JP2015056087A - Server load balancing method and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To determine a server which should execute information processing requested from a client, in consideration of contents of the information processing requested from the client and relations between loaded states and response times.SOLUTION: A server load balancing system includes a response record data storage unit in which contents of information processing, loaded state information, and averages (records) of response times from acceptance of requests to response to them are stored per server. The server load balancing system receives a request for information processing from a client, receives loaded state information from each of a plurality of servers, and determines a server which should execute the requested information processing on the basis of contents of information processing relating to the received request for information processing, the received loaded state information of respective servers, and information in the response record data storage unit.

Description

  The present invention relates to a server load distribution technique for determining an optimum server for executing requested information processing from a plurality of servers.

  A server load balancer is a general system that achieves high-speed and efficient processing in a server-client type computer system by performing requests requested by clients in parallel on a plurality of servers having the same function. In particular, it is a key technology for speeding up and improving the efficiency of a large amount of data called big data.

  In recent years, with the development of sensor technology, a mechanism that provides services such as monitoring and control based on information from sensors equipped with devices and devices (things) called M2M (Machine to Machine) has been attracting attention. Yes. Also, in the construction equipment industry, in order to minimize the downtime of construction equipment for mine development, analyzing information collected from sensors installed in motors, hydraulic gauges, etc. in real time leads to abnormalities and failures. It is expected to detect signs and perform control to avoid them.

  However, in order to realize a service for minimizing the downtime of mining development construction equipment as described above, an information processing device such as a server is sometimes used from many construction equipment operating at the mining development site. It is necessary to perform information processing such as analysis of a large amount of data generated in real time in real time. Therefore, there is a demand for a system that can perform information processing faster and more efficiently than before.

  Therefore, paying attention to the information processing contents and server load information that are considered to be closely related to the response time required for the information processing requested by the client, the result of accumulating and analyzing the processing results is the information requested by the client. If it is used when determining a server to perform processing, there is a possibility that efficient information processing can be performed at a higher speed than a conventional system. Also, if this system that can process information faster and more efficiently than before is applied to a service that minimizes the downtime of construction equipment for mining development, it will lead to abnormalities and failures compared to conventional M2M. There is a possibility that control can be performed to detect a sign earlier and to increase the probability of avoiding an abnormality or failure.

  As a related technology, considering each real-time server load state and the content of the information processing requested by the client, each server that performs information processing is determined, and the server load that performs the requested information processing on that server A dispersion technique is known (see, for example, Patent Document 1).

JP 2011-138202 A

  In recent years, due to the development of sensor technology, a mechanism for providing services such as monitoring and control based on information from sensors equipped with devices and devices (things) called M2M has been attracting attention. Also, in the construction equipment industry, in order to minimize the downtime of construction equipment for mine development, analyzing information collected from sensors installed in motors, hydraulic gauges, etc. in real time leads to abnormalities and failures. It is expected to detect signs and perform control to avoid them. However, in order to realize a service for minimizing the downtime of mining development construction equipment as described above, an information processing device such as a server is sometimes used from many construction equipment operating at the mining development site. It is necessary to perform information processing such as analysis of a large amount of data generated in real time in real time, and a system capable of responding faster than before is required. As a conventional server load distribution device, there is a mechanism for distributing processing to each server in consideration of the real-time server load state and the information processing information requested from the client, as disclosed in Patent Document 1, but the request from the client The problem is that the requested information processing cannot be performed by the server with the fastest response even if the load state is high, without considering the relationship between the contents of the processed information, the load state of the server, and the response time. There is.

  The present invention has been made in view of such circumstances, and performs information processing requested by a client in consideration of the contents of information processing requested from the client, the load state of the server, and the response time. It is an object of the present invention to provide a server load distribution method and program for determining a server.

  A typical example of the present invention is as follows. That is, the present invention is a server load distribution method in a server load distribution apparatus that determines a server that performs information processing requested by a client from a plurality of servers. The server load distribution apparatus includes a response record data storage unit that stores information processing contents, load state information, and an average value (result) of response time from receiving a request to responding for each server. Then, the server load distribution device receives an information processing request from the client, receives load status information from each of the plurality of servers, and receives the information processing content and the received information processing request. The server that executes the requested information processing is determined based on the load status information of each server and the information in the response record data storage unit.

  According to the present invention, it is possible to determine the server that performs the information processing requested from the client in consideration of the relationship between the content of the information processing requested from the client, the load state of the server, and the response time. Accordingly, the requested information processing can be performed on the server that is predicted to perform the requested information processing earliest.

It is a figure which shows the example of whole structure of the computer network system 1 which concerns on this embodiment. FIG. 2 is a diagram illustrating a hardware configuration example of a server load distribution device 16. It is a figure which shows the structure and flow of each process of the server load distribution apparatus. It is a figure which shows the data structural example of the process request table 210 with which the server load distribution apparatus 16 is provided. It is a figure which shows the data structural example of the real-time server load information table 220 with which the server load distribution apparatus 16 is equipped. It is a figure which shows the example of a data structure of the past response performance table 230 which the server load distribution apparatus 16 comprises. It is a figure which shows the example of a data structure of the past response performance space table 240 which the server load distribution apparatus 16 comprises. It is a figure which shows the concept of the past response performance space table 240 which the server load distribution apparatus 16 comprises. It is a figure which shows the example of a data structure of the present information and past results mapping result table 250 which the server load distribution apparatus 16 comprises. It is a figure which shows the example of a data structure of the response time table 270 of each server with which the server load distribution apparatus 16 is provided. 6 is a diagram illustrating an example of a data configuration of a response performance table 270 by the server load distribution device 16. FIG. 10 is a flowchart showing processing of the entire processing distribution destination server determination function 1610 provided in the server load distribution device 16. 10 is a flowchart showing processing of a subroutine past information statistical processing subroutine of the processing distribution server decision function 1610 by the server load distribution device 16. 15 is a flowchart showing processing of a subroutine past information mapping subroutine of the server load distribution device 16 of a process distribution destination server determination function 1610. 10 is a flowchart showing processing of a subroutine response record generation / storage subroutine of the processing distribution destination server determination function 1610 by the server load distribution device 16.

  Hereinafter, an embodiment of the present invention will be described in detail with reference to the drawings.

  FIG. 1 is a diagram illustrating an example of the overall configuration of a computer network system 1 according to the present embodiment. The server load balancer 16 and the sensors (client computers) 11, 13, 15 installed in the mining construction devices 10, 12, 14 and the information processing requested by the mining construction devices 10, 12, 14 Servers (information processing devices) 17, 18, and 19 that perform information processing such as analysis of mining construction equipment information based on the contents are connected to a network. The server load balancer 16 responds to requests received from the sensors 11, 13, and 15 installed in the mining construction devices 10, 12, and 14, and details of requested information processing and real-time load status of each server Is mapped to the past response record, and the server that is estimated to have the shortest response time is determined as the server that performs the requested information processing. Further, the server load balancer 16 receives the result (response history) of the requested information processing from the server, and accumulates it as past response history data for determining the next processing distribution destination.

  FIG. 2 is a diagram illustrating a hardware configuration example of the server load distribution device 16. The server load balancer 16 can be realized by a general computer, and includes a processing allocation server determination function 1611 as a program 1610 in a control unit (CPU) 160 and a storage unit 161, an input unit 162, and a communication unit 163. . The storage unit 161 also includes a processing distribution destination server determination function 1611 and a past response record accumulation database 1612. The program may be stored in advance in the storage device 161, or stored in a portable storage medium that can be used by a computer, and read out as necessary via a reading device (not shown). Alternatively, it may be downloaded as necessary from another device connected to a network, which is a communication medium that can be used by a computer, and stored in the storage device 161.

  FIG. 3 is a diagram showing the configuration and flow of each process of the server load distribution device 16. The request reception processing 21 operates on the processing distribution destination server determination function 1611 shown in FIG. 2, and based on the content of the request transmitted from the sensor attached to each mine construction equipment, the processing request shown in FIG. This is processing for generating the table 210. The processing request table 210 includes a processing ID 211 for uniquely identifying a request waiting for processing as a main key, and a construction device that has transmitted the request as another column. The construction equipment ID 212 for identifying, the time 213 when the request is received, the construction equipment operation information 214, and the processing content 215 requested by each mine construction equipment are stored.

  The server load state reception process 22 operates on the processing distribution destination server selection function 1611 shown in FIG. 2, and acquires, for example, CPU usage rate, memory usage rate, and disk usage rate as real-time server load status information from each server. The real-time server load information table 220 shown in FIG. The real-time server load information table 220 stores the date and time 211 and the server name 212 as primary keys, and other columns such as the CPU usage rate 213, the memory usage rate 214, and the disk usage rate 215 as real-time server load status information. To do. The server load state information may be any combination of the CPU usage rate, the memory usage rate, and the disk usage rate, and may include parameters other than the CPU usage rate, the memory usage rate, and the disk usage rate. .

  The past response record statistics process 23 operates on the processing distribution destination server determination function 1611 shown in FIG. 2, and the server name 232 and the past response record table 230 in the past response record storage database 1612 shown in FIG. This is a process of generating the past response record space table 240 shown in FIG. 7 from the past record obtained by using the process contents 233 as a key. The response result space table 240 includes, as main keys, a server name 241, processing contents 242, a CPU usage level 243 indicating the level of CPU usage, a memory usage level 244 indicating the level of memory usage, and a level of disk usage. Is stored, and the average response time 246 of the corresponding divided section is stored as other columns. Here, FIG. 8 shows a conceptual diagram regarding the load level (CPU usage rate level 243, memory usage rate level 244, disk usage rate level 245) of each server. The load state level of each server is a rectangular parallelepiped space generated from the CPU usage rate, the memory usage rate, and the disk usage rate. The CPU usage rate is equally divided into i, the memory usage rate is divided into j, and the disk usage rate is related. It shows which box belongs when divided into k × equally, i × j × k solid (for example, rectangular parallelepiped) boxes as a whole. Further, the average response time 246 of the server in the corresponding divided section is calculated by calculating the average value of the response time of the server in the corresponding space for each of the cuboid spaces divided into the i × j × k. The reason why the i, j, and k are arbitrarily determined in the system unit to which the device 10-19 is connected as shown in FIG. 1 is that the device 10-19 is connected as shown in FIG. This is because parameters indicating each server load state that can affect the response time differ depending on the system configuration. For example, when a general WEB / AP server is considered as the processing distribution destination server 17-19, it is considered that the CPU usage rate affects the response time rather than the disk usage rate. In this case, the CPU usage rate is divided into 10 (i = 10) and the disk usage rate is not divided (k = 1), so that the load status of the CPU usage rate, which is considered to strongly influence the response time of the WEB / AP server, is determined. It is possible to assign processing with importance. Conversely, when a general database server is considered as the processing distribution destination server 17-19, it is considered that the disk usage rate also greatly affects the response time of the database server. In this case, the CPU usage rate is divided into 10 (i = 10) and the disk usage rate is also divided into 10 (k = 10), so that processing is performed with the same importance as the disk usage rate. It becomes possible to distribute.

  Further, for example, a WEB / AP server group that implements an application that expands most of the results of the issued SQL statement to a memory for a linked database server is set as a processing distribution destination server 17-19. The server allocating server 17 is a WEB / AP server group in which an application that expands only a part of the data of the issued SQL statement to the memory is connected to the system configuration A and the linked database server. Consider system configuration B as -19. In this case, when processing is distributed in the system configuration A, it is necessary to consider the memory usage rate of each WEB / AP server than when processing is distributed in the system configuration B. Therefore, in the system configuration B, j = 2. If the memory usage rate does not exceed 50%, there is no problem so far. However, in the system configuration A, it is possible to make adjustments such as j = 10 and determining whether to distribute the processing with higher sensitivity.

  The current information / past response record mapping process 24 operates on the process distribution destination server determination function 1611 shown in FIG. 2, and requests the process request table 210, the real-time server load information table 220, and the response record space table 240. This is a process of generating the current information / past performance mapping result table 250 shown in FIG. 9 by mapping the processing contents, server name, CPU usage level, memory usage level, and disk usage level to keys. The current information / past performance mapping result table 250 includes a processing ID 251 and a server name 252 as primary keys, a date and time 253, a processing content 254, a CPU usage rate level 255, a memory usage rate level 256 as other columns, The disk usage rate level 257 and the average response time 258 of the corresponding divided section are stored. Based on the current information / past performance mapping result table 250, it can be predicted how long the response time will be required when the corresponding process is executed on which server. As described above, the real-time request processing contents and the real-time load state of each server are mapped to the corresponding divided boxes in the server load state space by the current information / past response actual result mapping process 24, thereby realizing real-time You can predict how long it will take to respond to a request.

  The process allocation destination server determination process 25 operates on the process allocation destination server determination function 1611 shown in FIG. 2 and uses the process ID 251 and the server name 252 as keys from the current information / past performance mapping result table 250. The average response time 258 is taken out, and the request receiving process 21 stores a process distribution waiting request so that the process is executed on the server considered to have the shortest response time from the comparison result of the average response time 258 of the corresponding divided section. Retrieve the request from the queue and execute the process. As a result, the processing is distributed to the server that is expected to process the processing contents earliest.

  The response information reception process 26 operates on the process allocation destination server determination function 1611 shown in FIG. 2, acquires the result executed by the corresponding process allocation destination server based on the notification result of the process allocation destination server determination process 25, It is a process which produces | generates the response time table 260 of each server shown in FIG. The response time table 260 includes a processing ID 261 as a primary key, a server name 262 as other columns, a processing content 263, a CPU usage rate 263 as a load status information of the server when processing, a memory usage rate 264, The disk usage rate 265 is stored as a response time 266 when the server completes the process and returns a response to the request.

  The response record generation process 27 operates on the process distribution destination server determination function 1611 shown in FIG. 2, calculates the response time required for the corresponding process from the process request table 210 and the response time table 260 using the process ID as a key. 11 is a process of generating a response record table 270 shown in FIG. The response record table 270 includes the date / time 271, the server name 272, and the processing content 273 as the main key, the CPU usage rate 274 when the corresponding processing is performed as the other columns, the memory usage rate 275, and the disk usage rate. The response time 277 required for the corresponding process is stored as the rate 276.

  The response record storage process 28 operates on the process assignment destination server determination function 1611 shown in FIG. 2, and the response record table 270 generated by the response record generation process 27 is stored in the past response record storage database 1612 shown in FIG. It is a process to store in.

FIG. 12 is a flowchart showing the entire processing of the processing distribution destination server determination function 1611 provided in the server load distribution device 16. The process distribution destination server determination function 1611 is started by the request reception process (step S100), and the construction equipment ID 212, the request time 213, the construction equipment operation information 214, and the processing contents transmitted from the sensor attached to each mine construction equipment. The process request table 210 is generated by assigning a process ID 211 that uniquely identifies the process awaiting execution to the process 215 (in this example, the records 2101 and 2102 are included). Next, in the server load state reception process (step S101), real time is obtained from the date and time 221, the server name 222, the CPU usage rate 223, the memory usage rate 224, and the disk usage rate 225 acquired from the servers 17, 18, and 19 that process the request. A server load state information table 220 (in this example, records 2201, 2202, and 2203 are generated) is generated. An overall flowchart of the past operation information statistical processing subroutine (step S102) is shown in FIG. In the past response record acquisition process (step S200), the past response record table 230 is acquired from the past response record storage database 1612 (in this example, records 2301 to 230C are included). In the past CPU usage level C calculation process (step S201), the past CPU usage level C is calculated from the CPU usage rate c in the past response record table 230 by the following formula. Calculate for each record (calculate for each record 2301 to record 230C).
C = [c / i]
Here, i ≠ 0 (10 in this example) is an integer representing the fineness of dividing the CPU usage rate arbitrarily determined by the user, and [c / i] is the value obtained when c is divided by i. It means the value rounded to the first decimal place. For example, in the record 2301 of the past response record table 230, since the CPU usage rate 234 is 12%, the past CPU usage rate level C = [12/10] = [1.2] = 1. In the past memory usage level M calculation process (step S202), M is calculated for each record of the past response record table 230 from the memory use rate m of the past response record table 230 by the following formula (record). 2301 to 230C for each record).
M = [m / j]
Here, j ≠ 0 (10 in this example) is an integer representing the fineness of dividing the memory usage rate arbitrarily determined by the user, and [m / j] is the value obtained when m is divided by j. It means the value rounded to the first decimal place. For example, in the record 2302 of the past response record table 230, since the memory usage rate 235 is 19%, the past memory usage rate level M = [19/10] = [1.9] = 2. In the past disk usage level D calculation process (step S203), the past disk usage level D is calculated from the disk usage rate d of the past response record table 230 by the following formula, and each record of the past response record table 230 is recorded. (Each record 2301 to record 230C is calculated).
D = [d / k]
Here, k ≠ 0 (10 in this example) is an integer representing the fineness of dividing the disk usage rate arbitrarily determined by the user, and [d / k] is the value obtained when d is divided by k. It means the value rounded to the first decimal place. For example, in the record 2303 of the past response record table 230, since the disk usage rate 236 is 13%, the disk usage rate level D = [13/10] = [1.3] = 1. In the average response time calculation process (step S204) for each divided box, the past CPU obtained in the CPU usage level C calculation process (step S201) from each process content, server name, and past response record table 230. Usage rate level C, past memory usage level M calculated by the memory usage level M calculation processing (step S202) from the past response record table 230, and past disk usage level D calculation from the past response record table 230 The average response time is calculated using the past disk usage rate level D obtained in the process (step S203) as a key. For example, the server A calculates the average value by the CPU usage level C calculation process (step S201), the memory usage level M calculation process (step S202), and the past disk usage level D calculation process (step S203). It can be seen that the record records of the past response record table 230 when records are processed at the CPU usage rate level 1, the memory usage rate level 1, and the disk usage rate level 1 are records 2301 and 2303. Therefore, the value of the average response time 246 of the corresponding divided section in the record 2401 of the past response record space table 240 is (5 + 7) / 2 = 6. Further, the server B performs an abnormality detection process by the CPU usage level C calculation process (step S201), the memory usage level M calculation process (step S202), and the past disk usage level D calculation process (step S203). It can be seen that the record 2306 is the only actual record in the past response record table 230 when processing is performed at the CPU usage rate level 1, the memory usage rate level 2, and the disk usage rate level 1. Therefore, the value of the average response time 246 of the corresponding divided section in the record 2405 of the past response record space table 240 is 60/1 = 60. In the past operation information statistical processing subroutine (step S102), the past response record space table 240 is obtained by the past response record acquisition process (step S200) to the average response time calculation process (step S204) for each of the divided boxes. Is generated. FIG. 14 shows an overall flowchart of the current information / past performance mapping processing subroutine (step S103). In the current CPU usage level C calculation process (step S301), C is calculated from the CPU usage rate c in the real-time server load state information table 220 by the same formula as in the past CPU usage level C calculation process (step S201). To do. For example, in the record 2201 of the real-time server load state information table 220, the current CPU usage rate level C is C = [10/10] = 1. In the current memory usage rate level M calculation process (step S302), M is calculated from the memory usage rate m in the real-time server load state information table 220 by the same formula as in the past memory usage level M calculation process (step S202). To do. For example, in the record 2202 of the real-time server load state information table 220, the current memory usage level M is M = [20/10] = 2. In the current disk usage rate level D calculation process (step S303), D is calculated from the disk usage rate d in the real-time server load state information table 220 by the same calculation formula as the past disk usage level D calculation process (step S203). To do. For example, in the record 2203 of the real-time server load information table 220, the current disk usage rate level D is D = [30/10] = 3. In the current information and past information mapping process (step S303), the process request table 210, the real-time server load state information table 220, and the past response record space table 240 are stored in the server name, processing content, CPU usage level, memory usage level, This is a process for generating the current information / past performance mapping result table 250 by performing the mapping process using the disk usage rate level as a key. In this mapping process, first, a simple join process (join) is performed on the process request table 210 and the real-time server load state information table 220. In this combination processing, the number of requests processing times the number of servers is generated. For example, in the example of the processing request table 210 and the real-time server load state information table 220, since the processing is two processing IDs 10000 and 10001, and the servers are server A, server B, and server C, 2 × 3 = 6 records. Become. Next, with respect to the combined processing results of the processing request table 210 and the real-time server load state information table 220, the past response result space table 240 is processed, the server name, the CPU usage level, the memory usage level, and the disk usage rate level. The current information / past performance mapping result table 250 is generated by performing a join process (join) using as a key. For example, consider the record 2501 of the current information / past performance mapping result table 250. In this example, when the average value calculation process is currently executed on the server A having the CPU usage rate level 1, the memory usage rate level 1, and the disk usage rate level 1, how much response time has been required in the past. Calculated. The processing content indicated by the record 2101 of the processing request table 210 is the average value calculation, the CPU usage level of the server A of the record 2201 of the real-time server load state information table 220 is 1, the memory usage level is 1, and the disk usage rate Since the level is 1, in the past response record space table 240, a record 2401 having the same processing content, server name, CPU usage rate level, memory usage rate level, and disk usage rate level is mapped. As a result of this mapping, a record 2501 of the current information / past performance mapping result table 250 is generated. Also, records of the past response record space table 240 with the processing contents of the records of the processing request table 210 and the records of the real-time server load state information table 220, the server name, the CPU usage rate level, the memory usage rate level, and the disk usage rate level. If there is nothing that matches the processing content, server name, CPU usage rate, memory usage level, and disk usage rate level, as shown in the record 2502 of the current information / past performance mapping result table 250, Cannot be mapped to average response time. However, when mapping is impossible, the response time results are positively accumulated by setting the response time results to 0. In the process assignment destination server determination process (step S104), a record is extracted from the result of the current information / past performance mapping result table 250 using the process ID 251 as a key, and the process is executed on the server having the fastest average response time 258 in the corresponding divided section. The request reception process 21 notifies the queue storing the process distribution waiting. For example, in the process ID 10001 of the current information / past performance mapping result table 250, when executed from the record 2504 on the server A, the average response time of the corresponding divided section is 6 seconds, and when executed on the server B from the record 2505, When the average response time of the divided section is 12 seconds and the server 2 C executes from the record 2506, the average response time of the corresponding divided section is calculated as 18 seconds. An execution request at the shortest server A is notified to the queue storing the process distribution waiting by the request reception process 21. Also, in the process ID 10000 of the current information / past performance mapping result table 250, when executed from the record 2501 on the server A, the average response time of the corresponding divided section is 30 seconds, and when executed on the server 250 from the record 2502, If the average response time of the divided section is 0 seconds (mapping is impossible), and the record 2503 is executed on the server C, the average response time of the corresponding divided section is calculated as 0 seconds (mapping is impossible). Execution is notified to the queue storing the process distribution waiting in the request reception process 21 (if the average response time of the corresponding divided section is the same, the distribution destination is determined in ascending order of the server name). FIG. 15 shows an overall flowchart of the response record generation / storage subroutine (step S105). The response information reception process (step S400) is a process of generating a response time table 260 of each server indicating the result of processing the process distributed by the process distribution destination server process (step S104) by the servers 17, 18, and 19. . For example, in the record 2601, the process ID 10000 is executed by the server A and the abnormality detection is executed in a state where the CPU usage rate is 10%, the memory usage rate is 10%, and the disk usage rate is 10%. 2013/1/21 10:02: 00 indicates that the processing result is returned from the server A. The response record generation process (step S401) is a process of calculating a response record using the process request table 210 and the response time table 260 of each server as a process ID. For example, the response time of the process ID 10000 is the response time 2013 of the record 2601 of the process ID 10000 of the response time table 260 of each server from the request time 2013/1/21 10:00 of the record 2101 of the process ID 10000 of the process request table 210. / 1/21 120: difference of 10:02:00 is stored in the response record table 270 in the form of a record 2701 as a response time. The response record storage process (step S402) is a process of storing the response record generated in the response record generation process (step S401) in the past response record storage database 1612.

  The embodiment of the present invention has been described above.

  According to the above embodiment, analysis of a large amount of data generated from sensors attached to construction equipment for each mine development is performed by causing a server that seems to be faster to respond according to past response results. Speed up and improve the availability of construction equipment for mine development.

  Although the embodiment of the present invention has been described above, the present invention is not limited to this, and various modifications can be made without departing from the scope of the invention.

16 ... Server load balancer, 10, 12, 14 ... Mining construction equipment, 11, 13, 15 ... Sensors attached to mining construction equipment, 11, 13, 15 ... Mining development Construction equipment information analysis server, 160 ... control unit (CPU), 161 ... storage unit, 162 ... control unit (CPU).

Claims (6)

A server load balancing method in a server load balancer that determines a server that performs information processing requested by a client from a plurality of servers,
The server load distribution apparatus includes a response result data storage unit that stores the average value (actual result) of response time from receiving the content of information processing, load state information, and a request for each server,
By the server load balancer,
Receiving an information processing request from the client;
Receiving load status information from each of the plurality of servers;
Based on the contents of the information processing of the received information processing request, the received load status information of each server, and the information in the response record data storage unit, the server that performs the requested information processing is determined. ,
A server load balancing method. By the server load balancer,
Responds after receiving the requests from a plurality of servers based on the contents of the information processing of the received information processing request, the received load status information of each server, and information in the response record data storage unit The server with the shortest response time is determined as the server that performs the requested information processing.
The server load distribution method according to claim 1. The load status information of each server includes a CPU usage rate, a memory usage rate, and a disk usage rate.
The server load distribution method according to claim 2, wherein: The average value (actual result) of the response time from the receipt of the request to the response stored in the response achievement data storage unit is the CPU usage rate, the memory usage rate, and the disk usage rate, i, j, k (i ≠ 0, j ≠ 0, k ≠ 0) are equally divided into i × j × k three-dimensional space average values (actual results),
By the server load balancer,
Receiving the load status information including the CPU usage rate, the memory usage rate and the disk usage rate from the plurality of servers;
Calculate the values obtained by dividing the received CPU usage rate, memory usage rate, and disk usage rate by i, j, and k, respectively.
A plurality of servers based on the information processing contents of the received information processing request, the values calculated based on the received load status information of the servers, and the response record data storage unit information To determine the server having the shortest response time from receiving the request to responding to the server that performs the requested information processing.
The server load distribution method according to claim 3. Receiving from the server information including information identifying the server, information processing information, the load status information, and a response time from receiving a request to responding;
Response history data using the information for identifying the received server, the content of the information processing, the load status information, information including the response time from when the request is received until the response is received, and the time when the request is received Produces
Storing the generated response record data in the response record data storage unit;
The server load distribution method according to claim 4, wherein:   A program for causing a computer to execute the server load distribution method according to any one of claims 1 to 5.

Images