JP5870635B2 - Processing request execution management apparatus and program - Google Patents

Description

  The present invention relates to a processing request execution management apparatus and a program.

  Patent Document 1 discloses a system for performing distributed processing with priority given to specific data processing even if data processing requests from each terminal apparatus are concentrated, to an allowed specific user. Has been. In this system, when a data processing request is received from each client, the user attribute of the data request source is determined based on the stored determination file, and the data processing request from a specific user is distributed to other information processing apparatuses. To give priority processing.

  Patent Document 2 discloses a system in which independent computing resources are allocated to each user company in a data center, and the allocation is automatically changed in real time according to a load. In this system, the control program creates a computer allocation control table for each user company identifier and sets it in the load balancer. Means are provided for identifying the user company identifier from the IP address in the user request packet. The load balancer identifies the user company from the user request packet and forwards it to the appropriate computer group. Further, the control program compares the service level contract for each user with the operation status monitor result of the computer, and increases the allocated computer when the contract condition is not met. Specifically, the control table is changed and reset in the load balancer. Further, the control program changes the billing information based on it.

  In the dynamic load distribution system disclosed in Patent Document 3, when a plurality of Web servers process a plurality of tasks in a distributed manner, the response is monitored outside the Internet, and resources such as the CPU usage rate on each server side are monitored. Monitors and distributes work to each server according to a predetermined schedule rule, and when a server whose response time is delayed by exceeding a preset response rule threshold is detected, or a preset server resource rule When a server having a high CPU usage rate exceeding the upper limit value is detected, the server monitoring / control agent dynamically changes the allocation of a plurality of tasks to each Web server.

JP 2002-140299 A JP 2002-024192 A JP 2005-182641 A

  The present invention accepts a processing request of a user in a customer organization outside the response guarantee time zone when the data processing system has sufficient capacity, but responds to a response guarantee time zone in a high load state. The purpose is to be able to guarantee the response of the customer organization within.

The invention according to claim 1, for each of the plurality of client organization, and guaranteed time period is a time period during which a response guarantee, a storage means for storing, the contents of the response assurance in the guarantee time period The content of the response guarantee includes at least the maximum number of simultaneous request processes assigned to the customer organization, and performs data processing in response to a processing request from each user in the plurality of customer organizations. Monitors the load of the data processing system to be executed and capable of simultaneously processing the maximum value of all the maximum simultaneous request processes of each customer organization at the same time in all time zones And a period during which a high load state of the data processing system is detected by monitoring by the monitoring means is a guarantee time that is a customer organization that includes the period within the guarantee time zone Processing request from the user within the organization may be executed in the data processing system, the processing requirements of the period from the user of the guaranteed time zone comprise not within the guarantee time outside the organization is the customer organization to the data processing system A processing request execution management device comprising: control means for performing control so as not to be executed.

  According to a second aspect of the present invention, the control means includes a queue that temporarily holds a processing request from each user in each customer organization, and a high load state of the data processing system is detected. When there is no period, the processing request is extracted from the queue in order from the top of the queue and executed by the data processing system without distinguishing whether the processing request is from a user within the guarantee time organization or an organization outside the guarantee time. In addition, during a period when a high load state of the data processing system is detected, a processing request from a user in an organization outside the guarantee time is not taken out from the queue, and only a processing request from a user in the organization within the guarantee time is The processing request execution management apparatus according to claim 1, wherein the processing request execution management apparatus extracts the data from the queue and causes the data processing system to execute the data processing system.

Invention of claim 3, prior Symbol control means, said data processing high load state of the system in the period that is detected when the guarantee time the organization there are a plurality, they in each guaranteed time from the queue ratio of processing requests tissue is taken out, according to claim 1 or 2 which is controlled so that the one corresponding to the ratio of the maximum simultaneous number of request processing between assigned to each guaranteed time within the organization, and wherein the The processing request execution management device described in 1.

  The invention according to claim 4 further includes a re-execution determination unit that determines whether or not the processing request can be re-executed, and the control unit is configured to detect a high load state of the data processing system. Of the processing requests from users in the non-guaranteed organization that are being executed by the data processing system, the data processing system is forcibly terminated to execute processing requests that are determined to be re-executable by the re-executability determination unit. The processing request execution management apparatus according to claim 1, wherein

  The invention according to claim 5 further comprises request type determination means for determining whether or not the processing request is a light load type that requires less than a predetermined time for execution by the data processing system. The monitoring means monitors the number of processing requests determined to be not of the light load type among the processing requests from the customer organization that are simultaneously executed by the data processing system for each customer organization as the number of simultaneous executions. And determining whether or not the data processing system is in a high-load state based on the number of concurrent executions of each customer organization being monitored. The processing request execution management device described in 1.

According to a sixth aspect of the present invention, there is provided storage means for storing a guarantee time zone, which is a time zone in which response is guaranteed, for each of a plurality of customer organizations, and contents of the response guarantee in the guarantee time zone. The content of the response guarantee includes at least the maximum number of simultaneous request processes allocated to the customer organization, and data in response to a processing request from each user in the plurality of customer organizations. The load of a data processing system that executes processing, and that has the ability to simultaneously process the maximum value in all time zones of the total value of the maximum simultaneous request processing counts of each customer organization at the same time The monitoring means for monitoring the period during which the high load state of the data processing system is detected by the monitoring means is a customer organization that includes the period within the guaranteed time zone. That assurance processing request from a user within a time set Onai is to execute on the data processing system, the processing requirements of the period from the user of the guaranteed time zone comprise not a customer organization guarantees overtime in tissue wherein It is a program for causing a data processing system to function as a control unit that performs control to prevent execution .

  According to the first or sixth aspect of the present invention, if the data processing system is not in a high load state, it accepts a user's processing request in the organization outside the guaranteed time, but if the data processing system is in a high load state, the organization within the guaranteed time It can be controlled to satisfy the contents of the response guarantee.

  According to the second aspect of the present invention, while the data processing system is in a high load state, the number of processing requests of the organization outside the guarantee time is reduced, and the processing capacity of the data processing system generated thereby is guaranteed. It can be directed to the processing request of the organization within time.

  According to the third aspect of the present invention, the processing request can be more easily executed and the response guarantee can be easily realized as the organization within the guaranteed time with the largest number of simultaneous request processing.

  According to the fourth aspect of the present invention, rather than simply waiting for the execution of the processing request of the organization outside the guarantee time, the load of the data processing system is rapidly reduced, and the remaining power generated by the load reduction is reduced to the guarantee time. It can be directed to the processing request of the internal organization.

  According to the invention according to claim 5, rather than determining whether or not the processing request is a light load type, it is more than a method of determining whether or not a high load state is based on the number of simultaneous executions. Can be determined more appropriately.

It is a figure which shows an example of the system configuration | structure of embodiment. It is a figure which shows the example of the service level data memorize | stored in a service level data holding part. It is a figure which shows typically the guarantee content which the service level data shown in FIG. 2 shows. It is a figure which shows an example of the data structure of the measured value data of the response time which a response time recording part manages. It is a flowchart which shows an example of the procedure of the mode switching of a process request | requirement schedule part. It is a flowchart which shows an example of the procedure of the extraction process of the process request in the queue at the time of guarantee mode. It is a flowchart which shows another example of the procedure of the extraction process of the process request in the queue at the time of guarantee mode.

  With reference to FIG. 1, an embodiment of an ASP (Application Service Provider) system including a processing request execution management apparatus 10 according to the present invention will be described. This ASP system is a system for providing a data processing service to a customer via a data communication network such as the Internet. Here, it is assumed that each customer is an organization such as a company ("Company A", "Company B", etc. in FIG. 1). The ASP system receives a processing request from each user's PC (personal computer) belonging to each customer, and executes data processing corresponding to the processing request. Each customer (and individual users belonging to the customer) is registered in advance in the ASP system (for example, the processing request execution management device 10). A user belonging to each customer logs in to the system, calls a required application, and requests data processing by the application. The ASP system executes data processing according to this processing request.

  This ASP system includes a processing request execution management device 10, a service level registration unit 20, and a data processing system 30.

  The data processing system 30 is a system that executes data processing for application services for customers, and includes a plurality of data processing units 32-1, 32-2, ... 32-N (N is an integer of 2 or more), and processing. And a data storage unit 34. The data processing units 32-1, 32-2,..., 32-N (hereinafter collectively referred to as “data processing unit 32” when there is no need for distinction) are individual computers that execute applications. The data processing unit 32 realizes individual data processing according to a processing request from the user by executing a necessary application. There is no particular limitation on the application executed by the data processing unit 32. The processing data holding unit 34 is a storage device that stores data used for data processing of each data processing unit 32, and is constructed as a database, for example. The processing data holding unit 34 may be physically a single device or may be composed of a plurality of devices. Each data processing unit 32 acquires necessary data from the processing data holding unit 34 in response to a processing request from the user, and executes data processing using the acquired data. Further, if the executed data processing changes data in the processing data holding unit 34, the data processing unit 32 updates the data in the processing data holding unit 34 with the result of the data processing.

  A specific example of the flow of processing executed by the data processing unit 32 is shown below. For example, when a user in a customer accesses a web server on the data processing unit 32 from a browser on his / her PC and sends a processing request for requesting a database search UI (user interface) screen in the form of an HTTP request, The data processing unit 32 provides a web page of a search UI screen according to the request. When the user inputs a search condition on the UI screen and issues a search processing request, the data processing unit 32 searches the database of the customer in the processing data holding unit 34 according to the processing request, and performs a search. A web page displaying the result is generated and returned to the user. When the user inputs data to the database input form and issues a data input request, the data processing unit 32 updates the customer database in the processing data holding unit 34 with the input data. Note that this specific example is merely an example for explanation, and the type of application executed by the data processing unit 32 and the type of processing request for the application are not limited to those exemplified above.

  The processing request execution management apparatus 10 performs processing for allocating processing requests from users belonging to customers to the data processing units 32-1, 32-2,..., 32-N in the data processing system 30. The processing request execution management apparatus 10 includes a processing request receiving unit 11, a processing request holding unit 13, a processing request scheduling unit 15, a service level data holding unit 17, and a response time recording unit 19.

  The processing request reception unit 11 receives processing requests from users in each customer (A, B, and C companies in the illustrated example) via the network. The accepted processing request is entered in the processing request holding unit 13. The processing request holding unit 13 is a queue (queue) that holds processing requests. Each time the process request accepting unit 11 accepts a process request, the newly accepted process request is added to the end of the queue. The processing request held in the processing request holding unit 13 is taken out by the processing request scheduling unit 15 described below, and is passed to any data processing unit 32. The extracted processing request is deleted from the processing request holding unit 13.

  The processing request scheduling unit 15 distributes and executes the processing requests in the queue in the processing request holding unit 13 to each data processing unit 32 in the data processing system 30. For example, when any one of the data processing units 32 can accept the next processing request, the processing request scheduling unit 15 extracts the processing request from the processing request holding unit 13 and sends the extracted processing request to the data processing unit 32. Assign. The processing request scheduling unit 15 has at least two modes called a normal mode and a guarantee mode as modes for extracting processing requests from the queue (processing request holding unit 13). The normal mode is a general FIFO (first-in first-out) mode in which processing requests are extracted in order from the head of the queue of the processing request holding unit 13. In the guarantee mode, in order to achieve the service level guaranteed to the customer, the processing request from the customer to be guaranteed is taken out from the queue in preference to the processing request from the customer not to be guaranteed. In one example, in a situation where the service level guaranteed to all customers can be achieved, the processing request scheduling unit 15 operates in the normal mode, and when the data processing system 30 is in a high load state, the processing request The schedule unit 15 operates in the guarantee mode. The high load state is a state in which the service level guaranteed to at least one customer is no longer achieved (or is almost impossible to achieve).

  The service level data holding unit 17 stores data indicating the contents of the service level guaranteed to each customer. The service level is the content of the ASP service quality that the provider operating this system guarantees to each customer.

  FIG. 2 shows an example of service level data stored in the service level data holding unit 17. In this example, the combination of the maximum number of simultaneous sessions and the target response time is used as the ASP service quality. Here, the maximum number of simultaneous sessions is an upper limit value of the number of simultaneous sessions guaranteed by the provider to the customer. The number of simultaneous sessions is the number of processing requests from one customer that the data processing system 30 executes simultaneously. The target response time is a time from when the user's PC issues a processing request until the data processing unit 32 returns the execution result of the processing request to the user's PC (this time is referred to as “response time”). The target upper limit value. In this embodiment, the processing request scheduling unit 15 does not exceed the target response time for the response time of the data processing system 30 to the processing request from the customer until the number of processing requests from the customer reaches the maximum number of simultaneous sessions. In order to assure such, the execution order of the processing requests in the processing request holding unit 13 is controlled. Here, the target response time may be about the response time for each processing request, or may be about the average response time for the processing request from the customer in a certain period of time. . In other words, the processing request scheduling unit 15 may control the response time for each processing request from the customer so that it does not exceed the target response time of the customer, or the customer in the most recent period. Control may be made so that the average response time for the processing requests from does not exceed the target response time. The processing request schedule unit 15 may be able to select which of these two methods is used.

  In the example of FIG. 2, the ASP service level (response guarantee content) is registered for each time zone. For example, for the customer company A, response guarantee details of a maximum number of sessions of 100 and a target response time of 5 seconds are registered in the time zone from 9 am to 10 am, and for the customer company C, from 8 am to 6 pm The response guarantee contents of the maximum number of simultaneous sessions 30 and the target response time of 8 seconds are registered in this time zone. On the other hand, for the customer company B, the response guarantee content that the maximum number of simultaneous sessions is 50 and the target response time is 5 seconds from 11:00 am to noon, the maximum number of simultaneous sessions is 100 from 1 pm to 2 pm The response guarantee content of the target response time of 5 seconds is registered. The time zone in which the service level (response guarantee content) is registered in this way is hereinafter referred to as “guaranteed time zone”. In the example of FIG. 2, customers A and C register only one guaranteed time zone, and customer B registers two guaranteed time zones.

  It should be noted that the user within the customer can use the data processing system 30 (that is, send a processing request to the data processing system 30 to execute the data processing) even in a time zone other than the guarantee time zone of a certain customer. However, there is no guarantee of service level. Therefore, even if the user uses the data processing system 30 in a time zone other than the guarantee time zone, the use is limited because the response guarantee of another customer is required within the guarantee time zone. (For example, a new processing request cannot be accepted).

  The guarantee contents indicated by the service level data shown in FIG. 2 are schematically shown in FIG. In the case where there is no setting of the guaranteed time zone, the data processing system 30 has a processing capacity to simultaneously execute 230 sessions (230 processing requests) in order to satisfy the maximum number of simultaneous sessions of the customers A, B, and C. On the other hand, when the guarantee time zone of each customer is set as shown in FIG. 2, the data processing system 30 only needs to have a processing capability to simultaneously execute 130 sessions. In other words, the data processing system 30 only needs to have the ability to simultaneously process the maximum value of the total number of simultaneous sessions of each customer at the same time in all time zones.

  The service level registration unit 20 is a UI for registering service level data in the service level data holding unit 17. The service level registration unit 20 may be provided to a user in the customer (for example, a system administrator in the customer) from the processing request execution management apparatus 10 in the form of a web page, for example. In this case, the user inputs the guaranteed time zone, the maximum number of simultaneous sessions, the target response time, etc. in the input fields in the web page, and transmits the input result, so that the service level data holding unit 17 stores the customer's Service level information is registered. Similarly, the customer's registered service level data may be updated. Further, the service level registration unit 20 may be provided in the ASP system, and the staff of the ASP system may register or update data in the service level data holding unit 17 in response to a request from a customer.

  The response time recording unit 19 measures the response time of the data processing system 30 with respect to the processing request issued by each user of each customer, and records the measured value of the response time of the processing request for each customer.

  An example of the data structure of the actually measured response time data managed by the response time recording unit 19 is shown in FIG. As shown in the figure, for each customer, the response time recording unit 19 displays the number of processing requests currently being executed by the customer (“the number of sessions being executed”) and the response time for the current processing request of the customer. Are measured and stored in the memory.

  For example, when the processing request receiving unit 11 receives a processing request from a user belonging to a customer, the response time recording unit 19 assigns an ID (identification information) to the processing request and associates the processing request with the processing request ID. The ID of the user who issued the processing request and the reception time of the processing request are recorded. When the processing request scheduling unit 15 retrieves a processing request from the processing request holding unit 13 and assigns it to one of the data processing units 32, the response time recording unit 19 displays the “number of running sessions” of the customer to which the processing request issuing user belongs Is increased by one. What customer the user who issued the processing request belongs to may be determined by referring to a database of correspondence relationship between the customer ID and the user ID. In addition, upon receiving a notification that the execution of the processing request has been completed (including the ID of the processing request) from the data processing unit 32, the number of running sessions of the customer to which the user who issued the processing request belongs is decreased by one. . At this time, the time from the reception time of the processing request to the completion time, that is, the response time for the processing request is calculated, and the measured value of the response time of the customer is updated with the response time. When the response time for a single processing request is used as the “response time actual value”, the calculated response time may be overwritten on the value of the “response time actual value” of the customer. Also, when using the average response time, for example, the response time of processing requests that have been executed during a certain period in the past is classified and stored for each customer, When the response time of the processing request is calculated, the average response time may be updated in consideration of the newly calculated response time. In this way, each time the processing request in the processing request holding unit 13 is transferred to the data processing system 30, the response time recording unit 19 increments the “number of running sessions” of the customer who issued the processing request to perform data processing. Each time the processing request is executed in the unit 32, the “number of sessions in execution” of the customer who issued the processing request is decremented, and the “measured response time value” is updated. By such processing, the number of sessions being executed and the actual response time of each customer are managed so as to always represent the latest situation.

  The processing request schedule unit 15 refers to the information on the number of sessions being executed and the measured response time of each customer managed by the response time recording unit 19 and the service level data of each customer held in the service level data holding unit 17. Then, control of fetching the processing request from the processing request holding unit 13 is performed.

  As described above, this take-out control has at least two operation modes of the normal mode and the guarantee mode. The processing request schedule unit 15 switches from the normal mode to the guarantee mode when the data processing system 30 is in a high load state. In one example, the data processing system 30 when the target response time of at least one of the customers within the warranty window is not achieved (or is determined to be likely not to be achieved in the near future). Is determined to be in a high load state. Whether or not the target response time of a certain customer has been achieved may be determined by comparing the target response time of the customer with the measured response time. If the measured response time is longer than the target response time, the target response time is not achieved, and if the measured response time is less than or equal to the target response time, the target response time is achieved. If the measured response time is longer than the threshold time set slightly shorter than the target response time, it is determined that there is a high possibility that the target response time will not be achieved soon, and it is determined that the load is high. May be. In addition, if a history of actual response time is recorded and it is determined from the history that the actual response time exceeds the target response time, the target response time may not be achieved. It may be determined that the load is high and the load may be determined to be high. In the following, in order to simplify the description, the state that the target response time is likely not to be achieved in the near future is also comprehensively expressed as “the target response time is not achieved”. To do. Similarly, a state where there is a low possibility that the target response time cannot be achieved in the near future will be comprehensively expressed by a description “target response time is achieved”.

  When it is determined that the load is high, the processing request scheduling unit 15 holds information such as the guaranteed time zone and target response time of each customer held in the service level data holding unit 17 and the response time recording unit 19. It is determined whether or not there is a high load state with reference to information such as measured response time of each customer. When it is determined that the load is high and the process is shifted to the guarantee mode, the process request schedule unit 15 sends a process request of a customer within the guarantee time zone (hereinafter referred to as “customer within the guarantee time period”) Hereinafter, the processing request is taken out from the processing request holding unit 13 in preference to the processing request of “customer outside warranty time” and is processed by the data processing system 30. In the guarantee mode, the processing request holding unit gives priority to the processing request of the customer whose target response time is not achieved among the customers within the same warranty time, over the processing request of the customer who has achieved the target response time. You may make it take out from 13. Moreover, in the guarantee mode, among the customers within the same guarantee time, the processing request of the customer whose number of running sessions (see FIG. 4) is less than the maximum number of simultaneous sessions is given priority over the processing request of the other customer. You may make it take out from the request | requirement holding | maintenance part 13. FIG.

  The priority relationship of processing request execution in the guarantee mode is summarized as follows.

  (1) The highest priority in the guarantee mode is a customer within the guarantee time when the target response time is not achieved and the number of running sessions does not reach the maximum number of simultaneous sessions. By preferentially processing such customer processing requests, the target response time is easily achieved.

  (2) The next priority is the customer within the guaranteed time when the number of running sessions has reached the maximum number of simultaneous sessions but the target response time has not been achieved. Even if processing requests exceeding the maximum number of simultaneous sessions are currently being processed, waiting for such customer processing requests will increase the measured response time and achieve the target response time. This is because it becomes more difficult.

  (3) The next priority is the customer within the guaranteed time period in which the target response time has been achieved but the number of running sessions has not reached the maximum number of simultaneous sessions. Since the target response time has been achieved for such a customer, there are fewer problems than the cases (1) and (2) above even if the execution of the processing request is delayed. Therefore, the priority is lower than the cases (1) and (2). However, since simultaneous execution up to the maximum number of simultaneous sessions is guaranteed, it is necessary to execute processing requests with a certain degree of priority so that the measured response time does not exceed the target response time. .

  (4) The next priority is the customer within the guaranteed time in which the number of running sessions has reached the maximum number of simultaneous sessions and the target response time has been achieved. Since the target response time has been achieved for such a customer, there are fewer problems than the cases (1) and (2) above even if the execution of the processing request is delayed. In addition, since processing requests exceeding the maximum number of simultaneous sessions are currently being processed, there is no need to further increase the number of simultaneous executions. Therefore, this case has a lower priority than the case (3).

  (5) The customer with the lowest priority is a customer outside the guaranteed time. Since the customer outside the guarantee time is out of the time zone for guaranteeing the response, there is no need to guarantee the response. Therefore, in the guarantee mode, the processing request of the customer outside the guarantee time is not newly taken out from the processing request holding unit 13 (and is executed by the data processing system 30). This allows the limited processing capability of the data processing system 30 to be directed to customers with higher priorities.

  As mentioned above, although the priority of cases (1) to (5) has been illustrated, it is not always necessary to distinguish all of these five stages when implemented in a system. For example, cases (1) to (4) may be treated the same (example in FIG. 6 to be described later), and handling of a plurality of cases having adjacent priorities may be set to the same level.

  Further, when there are a plurality of customers having the same priority (handling level), the ratio of the number of processing requests taken out from the processing request holding unit 13 is the ratio of the maximum number of simultaneous sessions of these customers. The processing request extraction from the processing request holding unit 13 may be controlled so as to be equal to.

  The control content of the processing request schedule unit 15 has been outlined above. Next, a specific example of the control will be described with reference to FIG.

  First, at the time of starting the system, the operation mode of the processing request scheduling unit 15 is set to “normal mode”. The process request schedule unit 15 periodically repeats the process of FIG. 5 in order to dynamically switch its operation mode in accordance with the latest state of the data processing system 30. In the procedure of FIG. 5, the process request schedule unit 15 compares the current time with the guarantee time zone of each customer stored in the service level data holding unit 17, and the customer within the guarantee time (the current time within its own guarantee time zone). The customer including the time is specified (S12). If there is no customer within the guarantee time in S12, the operation mode is set to the normal mode (S14), and the process returns to S12.

  If at least one customer within the guaranteed time is found in S12, the response time measured value of each customer within the guaranteed time is acquired from the response time recording unit 19 (S16), and the acquired response time measured value is stored in the service level data holding unit. By comparing with the target response time of the customer in 17, it is determined whether or not there is a customer within the guaranteed time for which the target response time has not been achieved (S 18). If there is no customer within the guaranteed time period for which the target response time has not been achieved, in this example, the data processing system 30 determines that it is not in a high load state, sets the operation mode to the normal mode (S14), and returns to S12. When it is determined in S18 that there is a customer within the guaranteed time for which the target response time has not been achieved, the process request scheduling unit 15 sets the operation mode to the guaranteed mode (S20), and returns to S12.

  Each data processing unit 32 in the data processing system 30 requests the processing request scheduling unit 15 to allocate a new processing request every time processing of a new processing request becomes possible. When a new processing request is requested from the data processing unit 32, the processing request scheduling unit 15 extracts the processing request from the processing request scheduling unit 15 and supplies the processing request to the requesting data processing unit 32.

  Here, when the operation mode is the normal mode, the processing request scheduling unit 15 takes out the first processing request in the order according to the order in the processing request holding unit 13 (queue), and requests the data processing unit 32 of the request source. To supply.

  On the other hand, FIG. 6 shows an example of the processing procedure of the processing request scheduling unit 15 when the operation mode is the guarantee mode.

  In this procedure, the processing request scheduling unit 15 waits for an allocation request from each data processing unit 32 in the data processing system 30 (S22), and when the allocation request arrives, sets the control variable n (n is an integer of 1 or more). It is initialized to 1 (S24). Then, the customer who issued the nth processing request from the head of the queue (processing request holding unit 13) is identified, and the service level data holding unit 17 is referenced to determine whether the customer is a customer within guaranteed time (S26). If the customer who issued the n-th processing request is not a customer within the guaranteed time, the processing request is not taken out from the processing request holding unit 13, and n is incremented by 1 (S28) and the process returns to S26, and the next in the queue Check the processing request.

  If it is determined in S26 that the nth processing request in the queue is for the customer within the guaranteed time, the processing request scheduling unit 15 extracts the nth processing request from the processing request holding unit 13 and issues an allocation request. The original data processing unit 32 is assigned (S30), and the process returns to S22.

  In the processing procedure of FIG. 6, the processing request of the customer within the guaranteed time that is found first when viewed from the head of the queue (corresponding to any of the cases (1) to (4) described above) is taken out from the processing request holding unit 13 and processed. Will be. On the contrary, the processing request of the customer outside the guarantee time (corresponding to the case (5) described above) remains in the processing request holding unit 13. According to such processing, during the guarantee mode, new processing requests for customers outside the warranty period are not executed, so processing of each processing request for the customer currently outside the warranty period is completed as time passes. As the process proceeds, the remaining capacity of the data processing system 30 increases. Then, the increased surplus capacity will be directed to new processing requests of customers within the guaranteed time. As a result, the execution of the processing request of the customer within the guarantee time is accelerated, and the target response time is achieved.

  In the processing procedure of FIG. 6, when there are a plurality of customers within the guarantee time, in the high load state, the processing requests of the customers within the guarantee time are simply processed according to the order in the queue. On the other hand, as another example, when there are a plurality of customers within the guaranteed time period during the period when the data processing system 30 is in a high load state, the rate at which the processing requests of each customer within the guaranteed time period are taken from the queue You may control so that it may become a thing according to the ratio of the maximum number of simultaneous sessions allocated to the customer within guarantee time. For example, thinning out of processing requests from queues of customers within each guaranteed time period may be performed so that the ratio of the number to be taken out from the queue is equal to the ratio between the maximum number of simultaneous sessions (for example, For example, the processing request of the customer within the guaranteed time period in which the ratio of the actual number of fetches is higher than the ratio of the number of fetches is set to a higher decimation rate from the queue).

  Next, an example of another processing procedure of the processing request schedule unit 15 in the guarantee mode will be described with reference to FIG.

  In this procedure, the processing request scheduling unit 15 waits for an allocation request from each data processing unit 32 in the data processing system 30 (S22), and when the allocation request arrives, sets the control variable n (n is an integer of 1 or more). It is initialized to 1 (S24). Then, the customer who issued the nth processing request from the head of the queue (processing request holding unit 13) is identified, and the service level data holding unit 17 is referenced to determine whether the customer is a customer within guaranteed time (S26). If the customer who issued the n-th processing request is not a customer within the guaranteed time, the processing request is not taken out from the processing request holding unit 13, and n is incremented by 1 (S28) and the process returns to S26, and the next in the queue Check the processing request.

  If it is determined in S26 that the n-th processing request in the queue belongs to the customer within the guaranteed time, the processing request scheduling unit 15 refers to the response time recording unit 19 and the service level data holding unit 17, and It is determined whether or not the target response time of the customer within the guarantee time has been achieved (S32). If not achieved, the n-th processing request is taken out from the processing request holding unit 13 and assigned to the data processing unit 32 of the allocation request issuing source (S30), and the process returns to S22.

  If it is determined in S32 that the target response time of the customer within the guaranteed time has been achieved, the processing request scheduling unit 15 further determines whether the number of running sessions of the customer within the guaranteed time has reached the maximum number of simultaneous sessions. Determine (S34). If it has reached, the n-th processing request is not taken out from the processing request holding unit 13, n is incremented by 1 (S28), and the process returns to S26. If not, the n-th processing request is taken out from the processing request holding unit 13 and assigned to the data processing unit 32 of the allocation request issuing source (S30), and the process returns to S22.

  In the processing procedure of FIG. 7, a customer processing request corresponding to the above-described cases (1) to (3), which is found first when viewed from the head of the queue, is taken out from the processing request holding unit 13 and processed. On the contrary, the customer's processing request corresponding to the above-described case (4) or (5) remains in the processing request holding unit 13. That is, in this procedure, the load on the data processing system 30 is reduced by waiting for the customer's processing request corresponding to the case (4) or (5), and the cases (1) to ( The processing request of the customer corresponding to 3) is executed quickly.

  In the example of FIGS. 6 and 7 described above, a delimiter as to whether or not to wait for a processing request in the processing request holding unit 13 is placed between cases (5) and (4) in the procedure of FIG. The procedure of 7 was placed between cases (4) and (3). However, these are merely examples. For example, as another example, the separation may be set between cases (3) and (2).

  In the above example, the control in the guarantee mode is a two-stage process in which non-prioritized customer processing requests are waited in the processing request holding unit 13 and the prioritized customer processing requests are executed in first-in first-out order. But this is just an example. As another example, for example, control may be performed so that processing requests are fetched from the processing request holding unit 13 at a relatively high frequency as the priority is higher. In this example, for example, the processing request for case (5) (customer out of warranty time) is completely waited in the queue (processing request holding unit 13), and the customers of cases (1) to (4) have a priority order. The higher the value, the easier the processing request is taken from the queue. For example, in the process of examining the queue from the top, every time a customer's processing request in case (1) is found, it is taken out from the queue, and every time a customer's processing request in case (2) is found twice, As the priority decreases, the rate of taking out from the queue is lowered, such as taking out from the queue.

  In addition, when there are multiple customers with warranty time of the same priority, the ratio of processing requests for each customer within the warranty time of the same rank from the queue is assigned to each customer within the warranty time. You may control so that it may become a thing according to the ratio of the maximum number of simultaneous sessions.

  Further, in the above example, a processing request that is currently being executed by a customer who is not prioritized, such as a customer outside the guarantee time, waits until execution is completed, but this is only an example. Instead of this, the following control may be performed.

  In other words, in this example, processing requests are classified into those that can be re-executed and those that cannot be re-executed, and in the guarantee mode, the processing requests that are currently being executed in the data processing system 30 are not preferred. Of these, re-executable processing requests are forcibly canceled (discarding the processing results up to the middle). As another example, a re-executable processing request may be forcibly interrupted (suspended by holding the processing result up to the middle). Processing requests that cannot be re-executed are not forcibly canceled or interrupted, but are waited for completion. The re-executable processing request whose execution has been canceled may be executed from the canceled processing request when the operation mode returns to the normal mode, for example, by returning to the head of the queue. The interrupted processing request is resumed using the processing result in the middle as soon as the operation mode returns to the normal mode. Note that the re-executable processing request includes, for example, the execution result at the first execution time and the execution time after the restart even if execution is canceled or interrupted and resumed later, such as browsing the contents of the database. There is a type of processing request that does not change. Non-re-executable processing requests include, for example, a request for database write processing.

  For this control, the process request schedule unit 15 only needs to hold classification information indicating whether each process request can be reexecuted.

  By canceling or interrupting the execution of a re-executable processing request of a customer that is not prioritized in this way, the remaining capacity of the data processing system 30 is increased more rapidly than when such cancellation or interruption is not performed, and the surplus power has priority. Will be directed to customers.

  In the above example, whether or not the data processing system 30 is in a high load state is determined based on the achievement status of each customer's target response time, but this is only an example. In addition to this, for example, when the sum of the number of running sessions of all customers exceeds a predetermined threshold, it may be determined that the load is high.

  Further, it may be determined whether or not a high load state is set based on a combination of the achievement status of the target response time of each customer and the total number of running sessions of all customers. In this case, for example, when the target response time of at least one customer is not achieved, it is determined that the load is high, and even if the target response time of all customers is achieved, the number of sessions being executed by all customers A determination condition such as determining a high load state may be used if the sum of the values exceeds a threshold value.

  Moreover, it is based on the measured value of the load condition of the hardware such as the usage rate of the CPU (central processing unit) of each data processing unit 32, not the achievement status of each customer's target response time and the number of running sessions of all customers. It may be determined whether or not the data processing system 30 is in a high load state.

  Some processing requests are completed in a very short time (that is, the load on the data processing system 30 is extremely low), such as a simple screen transition request. On the other hand, there are processing requests that require a certain amount of processing time (that is, processing load is high), such as processing involving a large amount of data processing or a large amount of data update. In general, processing requests from users are mixed with those having different loads. In such a situation, it may be inconvenient if a processing request with a very light processing load and a processing request that is not so are handled in the same row. For example, in the guarantee mode, even if a simple screen transition request of a customer outside the guarantee time is kept waiting in the queue, it does not contribute much to the load reduction of the data processing system 30, but the comfort of operation for the customer is It drops significantly. Further, when the number of running sessions is used for determining whether or not the data processing system 30 is in a high load state, if a processing request with little load such as a simple screen transition request is counted as one session, the count result Deviation from the actual load becomes severe.

  Therefore, processing requests may be classified into those that complete processing in a very short time and those that do not, and the former may not be counted as the number of running sessions. In this case, processing requests that can be processed in an extremely short time may be executed in the first-in first-out order without waiting in the queue even if the request is from a customer outside the guaranteed time.

  The processing request execution management apparatus 10 exemplified above may be realized by causing a general-purpose computer to execute a program representing the processing of each functional module described above, for example. Here, the computer includes, as hardware, a microprocessor such as a CPU, a memory (primary storage) such as a random access memory (RAM) and a read only memory (ROM), an HDD controller that controls an HDD (hard disk drive), Various I / O (input / output) interfaces, network interfaces that perform control for connection to a network such as a local area network, and the like have a circuit configuration connected via a bus, for example. Also, portable non-volatile recording of various standards such as a disk drive and a flash memory for reading and / or writing to a portable disk recording medium such as a CD or a DVD via the I / O interface, for example. A memory reader / writer for reading from and / or writing to a medium may be connected. A program in which the processing contents of each functional module exemplified above are described is stored in a fixed storage device such as a hard disk drive via a recording medium such as a CD or DVD, or via a communication means such as a network, and stored in a computer. Installed. The program stored in the fixed storage device is read into the RAM and executed by a microprocessor such as a CPU, thereby realizing the functional module group exemplified above.

  DESCRIPTION OF SYMBOLS 10 Process request execution management apparatus, 11 Process request reception part, 13 Process request holding part, 15 Process request schedule part, 17 Service level data holding part, 19 Response time recording part, 20 Service level registration part, 30 Data processing system, 32 Data processing unit, 34 processing data holding unit.

Claims (6)

For each of a plurality of customer organizations, storage means for storing a guarantee time zone that is a time zone for performing a response guarantee and the content of the response guarantee in the guarantee time zone , the content of the response guarantee includes: A storage means including at least the maximum number of simultaneous request processes assigned to the customer organization ;
A data processing system that executes data processing in response to a processing request from each user in the plurality of customer organizations, the total value of the maximum simultaneous request processing number of each customer organization at the same time, in all time zones A monitoring means for monitoring the load of a data processing system capable of simultaneously processing the maximum value among them ,
The period of high load state of the data processing system has been detected by the monitoring of the monitoring means, the processing request from the user guarantee time the set Onai a customer organization including the period in the warranty time period the it is performed on the data processing system, a control unit processing request for the period from the user of the guaranteed time zone comprise not a customer organization guarantees overtime in the organization to perform control so as not to execute the data processing system ,
A processing request execution management device comprising:   The control means includes a queue that temporarily holds a processing request from each user in each customer organization, and a period during which a high load state of the data processing system is not detected The processing request is taken out from the queue in order from the top of the queue without distinguishing which user from the organization outside the guarantee time, and the data processing system is subjected to high load. During the period in which the state is detected, processing requests from users within the organization outside the guarantee time are not taken out from the queue, but only processing requests from users within the organization within the guarantee time are taken out from the queue and sent to the data processing system. The processing request execution management apparatus according to claim 1, wherein the processing request execution management apparatus is executed. The control means, when there are a plurality of organizations within the guaranteed time period during the period when a high load state of the data processing system is detected, the ratio of processing requests for each organization within the guaranteed time period from the queue, Control to be in accordance with the ratio between the maximum simultaneous request processing number allocated to the organization within each guarantee time,
The processing request execution management apparatus according to claim 1 or 2, characterized in that Re-executability determination means for determining whether or not the processing request is re-executable;
The control means is the re-executability determination means among processing requests from users in the non-guaranteed organization that is being executed by the data processing system during a period when a high load state of the data processing system is detected. Instructing the data processing system to forcibly terminate execution of a processing request determined to be re-executable;
The processing request execution management apparatus according to any one of claims 1 to 3, wherein Request type determination means for determining whether or not the processing request is a light load type that requires only a predetermined time or less for execution by the data processing system;
The monitoring means monitors the number of processing requests determined to be not of light load type among the processing requests from the customer organization that are simultaneously executed by the data processing system for each customer organization as the number of simultaneous executions. Determining whether the data processing system is in a high load state based on the number of concurrent executions of each customer organization being monitored,
The processing request execution management apparatus according to claim 1, wherein the processing request execution management apparatus is a management request execution management apparatus. Computer
For each of a plurality of customer organizations, storage means for storing a guarantee time zone that is a time zone for performing a response guarantee and the content of the response guarantee in the guarantee time zone , the content of the response guarantee includes: Storage means including at least the maximum number of concurrent request processes assigned to the customer organization ;
A data processing system that executes data processing in response to a processing request from each user in the plurality of customer organizations, the total value of the maximum simultaneous request processing number of each customer organization at the same time, in all time zones Monitoring means to monitor the load of the data processing system capable of simultaneously processing the maximum value among them ,
The period of high load state of the data processing system has been detected by the monitoring of the monitoring means, the processing request from the user guarantee time the set Onai a customer organization including the period in the warranty time period the is performed on the data processing system, the control means processing request for the period from the user of the guaranteed time zone comprise not a customer organization guarantees overtime in the organization to perform control so as not to execute the data processing system,
Program to function as.

Images