 is at least stored. However, the details of this information will be described later.

Further, it is assumed that the program 102 includes the machine learning algorithm 1021. As the machine learning algorithm 1021, for example, the objective variable which is the item to be estimated (example: the number of active contractors, the daily transaction volume, the allocated resource value, etc. described later) is set as the explanatory variable (example: financial information and the calendar described later). Regression problems explained by characteristics, predicted number of active subscribers, predicted daily transaction volume, etc.) are solved by so-called supervised learning, and linear regression model, MARS, support vector regression, regression tree, model tree, gene programming, etc. are appropriately performed. It should be adopted for.
--- Data structure example ---

Subsequently, the tables used by the resource management support device 100 of the present embodiment will be described. FIG. 3 shows an example of the number of active contractors information 125 in this embodiment.
The number of active contractors information 125 of the present embodiment is a table accumulating information on the number of daily active contractors and the transaction volume in the target system 300.
The data structure is a collection of records consisting of data such as daily transaction volume, number of active subscribers, year, month, day, and day of the week, with the day as the key.

Further, FIG. 4 shows an example of the financial information 126 in the present embodiment. The financial information 126 of the present embodiment is a table accumulating information such as prices and interest rates of financial products and the like on a monthly basis.
The data structure is a collection of records consisting of data such as the previous month's average stock price and the previous month's interest rate for the month, with the month as the key.

Further, FIG. 5 shows an example of sales promotion information 127 according to the present embodiment. The sales promotion information 127 of the present embodiment is a table in which information on sales promotion performed by a financial institution or the like with respect to the service provided by the target system 300 is accumulated monthly.

The data structure is a collection of records consisting of data such as the number of CMs in the previous month and the number of DMs in the previous month, with the month as the key. CM is a commercial film broadcast on the mass media, and DM is a direct mail sent to each customer.

Further, FIG. 6 shows an example of the resource-related information 129 in this embodiment. The resource-related information 129 of the present embodiment is a table in which the types and amounts of various resources set in the target system 300, the usage status of the resources in the period, and the like are accumulated monthly.

The data structure consists of the type and amount of resources allocated to the target system 300 in the month, the maximum usage rate / usage amount at the peak time, and the average usage rate / usage amount at the peak time in the target system 300, with the month as the key. It is a collection of records consisting of data such as batch processing time and waiting time.

Of the above, the batch processing time is the time required for batch processing in the corresponding month in the target system 300 (eg, daily processing related to business data in the Internet banking service), and is the average time in the target system 300. The ratio when the batch processing time or the predetermined normal processing time is 1.0 is shown.

Further, the waiting time is a waiting time (latency) generated in response to a request from the user when the service in the target system 300 is used by the user, and is an average waiting time in the target system 300 or a predetermined time. The ratio when the normal waiting time is 1.0 is shown.
－－－ Flow example －－－

Hereinafter, the actual procedure of the resource management support method in the present embodiment will be described with reference to the figure. Various operations corresponding to the resource management support method described below are realized by a program read by the resource management support device 100 into a memory or the like and executed. The program is composed of codes for performing various operations described below.

FIG. 7 is a diagram showing a flow example 1 of the resource management support method in the present embodiment. First, the resource management support device 100 receives, for example, the arrival of a predetermined time by its own calendar function (a function naturally provided as a computer), or receives a user instruction from the input unit 105, and the storage unit 101 Information (past information) in the past predetermined period is extracted from the number of active contractors information 125, financial information 126, and sales promotion information 127 (s10).

Here, as illustrated in Table 150 of “At the time of learning” in FIG. 8, the number of active subscribers (the number of active subscribers) for each month from “January 2018” to “April 2018”, which is a predetermined period in the past. Number of users: Number of active subscribers (extracted from information 125), previous month's average stock price and interest rate (external environment information: both extracted from financial information 126), previous month's number of CMs and previous month's DM (external environment information: all) (Extracted from sales promotion information 127)) is obtained.

Subsequently, the resource management support device 100 uses the number of active subscribers as the objective variable, the previous month's average stock price and the previous month's interest rate, and the previous month's CM number and the previous month's DM number as explanatory variables in the information obtained in s10, and is a machine learning algorithm. Enter in 1021, for example, estimate the active subscribers in "May 2018", which is the next month from the present time (s11).

The explanatory variables at this time are the average stock price and interest rate of "April 2018", which is the latest of "May 2018", as shown in Table 151 of "Forecast time" in FIG. The previous month's stock price and the number of CMs and DMs for "April 2018", the number of CMs for the previous month and the number of DMs for the previous month.

Subsequently, the resource management support device 100 extracts the daily transaction volume (data processing volume), the number of active contractors, and the calendar characteristics of the period from the active contractor number information 125 (s12). ..

Here, as illustrated in Table 160 of “At the time of learning” in FIG. 9, each day from “May 1, 2008” to “April 30, 2018”, which is a predetermined period in the past, is a day. Next transaction volume (data processing volume: extracted from active contractor number information 125), active contractor number (user number: extracted from active contractor number information 125), date, day (calendar characteristics: all active contracts) Extracted from the number of persons information 125), is obtained.

Subsequently, the resource management support device 100 inputs the information obtained in s12 into the machine learning algorithm 1021 with the date and time and the daily transaction volume as objective variables and the number of active contractors, the date, and the day as explanatory variables. Then, for example, the daily transaction volume for each day from "May 1, 2018" to "May 31, 2018", which is tomorrow from the present time, is estimated (s13).

The explanatory variables at this time are the number of active subscribers estimated in s11 and the numbers from "May 1, 2018" to "May 31, 2018", as shown in Table 161 of "Prediction time" in FIG. Each day of the week.

Subsequently, the resource management support device 100 uses the resource-related information 128 of the storage unit 101 to provide past information on the resource status for each month from "May 2008" to "April 2018", which is a predetermined period in the past. As the peak usage rate or average usage rate, and batch processing time or latency (both processing times) information is extracted (s14).

Here, as illustrated by the table 170 in FIG. 10, the resource type and its allocated amount for each month from "May 2008" to "April 2018", which is a predetermined period in the past.
It is assumed that the maximum usage rate / usage amount at the peak time, the average usage rate / usage amount, the batch processing time, and the latency (past information of the resource status: all extracted from the resource-related information 128) are obtained.

Subsequently, in the resource management support device 100, among the information obtained in s14, the maximum usage rate / usage amount or the average usage rate / usage amount at the peak time and the batch processing time or latency are both equal to or higher than the first standard. Is determined (s15).

Here, the first criterion is, for example, "100%" for the maximum usage rate / usage amount at the peak time, "100%" for the average usage rate / usage amount at the peak time, and the batch processing time and latency. Is "1.3".

As a result of the above determination, when neither the maximum usage rate / usage amount or the average usage rate / usage amount at the peak and the batch processing time or latency are equal to or higher than the first standard (s16: No), the resource management support device 100 Transitions the process to s18.

On the other hand, as a result of the above determination, when both the maximum usage rate / usage amount or the average usage rate / usage amount at the peak and the batch processing time or latency are equal to or higher than the first standard (s16: Yes), resource management support The apparatus 100 estimates that the required allocation amount of the resource indicated by the above-mentioned past information is increased by a predetermined ratio from the value of the corresponding period (which may include the current concept) (s17), and the process transitions to s22. ..

As the "predetermined ratio" to be increased, for example, a ratio at which the batch processing time or latency exceeds the first criterion can be assumed. Therefore, if the batch processing time or latency exceeds the first criterion by 10%, the required allocation of the resource is increased by 10% from the value of the corresponding period.

On the other hand, as a result of the determination in step s15 described above, when both the maximum usage rate / usage amount or the average usage rate / usage amount at the peak and the batch processing time or latency are not equal to or higher than the first criterion (s16: No). , The resource management support device 100 determines whether the maximum usage rate / usage amount or the average usage rate / usage amount at the peak time and the batch processing time or latency are both equal to or less than the second reference (s18).

Here, the second criterion is, for example, "50%" for the maximum usage rate / usage amount at the peak time, "50%" for the average usage rate / usage amount at the peak time, and the batch processing time and latency. Is "0.5".

As a result of the above determination, when neither the maximum usage rate / usage amount or the average usage rate / usage amount at the peak and the batch processing time or latency are equal to or less than the second standard (s19: No), the resource management support device 100 Estimates that the required allocation amount of the resource indicated by the above-mentioned past information is appropriate as a value of the corresponding period (which may include the current concept) (s20), and transitions the process to s22.

On the other hand, as a result of the above determination, when both the maximum usage rate / usage amount or the average usage rate / usage amount at the peak and the batch processing time or latency are equal to or less than the second standard (s19: Yes), resource management support It is estimated that the apparatus 100 reduces the required allocation amount of the resource indicated by the above-mentioned past information by a predetermined ratio from the value of the corresponding period (which may include the current concept) (s21).

The "predetermined percentage" to be reduced includes, for example, batch processing time or latency.
A ratio below the second standard can be assumed. Therefore, if the batch processing time or latency exceeds the second criterion by 10%, the required allocation of the resource is reduced by 10% from the value for that period.

Subsequently, the resource management support device 100 sets the required allocation amount of each resource in the target system 300 estimated for a predetermined period by at least one of the above-mentioned s17, s20, and s21 as the objective variable, and the data processing amount in the period. The peak value of (estimated in s13, the maximum value of the daily transaction volume on each day of the corresponding period) is shown as an explanatory variable in Table 180 at the time of learning and Table 181 at the time of prediction in FIG. As described above, the data is input to the machine learning algorithm 1021 and, for example, the required allocation amount (“allocation resource value” in FIG. 11) in “May 2018” is estimated (s22).

The resource management support device 100 displays, for example, the required allocation amount (table 181 in FIG. 11) estimated for the resources of the target system 300 on the output unit 106 (s23), and ends the process.

As shown in the flow of FIG. 12, instead of steps s10 to s13 in flow example 1 of FIG. 7, the past of the number of active contractors, financial information, sales promotion information, and daily transaction volume in the past predetermined period. The information and the calendar characteristics of the relevant period are acquired (s101), of which machine learning is executed by the machine learning algorithm 1021 with the daily transaction volume as the objective variable, the number of active contractors and the calendar characteristics as the explanatory variables, and the future The daily transaction volume at the number of active contractors on each day of each calendar characteristic at a predetermined time may be estimated (s102). In this case, steps s14 and subsequent steps are the same as in Flow Example 1.

Although the best mode for carrying out the present invention has been specifically described above, the present invention is not limited to this, and various modifications can be made without departing from the gist thereof.

According to this embodiment, it is possible to automate the analysis of periodic operation information regarding the target system and accurately determine whether or not the resource setting needs to be changed based on the result. This reduces the load on the system engineer and the like, leads to the reduction of human costs and the like, and at the same time, it is possible to set and deal with the resource configuration that can be accurately dealt with at the peak time in the future. For the end user of such a system, there is an advantage that smooth and comfortable service use is possible without encountering unfavorable situations such as difficulty in access due to high load and latency exceeding the allowable range.
That is, the resource management of the system can be made efficient and accurate.

The description herein reveals at least the following: That is, in the resource management support device of the present embodiment, when the calculation unit is equal to or more than the first reference in both the usage status and the processing time in the fourth processing, the required allocation of the resource indicated by the past information is performed. It is estimated that the amount is increased by a predetermined ratio from the current state, and if both the usage status and the processing time are equal to or less than the second standard, the required allocation amount of the resource indicated by the past information is decreased by a predetermined ratio from the current state. It may be presumed that.

According to this, for example, when the resource usage remains high (eg 100%), it may be a state of overload in which the resource is completely used up but the performance is adversely affected, or the resource is completely used up. It is possible to make an appropriate decision by using the processing time as a guide for a situation that was not clear in the past, whether it is an overload state that adversely affects the performance above. As a result, system resource management can be made more efficient and accurate.

Further, in the resource management support device of the present embodiment, in the fourth processing, the calculation unit has the maximum usage rate or the average usage rate at the peak time, which is the usage status, and the batch processing time or the batch processing time, which is the processing time. If both the latency is equal to or higher than the first standard, it is estimated that the required allocation amount of the resource indicated by the past information is increased by a predetermined ratio from the current state, and the maximum usage rate or the average usage at the peak time, which is the usage status, is estimated. When both the rate and the batch processing time or latency, which are the processing times, are equal to or less than the second standard, it is estimated that the required allocation amount of the resource indicated by the past information is reduced by a predetermined ratio from the current state. There may be.

According to this, for example, when the resource usage remains high (eg 100%), it may be a state of overload that completely uses up the resource but adversely affects performance, or the resource is completely used up. It is possible to appropriately deal with and judge whether the overload condition is such that the performance is adversely affected above, using the batch processing time or the excessive length of latency as a guide for situations that were not clear in the past. Become. As a result, system resource management can be made more efficient and accurate.

Further, in the resource management support device of the present embodiment, the calculation unit performs the first processing, the number of users in the past predetermined period, the information of the financial product which is the information of the external environment, and the sales promotion regarding the system. By machine learning using past information of activity information as input, the estimated number of users of financial products and sales promotion activities under predetermined conditions at a predetermined time in the future from the predetermined period in the past is estimated. May be good.

According to this, for example, the number of users of financial services provided by a financial institution is sensitive to financial environment information such as stock prices and interest rates, that is, financial information, and sales promotion activities (commercial in mass media) conducted by the financial institution. It will be possible to accurately estimate the number of future users in light of the situation where many people are affected by broadcasting, sending direct mail, etc.). As a result, system resource management can be made more efficient and accurate.

Further, in the resource management support device of the present embodiment, the calculation unit merges the first process and the second process, and the number of users, the external environment, and the data processing amount in the past predetermined period. By machine learning with the past information of the above and the calendar characteristics of the period as input, the amount of predicted data processing with a predetermined number of predicted users at a predetermined time in the future from the past predetermined period is estimated. May be good.

According to this, it becomes possible to estimate the number of users in the future more efficiently, and by extension, the resource management of the system can be made more efficient and accurate.

Further, in the resource management support method of the present embodiment, when the information processing apparatus has both the usage status and the processing time equal to or higher than the first reference in the fourth processing, the resource indicated by the past information is required. It is estimated that the allocation amount is increased by a predetermined ratio from the current state, and when both the usage status and the processing time are equal to or less than the second standard, the required allocation amount of the resource indicated by the past information is decreased by a predetermined ratio from the current state. It may be presumed to be.

Further, in the resource management support method of the present embodiment, when the information processing apparatus performs the fourth processing, the maximum usage rate or the average usage rate at the peak time, which is the usage status, and the batch processing time, which is the processing time, are used. Or, if both the latency and the latency are equal to or higher than the first standard, it is estimated that the required allocation amount of the resource indicated by the past information is increased by a predetermined ratio from the current state, and the maximum usage rate or the average at the peak time, which is the usage status. When both the usage rate and the batch processing time or latency, which are the processing times, are equal to or less than the second standard, it is estimated that the required allocation amount of the resource indicated by the past information is reduced by a predetermined ratio from the current state. May be.

Further, in the resource management support method of the present embodiment, the information processing apparatus relates to the number of users in the past predetermined period, information on financial products which is information on the external environment, and the system in the first processing. It may be possible to estimate the estimated number of users of financial products and sales promotion activities under predetermined conditions at a predetermined time in the future from the predetermined period in the past by machine learning using past information of sales promotion activities as input. ..

Further, in the resource management support method of the present embodiment, the information processing apparatus merges the first process and the second process, and the number of users, the external environment, and the data process in the past predetermined period. It may be possible to estimate the predicted data processing amount with a predetermined number of predicted users at a predetermined time in the future from the predetermined period in the past by machine learning using the past information of the amount and the calendar characteristics of the period as input. ..

10 Network 100 Resource management support device 101 Storage unit 102 Program 1021 Machine learning algorithm 103 Memory 104 Calculation unit 105 Input unit 106 Output unit 107 Communication unit 125 Number of active subscribers Information 126 Financial information 127 Sales promotion information 128 Resource-related information 200 User terminal 300 Target system

Claims (8)

A storage unit that stores past information on the number of users, data processing volume, resource status, and external environment in the managed system.
By machine learning by inputting past information of the number of users in the past predetermined period , information on financial products which is information on the external environment, and information on sales promotion activities related to the system, a predetermined period in the future from the predetermined period in the past. The first process of estimating the estimated number of users of a given financial product and sales promotion activity under a given situation at a certain time, the data processing amount in the past predetermined period, the past information of the number of users, and the calendar of the period. The second process of estimating the predicted data processing amount for the predicted number of users in the predetermined period in the future from the predetermined period in the past by machine learning using the characteristics as input, and the past of the resource status in the predetermined period in the past. Based on the information, it is determined whether the resource usage status and processing time are equal to or longer than the predetermined standard, and the required allocation amount of the resource indicated by the past information is increased or decreased by a predetermined rate from the current state according to the result of the determination. By machine learning with the input of the third process to be estimated, the required allocation amount of the resource estimated for the predetermined period in the past, and the peak value of the data processing amount in the period, at a predetermined time in the future from the predetermined period in the past. A calculation unit that executes the fourth process of estimating the predicted required allocation amount in the predicted data processing amount, and
A resource management support device characterized by being equipped with. The arithmetic unit
In the third processing, when both the usage status and the processing time are equal to or longer than the first reference, it is estimated that the required allocation amount of the resource indicated by the past information is increased by a predetermined ratio from the current state, and the usage status is described. When both the processing time and the processing time are equal to or less than the second standard, it is estimated that the required allocation amount of the resource indicated by the past information is reduced by a predetermined ratio from the current state.
The resource management support device according to claim 1. The arithmetic unit
In the third process,
When the maximum usage rate or average usage rate at the peak time, which is the usage status, and the batch processing time or latency, which is the processing time, are both equal to or higher than the first criterion, the required allocation amount of the resource indicated by the past information. Is estimated to have been increased by a predetermined percentage from the current situation,
When the maximum usage rate or average usage rate at the peak time, which is the usage status, and the batch processing time or latency, which is the processing time, are both equal to or less than the second criterion, the required allocation amount of the resource indicated by the past information. Is estimated to have been reduced by a predetermined percentage from the current situation.
The resource management support device according to claim 2. The arithmetic unit
Machine learning by merging the first process and the second process and inputting past information of the number of users, the external environment, and the amount of data processed in the past predetermined period, and the calendar characteristics of the period. Therefore, the amount of predicted data processing with a predetermined number of predicted users in a predetermined time in the future is estimated from the predetermined period in the past.
The resource management support device according to claim 1. An information processing device equipped with a storage unit that stores past information on the number of users, data processing amount, resource status, and external environment in the managed system.
By machine learning using past information of the number of users in the past predetermined period , information on financial products which is information on the external environment, and information on sales promotion activities related to the system, a predetermined period in the future from the predetermined period in the past. The first process of estimating the estimated number of users of a given financial product and sales promotion activity under a given situation at a certain time,
By machine learning using the past information of the data processing amount and the number of users in the past predetermined period and the calendar characteristics of the period as input, the predicted number of users in the predetermined time in the future from the past predetermined period The second process for estimating the amount of predicted data processing and
Based on the past information of the resource status in the past predetermined period, it is determined whether the resource usage status and the processing time are equal to or longer than the predetermined standard, and the required allocation amount of the resource indicated by the past information is determined according to the result of the determination from the current state. The third process, which is presumed to have increased or decreased by a predetermined rate,
By machine learning using the required allocation amount of the resource estimated for the past predetermined period and the peak value of the data processing amount in the period as inputs, the predicted data processing amount at the predetermined time in the future from the past predetermined period. The fourth process for estimating the expected required quota, and
A resource management support method characterized by executing. The information processing device
In the third processing, when both the usage status and the processing time are equal to or longer than the first reference, it is estimated that the required allocation amount of the resource indicated by the past information is increased by a predetermined ratio from the current state, and the usage status is described. If both the processing time and the processing time are equal to or less than the second standard, it is estimated that the required allocation amount of the resource indicated by the past information is reduced by a predetermined ratio from the current state.
The resource management support method according to claim 5, characterized in that. The information processing device
In the third processing, when the maximum usage rate or the average usage rate at the peak time, which is the usage status, and the batch processing time or the latency, which is the processing time, are both equal to or higher than the first reference, the past information indicates. It is estimated that the required allocation amount of the resource is increased by a predetermined ratio from the current state, and both the maximum usage rate or the average usage rate at the peak time, which is the usage status, and the batch processing time or latency, which is the processing time, are the first. If it is less than or equal to the standard of 2, it is estimated that the required allocation amount of the resource indicated by the past information is reduced by a predetermined ratio from the current state.
The resource management support method according to claim 6, wherein the resource management support method is characterized. The information processing device
Machine learning by merging the first process and the second process and inputting past information of the number of users, the external environment, and the amount of data processed in the past predetermined period, and the calendar characteristics of the period. To estimate the amount of predicted data processing with a predetermined number of predicted users in a predetermined time in the future from the predetermined period in the past.
The resource management support method according to claim 5, characterized in that.

Images