JP4752767B2 - System configuration candidate derivation device, method and program - Google Patents

Description

  The present invention relates to a system configuration candidate derivation device, a system configuration candidate specifying method, and a system configuration candidate specifying program for specifying a configuration candidate after change when the configuration of a system to be managed is changed.

  An example of a conventional operation management system is described in Patent Document 1. The operation management system described in Patent Document 1 extracts a business pattern (a ratio of which business is being executed and how much) by monitoring a log of the management target system. Then, the performance prediction simulation unit calculates response time and throughput in the business pattern. If the predicted performance does not satisfy the required performance of the system, the configuration to be changed is specified according to the conditions for specifying the hardware configuration to be expanded or reduced. The above operation is repeated until the predicted performance calculated by the performance prediction simulation unit satisfies the required performance of the system. If the predicted performance satisfies the required performance, an optimum configuration is presented.

  FIG. 17 is an explanatory diagram illustrating an example of a system to be managed. The management target system illustrated in FIG. 17 includes a load balancer 240, a Web layer 210, an AP layer 220, and a DB layer 230. A layer is a set of computers that are loaded with the same kind of software and operate according to the software. The web layer 210 is a set of web servers 211 loaded with a program that causes a computer to execute processing for providing a web page to a terminal in response to a request from a terminal (not shown). The AP layer 220 is a set of servers (hereinafter referred to as AP (application) servers) 221 loaded with a program that causes a computer to execute processing in response to a request from a Web server. The DB layer 230 is a set of servers (hereinafter referred to as DB (database) servers) 231 loaded with a program that causes a computer to execute processing for extracting data from a database (not shown).

  The web server 211 transmits a web page to the terminal in response to a request from the terminal (not shown). The AP server 221 executes various processes according to requests from the Web server. The DB server 231 reads data from the database.

  The web server 211 is connected to the first communication network 212. The AP server 221 is connected to the second communication network 222. The DB server 231 is connected to the third communication network 232. The first communication network 212 and the second communication network 222 are connected by a router 251, and the second communication network 222 and the third communication network 232 are connected by a router 252. The load balancer 240 distributes requests from the terminal (not shown) to the Web server 211 so that the load on each Web server 211 is distributed.

  When any web server 211 receives a request from a terminal, the web server 211 causes any AP server 221 to execute processing according to the request. The AP server 221 causes the DB server 231 to extract data necessary for processing from a database (not shown), and proceeds with the processing using the data. The Web server 211 transmits a Web page reflecting the processing result by the AP server 221 to the terminal.

  When monitoring such a system, a monitoring system 170 that monitors the load status of each server included in the system is provided. The monitoring system 170 is connected to each communication network 212, 222, 232, for example, and monitors the load status of each server. The load status is, for example, the operation rate of the CPU of each server and the number of various commands such as requests received by each server.

Japanese Patent Laying-Open No. 2005-99973 (paragraphs 0022-0035, FIG. 1 and FIG. 2)

  The operation management system described in Patent Document 1 predicts the performance according to the business pattern, compares the predicted performance with the required performance required for the system to be managed, and the predicted performance satisfies the required performance. Until it is satisfied, the configuration to be changed is identified, and the performance prediction and comparison with the required performance are repeated. However, depending on the conditions for specifying the configuration to be expanded or reduced, it may not be possible to specify a configuration that satisfies the required performance. For example, when a certain configuration is specified and the system configuration is switched, a new problem may appear even if an event that has been a problem before the switching is solved. Specifically, even if the bottleneck before switching can be eliminated, another component may become a bottleneck by switching the configuration. In addition, when a system configuration is switched, a combination of components included in the system after switching the configuration affects one component, and the desired required performance cannot be obtained. There is also a possibility. As a result, there may be a case where a configuration that satisfies the required performance cannot be specified even if the process of calculating the predicted performance and comparing with the required performance is repeated.

  If the components belonging to the system are not balanced, the performance of the entire system will be affected. For example, if the number of servers in some layers of the system illustrated in FIG. 17 is small or performance deficiencies occur in some layers, a satisfactory response cannot be obtained as a whole system. In addition, if there are excessive servers in a certain layer, some of the servers may not contribute to the performance of the entire system. Such a state means that useless costs have occurred.

  As a result of the monitoring system 170 monitoring the load status of each server, it is clarified that the number of servers in any layer is insufficient or excessive. In this case, the system administrator may add a server to the layer or delete the server. However, each time an excess or deficiency of a certain component is recognized, the number of each component is appropriate after changing the component of the system simply by increasing or decreasing the component. Not necessarily.

  Also, a system configuration candidate derivation device and system configuration candidate identification that can derive a combination of system components included in a managed system and that does not cause excessive or small number of system components in some layers It is an object to provide a method and a system configuration candidate specifying program.

The system configuration candidate derivation apparatus according to the present invention is a system configuration candidate for deriving a system configuration candidate that is a candidate for the configuration of a managed system including a plurality of layers that are sets of system components that are loaded with the same kind of software and operate according to the software. An derivation device that measures a metric of each layer indicating a load status of a managed system as a value, and derives a correlation between values of metrics in a set of metrics measured from different layers ; System configuration candidates for deriving system configuration candidates based on performance profiles that are information that can derive the relationship between the number of system components and the processing amount of the system components, and the correlation of values between metrics And enumeration means.

  According to such a configuration, the system configuration candidate enumeration means includes the performance profile, which is information capable of deriving the relationship between the number of system components and the processing amount of the system components, and the metric correlation. Based on this, system configuration candidates are derived, and therefore system configuration candidates that do not have an excessively large or small number of system components in some layers can be derived.

  The system configuration candidate derivation device of the present invention includes, for example, performance profile means for providing a performance profile.

In addition, the inter-element correlation deriving means continuously measures the metric as time passes, calculates the coefficient by substituting the metric at each time into an expression including multiplication of the variable and the coefficient, using the metric as a variable. If the calculated variance of the coefficients is equal to or less than a threshold value, the correlation between the values of the metrics may be derived based on the coefficients.

The system configuration candidate enumeration means obtains the processing amount in the number of units in the reference layer based on the performance profile for each number of system components in one reference layer, and the processing amount and the value between metrics The system configuration candidate may be derived by calculating the processing amount in the other layer based on the correlation between the two, and deriving the number of units capable of processing the processing amount in the other layer from the performance profile.

The system configuration candidate enumeration means obtains the throughput ratio of each layer from the correlation of values between metrics, sets the number of system components in each layer to the initial value, and obtains the throughput ratio of each layer from the performance profile. Identify the system component of the layer corresponding to the term with the smallest division result by dividing each term of the obtained throughput ratio by each corresponding term of the throughput ratio obtained from the correlation of values between metrics. The configuration may be such that system configuration candidates are derived by repeating the addition of the number of system component elements in the layer.

Further, the system configuration candidate derivation program of the present invention is a computer for deriving a system configuration candidate that is a candidate for configuration of a managed system including a plurality of layers that are a set of system components that are loaded with the same kind of software and operate according to the software. Is a system configuration candidate derivation program installed in, which measures the metric of each layer that shows the load status of the managed system as a value on the computer, and correlates the value between the metrics in the metric set measured from different layers Based on the correlation between the metrics and the performance profile that is the information that can derive the relationship between the number of system components to be derived and the relationship between the number of system components and the amount of processing of the system components Execute system configuration candidate enumeration process for deriving system configuration candidates And wherein the door.

In the correlation derivation process between elements, the metric is continuously measured over time, and the coefficient at each time is calculated by substituting the metric at each time into a formula that includes the multiplication of the variable and the coefficient. If the calculated coefficient variance is less than or equal to a threshold value, a program that executes a process for deriving the correlation of values between metrics based on the coefficient may be used.

In the system configuration candidate enumeration process, the amount of processing in the number of system components in one reference layer is obtained from the computer based on the performance profile, and the processing amount and metric are calculated. Based on the correlation of the values between them, the amount of processing in other layers is calculated, and the number of units that can process the amount of processing in other layers is derived from the performance profile, and the system configuration candidate is derived. It may be a program.

In the system configuration candidate enumeration process, calculate the throughput ratio of each layer from the correlation of values between metrics, determine the number of system components of each layer as the initial value, determine the throughput ratio of each layer from the performance profile, Divide each term of the throughput ratio obtained from the performance profile by each corresponding term of the throughput ratio obtained from the correlation of values between metrics, and identify the system component of the layer corresponding to the term with the smallest division result Then, the program may be a program that executes a process of deriving system configuration candidates by repeating the addition of the number of system components of the layer.

The system configuration candidate derivation method of the present invention is a system for deriving a system configuration candidate that is a candidate for configuration of a managed system including a plurality of layers that are a set of system components that are loaded with the same kind of software and operate according to the software. A configuration candidate derivation method in which inter-element correlation deriving means measures a metric of each layer indicating the load status of a managed system as a value, and correlates a value between metrics in a set of metrics measured from different layers. The system configuration candidate enumeration means is derived based on the performance profile, which is information that can derive the relationship between the number of system components and the amount of processing of the system components, and the correlation of values between metrics. The system configuration candidate is derived.

  According to the present invention, the system configuration candidate listing means is based on the performance profile, which is information that can derive the relationship between the number of system components and the processing amount of the system components, and the metric correlation. Since the system configuration candidate is derived, it is possible to derive a system configuration candidate that does not have an excessively large or small number of system component elements in some layers.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

Embodiment 1 FIG.
FIG. 1 is a block diagram illustrating an example of a system configuration candidate derivation device according to the first embodiment of this invention. The system configuration candidate derivation device of the present invention includes a data processing device 100 and system operation storage means 140.

  The data processing apparatus 100 is connected to a management target system (hereinafter referred to as a target system) 160 that is a management target. FIG. 2 is an explanatory diagram illustrating an example of the target system 160 connected to the data processing apparatus 100. In FIG. 2, servers, communication networks, routers, load balancers, and layers constituting the target system are denoted by the same reference numerals as those in FIG. 17, and detailed descriptions of these servers and other communication networks are omitted. To do. However, since the layers belonging to the target system 160 are not limited to the Web layer, the AP layer, and the DB layer, the layers belonging to the target system 160 are hereinafter referred to as a first layer, a second layer, and a third layer. Although FIG. 2 illustrates the case where the target system 160 includes these three layers, the number of layers included in the target system 160 is not limited to three.

  As already described, a layer is a set of computers that are loaded with the same kind of software and operate according to the software. The same software may be installed in devices belonging to the same layer. Hereinafter, a server corresponding to a computer constituting a layer (a computer in which the same kind of software is installed and operates according to the software) is referred to as a system component. An apparatus or software that is an operation target for adding or excluding system components (servers) is referred to as an operation target element. Examples of the operation target element include a load balancer 240 and routers 251 and 252 included in the target system 160. Each server 211, 221, 231 also corresponds to the operation target element. In addition, software (for example, middleware, application software, etc.) installed in each of the servers 211, 211, 231, the load balancer 240, and the routers 251, 252 also corresponds to the operation target element.

  An operation for adding or excluding a system component (server) is a set of operations for the operation target element. For example, an operation of adding a server to a certain layer is “setting change so that a server that is physically connected to the communication network but is not communicable via the communication network can be communicated”. “Install the software on the server”, “Start the software”, “Change the server settings to disable communication”, “Exit the software”, “Delete the software from the server. "Is represented by a set of operations for the operation target element. The system operation storage unit 140 is a storage device that stores operation information for an operation target element when a server is added and when a server is excluded for each layer. The operation information stored in the system operation storage unit 140 is described in a format such as a shell script in UNIX (registered trademark) or a workflow language such as WS-BPEL (Web Services Business Process Execution Language). However, the shell script and WS-BPEL shown here are examples, and the operation information stored in the system operation storage unit 140 may be described in other formats.

  The data processing apparatus 100 monitors (measures) the load status of the target system 160 and generates a system configuration change rule based on the correlation of the measured items.

  The data processing apparatus 100 includes an inter-element correlation deriving unit 110, a performance profile unit 120, a system configuration candidate listing unit 131, an effective range calculating unit 132, a trigger generating unit 133, an operation sequence generating unit 134, a rule Generating means 130.

  The inter-element correlation deriving unit 110 measures a metric of a server that is a system component of the target system 160, specifies a combination of metrics having correlation, and derives a correlation between the metrics.

  A metric is a measurement item that represents the load status of a managed system. Examples of metrics include the CPU usage rate and memory usage of each server belonging to each layer 210, 220, 230, and the total number of requests received by each server belonging to the same layer per unit time. These are examples of metrics, and the metrics measured by the inter-element correlation deriving unit 110 are not limited to the above-described metrics.

  Correlation is a relationship between metrics in which when a value of a certain metric changes, the value of another metric also changes with the change. For example, it is assumed that the number of requests that each server 211 of the first layer 210 receives from the load balancer 240 per unit time has increased. Then, it is assumed that each server 211 requests processing from the server 221 of the second layer 220 in response to the request. Then, the number of requests per unit time that the server 221 in the second layer 220 receives from each server 211 in the first layer also increases. Therefore, the total number of requests received per unit time by each server 211 of the first layer 210 from the load balancer 240 and the requests received by the server 221 of the second layer 220 from each server 211 of the first layer per unit time. There is a correlation between the total number. The inter-element correlation deriving means 110 identifies a combination of metrics having these correlations, and derives the correlation between the metrics.

  The performance profile unit 120 provides the performance profile to the system configuration candidate listing unit 131 and the effective range calculation unit 132.

  The performance profile is information that can derive the relationship between the number of system components (servers) and the amount of processing of the system components of that number. The performance profile may be information that directly indicates the relationship between the number of servers that are system components and the processing amount of the servers. Further, the performance profile may be information indicating the processing amount per unit (one) of servers. By multiplying the processing amount per server by the number of servers, the relationship between the number of servers and the processing amount of the servers of that number can be derived.

  FIG. 3 is an explanatory diagram showing an example of a performance profile that directly shows the relationship between the number of servers and the processing amount of the servers. When directly showing the relationship between the number of servers and the processing amount of the servers, information that summarizes the relationship between the number of servers and the processing amount of the server is obtained for each metric as shown in FIG. Become. For example, it is assumed that the metric 1 illustrated in FIG. 3 is the total number of requests that the server 211 of the first layer 210 receives from the load balancer 240 per unit time. In the example shown in FIG. 3, the number of servers and the amount of processing, such as processing up to 70 requests per unit time with one server 211 and processing up to 140 requests per unit time with two servers 211, Showing the relationship. Similarly, for example, it is assumed that the metric 2 illustrated in FIG. 3 is the total number of transactions performed by each server 221 of the second layer 220 per unit time. In the example shown in FIG. 3, the number of servers and the amount of processing are such that one server 221 can execute 25 transactions per unit time, and two servers 221 can execute 50 transactions per unit time. Showing the relationship. In the example shown in FIG. 3, the number of transactions per unit time does not change even when the number of servers is changed from 4 to 5, but this is the layer to which the server belongs even if the number of servers is increased by 4 or more. This means that the overall throughput (performance) cannot be increased. In other words, when the number of servers is increased to four, the processing amount of the entire layer to which the servers belong is saturated.

  FIG. 4 is an explanatory diagram showing an example of a performance profile showing the processing amount per unit number of servers. When the processing amount per unit number of servers is shown, the metric, the processing amount per unit number, and the limit number are associated as illustrated in FIG. As in the case illustrated in FIG. 3, in FIG. 4 as well, the metric 1 is the total number of requests that the server 211 of the first layer 210 receives from the load balancer 240 per unit time. The example illustrated in FIG. 4 indicates that up to 70 requests can be processed per unit time (unit) of the server 211. Further, the limit number means the number of servers in which the processing amount of the entire layer to which the server belongs is saturated. Therefore, in this example, if the number of servers 211 is increased to 10, the processing amount can be increased to 70 · 10 = 700. However, even if the number of servers is increased, the processing amount cannot be increased any further.

  3 and 4 exemplify performance profiles related to two types of metrics, the performance profile is not limited to only those related to two types of metrics.

  In addition, for example, when the number of requests, the number of transactions, the number of queries, or the like is indicated as the processing performance, it may be indicated as the number that can be processed per unit time.

  The system configuration candidate listing unit 131 derives system configuration candidates. The system configuration candidate is a configuration candidate for the target system. The target system 16 includes system components (servers) for each layer, but if the number of servers in each layer is an arbitrary number, the system configuration is not balanced. For example, if the processing performance realized by the number of servers belonging to one of the tiers is lower than the processing performance realized by the number of servers belonging to other tiers, the layer with the lower processing performance becomes a bottleneck. Therefore, the overall system configuration is not balanced. For example, when the processing performance realized by the number of servers belonging to any layer is higher than the processing performance realized by the number of servers belonging to other layers, the number of servers in the layer with high processing performance is There are too many servers that do not contribute to processing performance. Even in this case, the system configuration is not balanced. A balanced system configuration is a configuration in which there is no server that does not contribute to processing performance in any layer.

  There may be an upper limit on the number of servers that can be included in one layer. For example, the number of servers in one layer may be limited due to physical reasons such as the number of ports of network devices such as routers being limited or the number of servers owned by the administrator being limited. Also, the server in one layer may be limited by the operational limitations of the software purchased for installation on the server (for example, the number and configuration of software that can be operated or the number of purchased licenses is limited). The number may be limited. In each of the layers 210, 220, and 230 shown in FIG. 2, it is assumed that the number of servers that can be installed is ten. Then, there are 10 × 10 × 10 = 1000 combinations of the number of servers installed in each layer, but not all combinations have a balanced configuration. For example, in a configuration in which there are 10 servers 211 on the first layer 210 and one server on each of the other layers, the number of servers on the first layer 210 is excessive, and the servers should be used effectively. Since it does not become, it cannot be said that it is a balanced composition.

  The system configuration candidate listing unit 131 derives a balanced system configuration candidate based on the metric correlation and the performance profile. The number of system configuration candidates derived by the system configuration candidate enumeration unit 131 is not limited to one, and the system configuration candidate enumeration unit 131 enumerates balanced system configuration candidates.

  The data processing apparatus 100 may include a storage device (not shown) that stores a predetermined upper limit value of the number of servers for each layer. In addition, this storage device may store a lower limit value of the number of servers for each predetermined layer. Hereinafter, this storage device (not shown) is referred to as the number limit value storage means.

  The effective range calculation unit 132 calculates the range of loads that can be processed in each layer for each system configuration candidate based on the system configuration candidates derived by the system configuration candidate listing unit 131, the performance profile, and the correlation of metrics. To do.

  The range of load that can be processed in each layer is not determined only by the processing performance and the number of servers in that layer, but also depends on the correlation of metrics. For example, in a system configuration candidate, it is assumed that the number of servers in the first layer 210, the second layer 220, and the third layer 230 is set to one each. Further, it is assumed that the processing numbers of the servers 211, 211, and 231 of each layer are 70, 50, and 25, respectively. In order for the first layer server 211 to execute one process, the second layer server 221 needs to execute two processes, and the third layer server 231 has to execute one process. The first layer server 211 can perform 70 processing by one unit. To that end, the second layer server 221 must execute 140 processing and the third layer server 231 must perform 70 processing. In the example, the second layer server 221 and the third layer server 231 execute only 50 processes and 25 processes, respectively. Therefore, “in order for the first layer server 211 to execute one process, the second layer server 221 must execute two processes, and the third layer server 231 must execute one process. Therefore, the range of loads that can be processed by each server in each layer from the first layer to the third layer (hereinafter referred to as an effective range) is 25 processing, 50 processing, and 25 processing, respectively.

  The effective range calculating unit 132 calculates such effective range combinations for each system configuration candidate derived by the system configuration candidate listing unit 131.

  It is assumed that a system configuration candidate has a configuration of “one first-layer server 211, one second-layer server 221 and one third-layer server 231”. The effective range in this system configuration candidate is calculated as “0 to 25 processes in the first layer server 211. 0 to 50 processes in the second layer server 221. 0 to 25 processes in the third layer server 231”. To do. Further, it is assumed that a certain system configuration candidate has a configuration of “one first-layer server 211, two second-layer servers 221 and two third-layer servers 231”. The effective range in this system configuration candidate is calculated as “0 to 50 processes in the first layer server 211. 0 to 100 processes in the second layer server 221. 0 to 50 processes in the third layer server 231”. To do.

  As described above, when switching from one system configuration candidate to another system configuration candidate, the effective range also changes. It is assumed that the system configuration candidates derived by the system configuration candidate listing unit 131 are arranged in ascending order so that the number of servers in each layer increases, and the effective ranges are arranged in that order. At that time, in the effective ranges arranged in order, the upper limit of the effective range arranged first is called a boundary of the effective range. The illustrated effective range of “0 to 25 processes for the first layer server 211. 0 to 50 processes for the second layer server 221. 0 to 25 processes for the third layer server 231.” With respect to the effective range of 0 to 50 processing at the server 211, 0 to 100 processing at the second layer server 221, and 0 to 50 processing at the third layer server 231, the boundary of the effective range is “in the first layer. 25, 50 for the second layer, 25 for the third layer. "

  Here, a system configuration change rule generated by the system configuration candidate derivation device of the present invention will be described. The system configuration change rule is an if-then format rule, and includes conditions regarding the load on the target system 16 and operation information regarding the target system when the conditions are satisfied. Hereinafter, the condition part of the system configuration change rule is referred to as a trigger. Further, when the trigger is satisfied, an operation for adding or excluding a server is performed. This operation is a set of operations for the operation target element. Furthermore, operations for adding or excluding servers are not necessarily performed for only one layer, but may be performed for a plurality of layers. A set of operations for the operation target element for each layer to be operated is referred to as an operation sequence. That is, the system configuration change rule includes a trigger and an operation sequence, and represents that each operation included in the operation sequence should be executed when the trigger is satisfied.

  The trigger generation unit 133 generates a trigger based on the boundary of the effective range calculated for each system configuration candidate. For example, the effective range of the first layer 210 is “0 to 25” for a certain system configuration candidate, “0 to 50” for another system configuration candidate, and the boundary of the effective range is “25”. . The trigger generation unit 133 obtains a value that is a predetermined number lower than the boundary or a predetermined ratio (for example, 90%) of the boundary value, and uses the value as a trigger. That is, the trigger is calculated as a value lower than the boundary value. When the load in the target system 16 is increased by obtaining the trigger in this way, the system configuration of the target system 16 can not withstand the load before switching to a system configuration candidate with a larger number of servers. It is possible to generate a system configuration change rule that instructs Further, when the load in the target system 16 has occurred, when the load is reduced to a load that can be processed by the configuration of the system configuration candidate after switching, an instruction to switch to a system configuration candidate with a smaller number of servers is given. A system configuration change rule can be generated.

  When the system configuration candidates derived by the system configuration candidate listing unit 131 are arranged in ascending order so that the number of servers in each layer increases, the operation sequence generation unit 134 obtains a difference between the system configuration candidates arranged in order. . That is, the system configuration elements (servers) included in only one system configuration candidate among the system configuration candidates arranged in order and adjacent to each other are obtained. The operation sequence generation unit 134 reads, from the system operation storage unit 140, operation information for adding a difference server to the layer. A set of information read from the system operation storage unit 140 as an operation information for adding a difference server to the layer is an operation sequence. Further, the operation sequence generation unit 134 also reads, from the system operation storage unit 140, operation information for excluding the server that becomes the difference from the layer. A set of information read from the system operation storage unit 140 is also an operation sequence as operation information for excluding the difference server from the layer.

  The rule generation unit 130 generates a system change rule by combining the trigger generated by the trigger generation unit 133 and the operation sequence read from the system operation storage unit 140 by the operation sequence generation unit 134.

  As shown in FIGS. 1 and 2, the data processing apparatus 100 (rule generation unit 130) may be connected to the monitoring system 170. Then, the rule generation unit 130 may transmit the generated system configuration change rule to the monitoring system 170. The monitoring system 170 monitors the load status of the target system 160. When the trigger is satisfied, the monitoring system 170 performs a server addition operation or an exclusion operation on the target system 160 according to an operation sequence corresponding to the trigger.

  In this example, the data processing apparatus 100 (rule generation unit 130) is connected to the monitoring system 170, and the rule generation unit 130 transmits the system configuration change rule to the monitoring system 170. The aspect provided to the monitoring system 170 may be another aspect. For example, the rule generation unit 130 may store the system configuration change rule in a storage medium, and the administrator may cause the monitoring system 170 to read the system configuration change rule from the storage medium.

  The inter-element correlation deriving unit 110, the performance profile unit 120, the system configuration candidate listing unit 131, the effective range calculation unit 132, the trigger generation unit 133, the operation sequence generation unit 134, and the rule generation unit 130 operate according to a program, for example. It is realized by a computer (central processing unit). Each of the above means may be realized by the same central processing unit.

  Next, the operation will be described. FIG. 5 is a flowchart illustrating an example of processing progress of the system configuration candidate derivation device according to the first embodiment.

  First, the inter-element correlation deriving unit 110 measures a metric of a server that is a system component of the target system 160 (step A1). Examples of metrics include the server CPU usage rate, memory usage, and the number of requests per unit time received by the server. The CPU of the server outputs a CPU usage rate and a memory usage amount in accordance with an OS (Operating System) installed in the server. For example, the inter-element correlation deriving unit 110 may request the CPU usage rate and the memory usage amount from the server and receive the information from the server. Further, the CPU of the server records the number of received requests and the like in a log according to the software installed in the server. For example, the inter-element correlation deriving unit 110 may request a log from the server, receive the log from the server, and obtain a metric such as the number of requests per unit time from the log. Further, the monitoring system 170 may collect the metrics as described above, and the inter-element correlation deriving unit 110 may receive the metrics from the monitoring system 170.

  The inter-element correlation deriving unit 110 does not measure only a metric at a certain time but continuously measures a metric as time passes. For example, the inter-element correlation deriving unit 110 periodically measures each metric.

  Next, the inter-element correlation deriving means 110 identifies a combination of metrics having correlation based on the measured metric, and derives the correlation between the metrics (step A2). Hereinafter, the operation of step A2 will be described in detail.

The inter-element correlation deriving unit 110 selects two metrics from the metrics measured in step A1, and determines whether the two metrics are correlated by regression analysis. If it is sufficient to determine whether or not there is a proportional relationship between the two metrics, a regression analysis may be performed using a linear expression y = a · x. This regression analysis may be performed as follows. The measurement results of the two metrics M _a and M _b at time t are M _a (t) and M _b (t), respectively. The inter-element correlation deriving unit 110 calculates the metrics M _a (t) and M _b (t) at each time (here, the time is represented as t = 0, 1, 2,... For convenience) by y. = M _a (t) and x = M _b (t) are substituted into y = a · x. Then, many equations are obtained as follows.

M _a (0) = a · M _b (0),
M _a (1) = a · M _b (1),
M _a (2) = a · M _b (2),
M _a (3) = a · M _b (3),
...

The inter-element correlation deriving unit 110 calculates the coefficient a of the linear expression from each of the above expressions. Since each value of M _a (0), M _b (0), etc. is measured in step A1, a in the above equations is calculated by, for example, division of M _a (0) / M _b (0). can do. Subsequently, the inter-element correlation deriving unit 110 calculates the variance of the coefficient a obtained from each equation, and determines whether the variance is equal to or less than a predetermined threshold. If the dispersion is below the threshold of a, element between correlation deriving means 110, two metrics M _a, is between M _b is correlative with the ratio of the throughput of M _a and M _b is 1: a Judge that there is. “A” in “1: a” obtained as the ratio of the throughput of M _a and M _b is the average value of a obtained from each equation.

For example, the metric M _a is the total number of requests received by each server of the first layer 210 (see FIG. 1) from the load balancer 240 per unit time, and the metric M _b is the first of the servers of the second layer 220. It is assumed that the total number of transactions executed per unit time in response to a request from a layer server. As described above, the ratio of the processing amount of M _a and M _b is 1: a. When the server of the first layer 210 receives one request, a number of transactions are received by the server of the second layer. It means to occur.

If the variance of a is larger than a predetermined threshold, it is determined that there is no correlation between the two metrics M _a and M _b , another metric combination is selected, and the same processing is performed. repeat. Note that the selection order of metric combinations may be determined in advance.

In the above example, the case where the presence / absence of correlation is determined by regression analysis using a simple linear expression y = a · x is exemplified, but the formula used for regression analysis (hereinafter referred to as regression formula) is as follows. An expression including a constant term b of y = a · x + b may be used. In this case, the inter-element correlation deriving unit 110 sets the metrics M _a (t) and M _b (t) at each time as y = M _a (t) and x = M _b (t), and y = a · x + b Assign to. Then, many equations are obtained as follows.

M _a (0) = a · M _b (0) + b,
M _a (1) = a · M _b (1) + b,
M _a (2) = a · M _b (2) + b,
M _a (3) = a · M _b (3) + b,
...

The inter-element correlation deriving means 110 obtains a and b from the above equations. The inter-element correlation deriving unit 110 may calculate a and b using, for example, a least square method. Further, a and b may be calculated as a solution of simultaneous equations consisting of a combination of two expressions. Next, the inter-element correlation deriving unit 110 calculates the calculated variance of each a and each b. The inter-element correlation deriving unit 110 determines whether or not the variances of a and b are less than or equal to a predetermined threshold, and if the variance of a and the variance of b are each less than or equal to the threshold, two metrics M It determines that a: _a, there is a correlation is between _{M b,} the ratio of the throughput of _{M a} and _{M b} is 1. “A” in “1: a” obtained as the ratio of the throughput of M _a and M _b is the average value of a obtained from each equation. If one or both of the variance of a and the variance of b exceed the threshold, it is determined that there is no correlation between M _a and M _b . Note that the threshold value used for comparison with the variance of a and the threshold value used for comparison with the variance of b may not be the same value.

Moreover, you may determine the presence or absence of a correlation using the regression formula which considered time delay as a regression formula. The regression equation considering the time delay is y (t) + a ₁ · y (t−1) + a ₂ · y (t−2) +... = B ₀ · x (t) + b ₁ · x (t− 1) It is expressed as + b ₂ · x (t−2) +. Here, y (t) + a ₁ · y (t−1) + a ₂ · y (t−2) = b ₀ · x (t) + b ₁ × x (t−1) + b ₂ × x (t−2) ) Will be described as an example. The inter-element correlation deriving unit 110 substitutes the metric at each time into the regression equation. Then, many equations are obtained as follows.

M _a (2) + a ₁ · M _a (1) + a ₂ · M _a (0) = b ₀ · M _b (2) + b ₁ · M _b (1) + b ₂ · M _b (0),
M _a (3) + a ₁ · M _a (2) + a ₂ · M _a (1) = b ₀ · M _b (3) + b ₁ · M _b (2) + b ₂ · M _b (1),
...

The inter-element correlation deriving unit 110 obtains coefficients a ₁ , a ₂ , and b _{0 to} b ₂ from the above equations. The inter-element correlation deriving unit 110 may calculate these coefficients using, for example, the least square method. Next, the inter-element correlation deriving unit 110 calculates variances of the calculated coefficients a ₁ , a ₂ , b _{0 to} b ₂ . The inter-element correlation deriving means 110 determines whether or not the variances of the coefficients a ₁ , a ₂ , b _{0 to} b ₂ are each equal to or less than a predetermined threshold, and if the variance of each coefficient is equal to or less than the threshold, respectively. For example, there is a correlation between the two metrics M _a and M _b , and it is determined that the ratio of the processing amount of M _a and M _b is 1: P. Here, P may be obtained as follows. If from the equation to determine the _{a 1, a 2, b 0} ~b 2, inter-component correlation deriving means 110 calculates the average value of _{a 1.} Similarly, average values are calculated for a ₂ , b ₀ , b ₁ , and b ₂ , respectively. The inter-element correlation deriving means 110 includes a ₁ + a ₂ = Σa _i which is the sum of the average values of the left side coefficients, and b ₀ + b ₁ + b ₂ = Σb _i which is the sum of the average values of the right side coefficients. And Σb _i / Σa _i = P, respectively, and the above processing amount ratio 1: P may be derived.

Each example Der h above, two metrics M _a, was determined whether the correlation is between M _b, further, for example, the metric M _b, to determine whether the correlation of M _c, correlated If it is determined that there _{is, M a,} M _b of the processing amount ratios and _M b, based on the ratio of throughput of _{M c, M a, M b} , to derive the amount of processing the ratio of _{M c} Also good.

  Further, the performance profile means 120 generates a performance profile (information that can derive the relationship between the number of servers and the processing amount of the number of servers) (step A3). The performance profile unit 120 passes the generated performance profile to the system configuration candidate listing unit 131 and the effective range calculation unit 132. The performance profile unit 120 may derive the performance profile by actually applying a load to the server of the target system 160. For example, the performance profile unit 120 acquires the configuration of the target system such as the number of servers operable in the target system 160. The performance profile means 120 sends a request to a server in a certain layer and increases the number of requests per unit time until the request cannot be processed. The number of requests may be determined. In addition, when it becomes impossible to process with one server, requests are transmitted to two servers to increase the number of requests per unit time, and per unit time that can be processed with two servers. The number of requests may be determined. Furthermore, if processing cannot be performed by a plurality of servers, the number of servers that transmit requests can be further increased, and similarly, the number of requests per unit time that can be processed is sequentially determined. Here, the number of requests per unit time that can be processed has been described as an example, but performance profiles related to other metrics (for example, the number of transactions per unit time and performance profiles related to queries per unit time) can be derived in the same way. That's fine.

  In the above description, the performance profile means 120 itself derives the performance profile. However, the performance profile is prepared in advance by an administrator or the like, and the performance profile means 120 stores the performance profile in advance. It may be. When the performance profile unit 120 stores a performance profile in advance, step A3 (performance profile generation processing) may not be performed. In this case, the system configuration candidate listing unit 131 and the effective range calculation unit 132 read the performance profile from the performance profile unit 120.

  As an aspect of preparing a performance profile without using the performance profile unit 120, the performance profile may be determined from the operation results of the target system 160 that is actually operated. For example, the load (for example, the number of requests, the number of transactions, the number of queries, etc.) when the operation of the target system 160 has slowed down to the grade may be read from the log, and the performance profile may be determined using the load.

  In addition, a server operated in the target system 160 is taken out from the target system, or a server having the same specifications as the server operated in the target system 160 is prepared, and a load is simulated on the server. The performance profile may be determined by measuring the processing performance over time. For example, a dummy request may be given to the server and the processing performance at that time (for example, response time, number of requests that can be processed per unit time, etc.) may be measured. In order to measure the processing performance of a desired number of servers, the number of servers may be increased or decreased manually or automatically, and a load (for example, a dummy request) may be applied.

  After step A3, the system configuration candidate listing unit 131 derives a balanced system configuration candidate based on the metric correlation and the performance profile (step A4). Here, the system configuration candidate listing unit 131 refers to the ratio of the processing amount of the metric derived in step A2 as the specific content of the metric correlation. The system configuration candidate listing unit 131 refers to the performance profile derived by the performance profile unit 120 (or stored in advance by the performance profile unit 120).

  The system configuration candidate listing unit 131 converts the ratio of the processing amount of the metric derived in step A2 into the ratio of the number of system components in each layer. The number of system components of each layer obtained by this conversion becomes a system configuration candidate. There are the following two methods for converting the ratio of the metric processing amount into the ratio of the number of system components in each layer.

  The first method is a method in which one reference layer is determined, the number of system components in that layer is tentatively determined, and the number of system components in each of the other layers is further determined.

In the first method, the system configuration candidate listing unit 131 determines a reference layer and the number n ₀ of system components E ₀ of the layer. The reference layer may be determined in advance. For example, among the layers shown in FIG. 2, the first layer 210 (which may be the second layer or the third layer) may be determined in advance as a reference layer. The system configuration candidate listing unit 131 provisionally determines the number n ₀ of system components E ₀ of the reference layer (in this example, the server 211 of the first layer 210).

The system configuration candidate enumeration unit 131 refers to the performance profile, and the server when the number of system component elements E ₀ (hereinafter referred to as the server 211) of the standard layer is n ₀ The processing amount of the metric 211 is referred to. Assume that this metric is M _0, and the processing amount of the metric M ₀ when the number of servers 211 is n ₀ is P ₀ . Further, the system configuration candidate listing unit 131 uses the processing amount of the metric M ₀ having a correlation with the metric of the server of the other layer as the processing amount of the metric M ₀ of the server 211 of the reference layer. Refer to

Next, the system configuration candidate listing unit 131 determines the processing amount required for the metric M ₁ of the other layer having a correlation with the metric M ₀ from the ratio of the processing amount of the metric derived in step A2. calculate. Assume that the ratio of the processing amounts of the metrics M ₀ and M ₁ derived in step A2 is j: k. The system configuration candidate listing unit 131 calculates the processing amount P ₁ required for the metric M ₁ as a value that satisfies P ₀ : P ₁ = j: k. Then, the system configuration candidate listing unit 131 obtains the number of system configuration elements (servers of other layers) of other layers that can realize the processing amount P ₁ with respect to the metric M ₁ with reference to the performance profile. Since the relationship between the system components become number of servers and the throughput of the server of the number is derivable information, it is possible to obtain the processing amount P ₁ a number of possible system components from the performance profile.

In this way, the system configuration candidate enumeration means 131, the number of system components other layers when the number of system components that layer was n _0, are sequentially calculated for each layer. A combination of the number n ₀ of system components of the reference layer and the number of system components of each other layer obtained in this way becomes one system configuration candidate.

  The system configuration candidate listing unit 131 changes the number of system configuration elements of the reference layer, and similarly derives system configuration candidates sequentially. For example, system configuration candidates when the number of system components in the reference layer is one, system configuration candidates when two units, system configuration candidates when three units, system configuration candidates when Derived.

  The second method is a method for obtaining the number ratio of each layer by increasing the number of system components having a relatively short processing amount from the minimum system configuration.

In the case of the second method, the system configuration candidate enumeration unit 131 obtains the ratio of the processing amount of the metric of the system configuration candidate of each layer based on the ratio of the processing amount of the metric of the system configuration candidate having different layers. For example, a metric M ₀ of the system components of the first layer only process the ratio of the amount of metrics M ₁ of the system components of the second layer, knowing the relationship between the metric of throughput of the system components of the third layer Absent. Therefore, the system configuration candidate enumeration means is obtained from the processing amount ratio between the metric M _{2 of} the third layer system configuration candidate and M ₁ (or M ₀ ) and the processing amount ratio of M ₀ and M _1. 131 calculates the ratio of the processing amounts of M ₀ , M ₁ , and M ₂ . And processing of the ratio of M ₀ and _{M 1,} and a processing amount of the ratio of _{M 1} and _{M 2,} when determining the ratio of M _0, M 1, _{M 2} of processing amount, the two ratios (in this example M The value of the metric M ₁ that is common to the ratio of ₀ to M _{1 and} the ratio of M ₁ to M ₂ may be shared, and the ratios may be connected. For example, if the ratio of the throughput of M ₀ and M ₁ is 1: 2 and the ratio of the throughput of M ₁ and M ₂ is 2: 1, the ratio of the throughput of M ₀ , M ₁ , M ₂ is What is necessary is just to obtain | require as 1: 2: 1. Hereinafter, this ratio is referred to as a processing amount ratio obtained from metric correlation.

The system configuration candidate enumeration unit 131 determines the number of system components of each layer so that the minimum system configuration is obtained. The system configuration candidate listing unit 131 may determine, for example, that the configuration having one system component in each layer is the minimum system configuration and the number of system components in each layer is one. Assume that the system components of each layer are E ₀ , E ₁ , E _2, and the number of system components of each layer is n ₀ , n ₁ , n ₂ . Here, n ₀ = 1, n ₁ = 1, and n ₂ = 1. This combination of “n ₀ = 1, n ₁ = 1, n ₂ = 1” is one of the system configuration candidates.

The system configuration candidate listing unit 131 refers to the performance profile to determine the ratio of the processing amounts of the metrics M ₀ , M ₁ , and M ₂ . Here, since n ₀ = 1, the processing amount (referred to as P ₀ ) relating to the metric M ₀ when there is one server may be obtained with reference to the performance profile. Similarly, the server throughput metrics M ₁ at the time of the one (the P _1.), And the server (and P _2.) Throughput metric M ₂ at the time one also obtains. The system configuration candidate listing unit 131 sets P ₀ : P ₁ : P _2, which is a processing amount ratio, as the processing amount ratio of each metric M ₀ , M ₁ , M ₂ . This P ₀ : P ₁ : P ₂ is described as the processing amount ratio obtained from the performance profile.

  Next, the system configuration candidate listing unit 131 compares the processing amount ratio obtained from the metric correlation with the processing amount ratio obtained from the performance profile, and the processing amount ratio obtained from the performance profile is relatively the highest. Identify the system components that are becoming smaller. This specifying method will be specifically shown. It is assumed that the processing amount ratio obtained from the metric correlation is p: q: r, and the processing amount ratio obtained from the performance profile is s: t: u. The system configuration candidate listing unit 131 obtains a relative ratio obtained by dividing each term in the processing amount ratio obtained from the performance profile by each corresponding term in the processing amount ratio obtained from the correlation of metrics. That is, the system configuration candidate listing unit 131 calculates s / p: t / q: u / r. The system configuration candidate listing unit 131 selects a system component corresponding to the smallest term among the terms of the relative ratio as “the system component having the smallest processing amount ratio obtained from the performance profile”. As specified.

  For example, it is assumed that the processing amount ratio p: q: r = 1: 1: 1 determined from the metric correlation and the processing amount ratio s: t: u = 1: 2: 3 determined from the performance profile. At this time, the relative ratio is 1: 2: 3, and the system component corresponding to the minimum term “1” in the relative ratio is set to “the system configuration in which the processing amount ratio obtained from the performance profile is the smallest. Element ".

  Further, for example, the processing amount ratio p: q: r = 1: 2: 3 obtained from the metric correlation, and the processing amount ratio s: t: u = 1: 1: 1 obtained from the performance profile. To do. At this time, the relative ratio is 1: 1/2: 1/3, and the system component corresponding to the minimum term “1/3” in the relative ratio is set to “relatively, the throughput ratio obtained from the performance profile is the highest. “System components that are getting smaller”.

If the “system component having the smallest processing amount ratio obtained from the performance profile” is specified, the system component is increased by one. For example, a "relatively performance profile than obtained processing amount ratio is the smallest system components" was E _1. Then, the number of E ₁ is increased by one for the system components “n ₀ = 1, n ₁ = 1, n ₂ = 1”, and “n ₀ = 1, n ₁ = 2, n ₂ = 1”. The combination of the numbers is added as one of the system configuration candidates.

  Thereafter, in the same manner, the system configuration candidate listing unit 131 identifies “the system component having the smallest processing amount ratio obtained from the performance profile” and increases the number of the system components by one. Then, the process of setting a new system configuration candidate is repeated. The system configuration candidate listing unit 131 repeats the above processing until the number of any of the system components reaches the upper limit value or the processing amount is saturated even when the number of system components is increased.

  Here, the number of system component elements to be increased is one by one, but system configuration candidates may be determined while increasing the system component elements by two or more.

  Further, the system configuration candidate listing unit 131 may determine whether or not the number of any of the system components has reached the upper limit value with reference to the above-mentioned number limit value storage unit (not shown). . Further, when the lower limit value of the number of servers in each layer is determined, the system configuration candidate listing unit 131 excludes the system configuration candidates in which the number less than the lower limit value is determined from the system configuration candidates.

  As will be described later, as a mode of connecting a server to a communication network, software that is physically connected to the communication network but is not set up for software installation or communication is used. There is a mode of setting for performing installation and communication. When adding a server in this manner, information indicating the number of servers for each layer that are prepared and physically connected to the communication network is stored in the storage device included in the data processing device 100. Good. And you may refer as an upper limit of the number of servers which this information shows.

  After the system configuration candidates are generated by the system configuration candidate listing unit 131 (after step A4), the effective range calculation unit 132 calculates the effective range (the range of load that can be processed by each server in each layer) (step A5). ). The effective range calculation process will be described below.

  The effective range calculation means 132 derives “a processing amount ratio obtained from metric correlation” in the same manner as in the second method of step A4.

  Further, the effective range calculation unit 132 selects one system configuration candidate from the plurality of system configuration candidates derived in step A4. Based on the number of system components of each layer determined by the selected system configuration candidate, the effective range calculation unit 132 calculates the “processing amount ratio obtained from the performance profile” in the same manner as the second method of step A4. To derive.

  Further, the effective range calculation means 132 determines “relatively, similarly to the second method of step A4, from the“ processing amount ratio obtained from the metric correlation ”and“ processing amount ratio obtained from the performance profile ”. The system component that has the smallest throughput ratio determined from the performance profile is identified. Then, the effective range calculation unit 132 determines how many specified system configuration elements are determined by the selected system configuration candidates.

  The effective range calculation unit 132 refers to the performance profile and obtains a processing amount corresponding to the number of units. Let Q be this processing amount. The effective range calculating unit 132 calculates the processing amount of the other layer based on the “processing amount ratio obtained from the correlation of metrics” with the processing amount Q as a reference. For example, “the system component having the smallest processing amount ratio obtained from the performance profile” is the first layer server 211, and the first corresponding to the number of servers 211 determined by the system configuration candidates. It is assumed that the processing amount Q of the layer is obtained. At this time, the processing amount Q ′ of the second layer and the processing amount Q ″ of the third layer are obtained. The effective range calculation means 132 determines the processing amount Q ′ of the second layer so that Q: Q ′: Q ″ is equal to “a processing amount ratio obtained from metric correlation (for example, p: q: r)”. And the processing amount Q ″ of the third layer is obtained. That is, the processing amounts Q ′ and Q ″ such that Q: Q ′: Q ″ = p: q: r are obtained.

  The effective range calculation unit 132 sets the processing amount of each layer obtained in this way as the upper limit of the effective range of each layer in the selected system configuration candidate. In addition, the effective range calculation unit 132 sets the lower limit of the effective range of each layer in the selected system configuration candidate to 0. In the above example, the effective range of the first layer 210 is 0 to Q, the effective range of the second layer 220 is 0 to Q ′, and the effective range of the third layer 230 is 0 to Q ″. .

  When the effective range calculation unit 132 determines the effective range of each layer with respect to the selected system configuration candidate, it selects an unselected system configuration candidate and similarly determines the effective range of each layer in the system configuration candidate. The effective range calculation unit 132 repeats the above processing until there is no unselected system configuration candidate. As a result, an effective range corresponding to the system configuration candidate is derived for each system configuration candidate.

After step A5, the trigger generation means 133 determines a trigger based on the boundary of the effective range (the upper limit of each effective range) (step A6). The trigger generation unit 133 arranges the system configuration candidates derived by the system configuration candidate listing unit 131 in ascending order so that the number of servers in each layer (the number of system components) increases. Furthermore, the trigger generation means 133 arranges the effective ranges (combinations of effective ranges of each layer) corresponding to each system configuration candidate in order according to the order of the system configuration candidates. And the boundary of the effective range of the system configuration candidate arranged in order is specified. For example, the effective range (combination of effective ranges of each layer) corresponding to a certain system configuration candidate A is “0 to Q ₁ for the first layer server. 0 to Q ₂ for the second layer server. Third layer server. Is 0 to Q _3. ”And the valid range corresponding to the next system configuration candidate B is“ 0 to Q ₁ ′ for the first layer server. 0 to Q ₂ ′ for the second layer server. It is assumed that 0 to Q ₃ ′ ”in the server of the layer. The trigger generation unit 133 specifies the upper limit of the effective range of the system configuration candidate having the smaller number of system components as a boundary. In this example, the boundary of the effective range is specified as “Q ₁ in the first layer, Q ₂ in the _second layer, and Q _{3 in} the third layer”.

Subsequently, the trigger generation unit 133 determines a value that is a predetermined number lower than the specified boundary as a trigger. In the case of the above example, for example, the first layer trigger is defined as Q ₁ -k, the second layer trigger is defined as Q ₂ -k, and the third layer trigger is defined as Q ₃ -k. Incidentally, the fact that the trigger of the first layer is Q ₁ -k, once the metric in the first layer is increased from a value lower than Q ₁ -k to Q ₁ -k, switched to a more effective range wide system configuration candidates It means that it should, the metric in the first layer After decreases from a value higher than Q ₁ -k to Q ₁ -k, which means that it should switch to a more effective range narrower system configuration candidates.

In addition, the trigger generation unit 133 may determine the trigger by multiplying the specified boundary by a predetermined ratio (a predetermined ratio lower than 100%). For example, in the case of the above example, for example, the trigger of the first layer is set to 0.9 · Q ₁ , the trigger of the second layer is set to 0.9 · Q ₂ , and the trigger of the third layer is set to 0.9 · Q _3. May be.

  The trigger generation means 133 determines a trigger for each boundary of the effective range of system configuration candidates.

  Next, the operation sequence generation unit 134 generates an operation sequence from the difference between the system configuration candidates (step A7). The operation sequence generation unit 134 obtains a difference between consecutive system configuration candidates among the system configuration candidates arranged in ascending order so that the number of system components of each layer increases. That is, with respect to two system configuration candidates arranged in a row, system configuration elements (servers) included only in one system configuration candidate are obtained. The operation sequence generation unit 134 reads operation information for adding a server that becomes a difference to a layer from the system operation storage unit 140, and a set of operation information for adding a server that becomes a difference between two system configuration candidates Is one operation sequence. Similarly, the operation information for excluding the server that becomes the difference from the layer is read from the system system operation storage unit 140, and a set of operation information for excluding the server that becomes the difference between the two system configuration candidates is stored in one The operation sequence.

  Similarly, the operation sequence generation unit 134 determines an operation sequence for adding a server as a difference and an operation sequence for excluding a server as a difference for each of two consecutive system configuration candidates arranged in series. To go.

  As a mode of connecting a server to a communication network, a person who is not physically connected can be physically connected to the communication network, and software can be installed on the server, or settings of the server can be connected to the communication network. There is a mode of setting so that communication can be performed. In this connection mode, as operations for the operation target element, for example, “a server is physically connected to a communication network”, “software is installed on a server”, “communication via a communication network”. Operations such as “change server settings” and “start software” are included. In the case where a human physically connects a server that is not physically connected to the communication network, information including operation information such as “connect the server physically to the communication network” is received from the system operation storage unit 140. Read and define an operation sequence including operation information “connect the server physically to the communication network”.

  In addition, as a mode of connecting the server to the communication network, software installation or communication is performed on a server that is physically connected to the communication network but is not set up for software installation or communication. There is a mode of performing setting for performing. In this case, for example, an operation sequence including information such as “change the setting so that the server can communicate”, “install software on the server”, “start software”, and the like may be defined. Good.

  It is determined in advance whether the server is connected to the communication network in any of the above two modes, and operation information corresponding to the mode is stored in the system operation storage unit 140 in advance.

  In step A 7, operation information for adding or excluding system configuration elements that are differences between system configuration candidates is not stored in the system operation storage unit 140, and the operation sequence generation unit 134 is stored in the system operation storage unit 140. When the operation information cannot be read, the operation sequence generation unit 134 outputs a message indicating that the operation information should be prepared and stored in the system operation storage unit 140 (for example, the message is displayed on the display device). Display).

  After step A7, the rule generation unit 130 generates a system configuration change rule by combining the trigger obtained in step A6 and the operation sequence obtained in step A7 (step A8). A trigger and an operation sequence are obtained for each of two system configuration candidates arranged in succession. The trigger obtained based on the boundary between the two system configuration candidates serves as a trigger for switching from one of the two system configuration candidates to the other. Of the two system configuration candidates, the system configuration change rule for switching from the system configuration candidate with the smaller system component to the system configuration candidate with the larger system component is based on the trigger obtained from the boundary of the effective range, the difference, What is necessary is just to produce | generate by combining the operation sequence for adding the server which becomes. Further, the system configuration change rule for switching from the system configuration candidate having the larger system component to the system configuration candidate having the smaller system component out of the two system configuration candidates has the trigger obtained from the boundary of the effective range, What is necessary is just to combine with the operation sequence for excluding the server used as a difference.

  For example, the rule generation unit 130 transmits the generated system configuration change rule to the monitoring system 170, and causes the monitoring system 170 to hold the system configuration change rule. The monitoring system 170 monitors the state of the target system 160, and when it becomes a state that satisfies the trigger of the held system configuration change rule, adds a server according to the operation sequence of the system configuration change rule, or Perform the operation to delete the server. For example, “change the setting so that the server can communicate”, “install the software on the server”, “start the software”, “change the setting of the server so that communication is disabled”. , According to the individual information included in the operation sequence such as “Exit software”, “Delete software from server”, etc., operations such as server setting change, software installation and deletion via the communication network Do. However, the monitoring system 170 cannot physically connect the server to the communication network. Therefore, regarding the operation of “connecting the server physically to the communication network”, the operation information is displayed on the display device (not shown in FIG. 1), and the server is physically connected to the administrator. Inform them that they should be connected.

  Further, in the above description, the inter-element correlation deriving unit 110 has shown a case where a combination of metrics having correlation and a correlation between metrics are derived by regression analysis. However, the inter-element correlation deriving unit 110 By analyzing the programs installed in the servers 211, 211, and 231 of each layer, a combination of metrics having correlation and a correlation between metrics may be derived.

Next, the effect of this embodiment will be described.
According to the present embodiment, the inter-element correlation deriving means 110 measures the server metrics of the target system 160 and derives the correlation between the measured metrics. Then, the system configuration candidate listing unit 131 derives system configuration candidates based on the metric correlation and the performance profile. Therefore, a combination that does not have too many or too few system components in some layers can be identified as a system configuration candidate.

  Further, according to the present embodiment, the effective range calculation unit 132 calculates the effective range (the load range that can be processed by each server in each layer) for each system configuration candidate, and the trigger generation unit 133 determines the effective range. A condition (trigger) for changing the configuration of the target system 160 is determined based on the boundary. Further, the operation sequence generation unit 134 reads operation information for adding or excluding differences between system configuration candidates (system configuration elements included only in one system configuration candidate) from the system operation storage unit 140, and autonomously The control rule generation unit 130 generates a system configuration change rule by combining a trigger and an operation sequence that is a set of operation information. Therefore, it is possible to generate a system configuration change rule without depending on human experience and results. Further, since the system configuration change rule can be generated regardless of human skill, an appropriate system configuration rule can be easily generated.

Embodiment 2. FIG.
FIG. 6 is a block diagram illustrating an example of a system configuration candidate derivation device according to the second embodiment of this invention. Components similar to those in the first embodiment are denoted by the same reference numerals as those in FIG. In the second embodiment, the data processing apparatus 100 includes an inter-element correlation deriving unit 110, a performance profile unit 120, a system configuration candidate listing unit 131, an effective range calculating unit 132, a trigger generating unit 133, and an operation sequence generating unit 134. In addition to the rule generation unit 130, a system configuration candidate narrowing-down unit 135 is provided.

  As already described, the effective range calculation means 132 sets the lower limit of the effective range of each layer in each system configuration candidate to be selected to 0. Accordingly, the lower limit of the effective range of each system configuration candidate is 0, and the effective range of a certain system configuration candidate includes the effective range of other system configuration candidates with fewer system components, and the two system configurations There are overlapping parts in the effective range of candidates.

  The system configuration candidate narrowing means 135 performs a process of excluding one of the system configuration candidates from the system configuration candidates when the overlapping portion between the valid ranges of different system configuration candidates is large.

  FIG. 7 is a flowchart illustrating an example of processing progress of the system configuration candidate derivation device according to the second embodiment. About the process similar to 1st Embodiment, the code | symbol same as the code | symbol shown in FIG. 5 is attached | subjected, and description is abbreviate | omitted. The system configuration candidate narrowing means 135 determines whether or not there is a large overlap between effective ranges of different system configuration candidates after the effective range of each system configuration candidate is calculated in step A5, and the overlap is large. If it is determined, either one of the system configuration candidates is excluded from the system configuration candidates, and only the other system configuration candidate is left as a system configuration candidate (step A5-a). Thereafter, the operations after Step A6 are performed.

  Hereinafter, the process of step A5-a will be described in detail. First, overlapping of effective ranges will be described. FIG. 8 is an explanatory diagram showing examples of effective ranges of two system configuration candidates. The portion shown in white represents the effective range of the system configuration candidate (hereinafter referred to as candidate 1) with the smaller number of system components. The hatched portion represents the effective range of the system configuration candidate (hereinafter referred to as candidate 2) having the larger number of system components. Further, FIG. 8 shows the effective range of correlated metrics from the first layer to the third layer. The valid range of the first layer metric of candidate 1 is 0 to h, the valid range of the first layer metric of candidate 2 is 0 to h ′, and the ranges of 0 to h overlap. The valid range of the metric of the second layer of candidate 1 is 0 to i, the valid range of the metric of the second layer of candidate 2 is 0 to i ', and the ranges of 0 to i overlap. The effective range of the third layer metric of candidate 1 is 0 to j, the effective range of the third layer metric of candidate 2 is 0 to j ', and the ranges of 0 to j overlap.

  The system configuration candidate narrowing-down means 135 has a predetermined ratio (for example, 90) of the effective range of the larger effective range (that is, the effective range of the candidate 2) with respect to each correlated metric of each layer. %) Or more, it is determined that the overlapping portion between the effective ranges is large. Hereinafter, the case where the predetermined ratio is 90% will be described as an example. In the case of the example shown in FIG. 8, if h> 0.9 · h ′, i> 0.9 · i ′, j> 0.9 · j ′ are all satisfied for the effective range of the metric of each layer, The system configuration candidate narrowing-down means 135 determines that the overlapping portion between the effective ranges is large.

  When it is determined that the overlapping portion between the effective ranges is large, the system configuration candidate narrowing means 135 excludes one of the candidates 1 and 2 from the system configuration candidates. At this time, it is preferable to exclude a system configuration candidate having a narrower effective range (a system configuration candidate having a smaller number of system components, in this case, candidate 1).

  Further, the system configuration candidate narrowing-down means 135 determines the effective range when the overlapping portion of the effective range is greater than or equal to a predetermined ratio of the larger effective range with respect to a predetermined metric among the correlated metrics of each layer. It may be determined that the overlapping portion is large. For example, when h> 0.9 · h ′ and i> 0.9 · i ′ are established for the first layer metric and the second layer metric, it is determined that the overlapping portion between the effective ranges is large. May be.

  Further, the system configuration candidate narrowing-down means 135 for each correlated metric of each layer, the upper limit value of the smaller effective range (that is, the effective range of candidate 1) and the larger effective range (that is, the effective range of candidate 2) ) When the difference from the upper limit value is a predetermined value (for example, 10) or less, it may be determined that the overlapping portion between the effective ranges is large. In the case of the example shown in FIG. 8, the effective ranges overlap when h′−h> 10, i′−i> 10, and j′−j> 10 are all established for the effective ranges of the metrics of the respective layers. You may determine that a part is large.

  Further, the system configuration candidate narrowing-down means 135 has a difference between the upper limit value of the smaller effective range and the upper limit value of the larger effective range that is less than or equal to a predetermined value for a predetermined metric among the correlated metrics of each layer. In some cases, it may be determined that the overlapping portion between the effective ranges is large. For example, when h′−h> 10 and i′−i> 10 hold for the first layer metric and the second layer metric, it may be determined that the overlapping portion between the effective ranges is large. .

  The predetermined ratio or the predetermined value in the above example may be different for each metric having a correlation.

  The system configuration candidate narrowing means 135 selects two system configuration candidates from the remaining system configuration candidates as system configuration candidates, determines whether or not the overlapping portion between the effective ranges is large, and the overlapping portion is If it is determined that it is larger, one of the system configuration candidates is deleted. This process is repeated until there are no system configuration candidates to be deleted. Thereafter, the operations after Step A6 are performed.

  Also in the second embodiment, as in the first embodiment, an effect that the system configuration rule can be easily generated is obtained. Furthermore, in the second embodiment, it is possible to narrow down the system configuration candidates by excluding one of the different system configuration candidates having many overlapping valid ranges from the system configuration candidates.

  The system configuration candidate narrowing means 135 is realized by, for example, a computer (central processing unit) that operates according to a program. The system configuration candidate narrowing down means 135 may be realized by the same central processing unit as other means.

Embodiment 3 FIG.
FIG. 9 is a block diagram illustrating an example of a system configuration candidate derivation device according to the third embodiment of this invention. Components similar to those in the first embodiment are denoted by the same reference numerals as those in FIG. In the third embodiment, the data processing apparatus 100 includes an inter-element correlation deriving unit 110, a performance profile unit 120, a system configuration candidate listing unit 131, an effective range calculating unit 132, a trigger generating unit 133, and an operation sequence generating unit 134. In addition to the rule generation unit 130, a system configuration candidate presentation unit 136 is provided.

  The system configuration candidate presentation unit 136 displays a list of system configuration candidates derived by the system configuration candidate 131, and displays the list of system configuration candidates as a user (for example, an administrator of the target system 106 and the system configuration candidate derivation device). To present. The system configuration candidate presenting unit 136 excludes the system configuration candidate from the plurality of derived system configuration candidates when the user inputs information on the system configuration candidate to be excluded.

  The system configuration candidate presentation unit 136 is realized by, for example, a computer (central processing unit) that operates according to a program, a display device, and an input device such as a mouse or a keyboard. The central processing unit for realizing the system configuration candidate presentation unit 136 may be the same as the central processing unit for realizing other units.

  FIG. 10 is a flowchart illustrating an example of processing progress of the system configuration candidate derivation device according to the third embodiment. About the process similar to 1st Embodiment, the code | symbol same as the code | symbol shown in FIG. 5 is attached | subjected, and description is abbreviate | omitted. The system configuration candidate narrowing means 135 displays a list of system configuration candidates after the effective range of each system configuration candidate is calculated in step A5, and when the information of system configuration candidates to be excluded is input by the user Then, the system configuration candidate is excluded from the plurality of derived system configuration candidates (step A5-b). Thereafter, the operations after Step A6 are performed.

  FIG. 11 is an explanatory diagram illustrating an example of a display screen of a list of system configuration candidates displayed by the system configuration candidate presenting unit 136. The system configuration candidate list display screen includes, for example, a system configuration candidate display column 401, a system configuration candidate effective range display column 402, a check box 403 for designating system configuration candidates to be excluded, and metrics. And a pull-down menu 404 for selecting.

  In step A5-b, the system configuration candidate presenting means 136 displays each display column 401 for displaying each system configuration candidate, and each display column 402 for displaying a check box 402 and an effective range corresponding to each display column 401. And a screen including a pull-down menu 404 is displayed on the display device. The system configuration candidate presenting means 136 displays the default metric name in the pull-down menu 404 and displays the effective range of the metric in each display column 402. When the user operates the pull-down menu 404 and designates another metric, the effective range of the metric is displayed in each display column 402.

  The user clicks the check box 403 corresponding to the system configuration candidate to be excluded with the mouse. When the check box 403 is clicked with the mouse, the system configuration candidate presentation unit 136 displays a check mark in the clicked check box. Then, the system configuration candidate corresponding to the clicked check box 403 is excluded from the system configuration candidates derived in step A4. Thereafter, the operations after Step A6 are performed.

  Also in the third embodiment, an effect that a system configuration rule can be easily generated is obtained as in the first embodiment. Furthermore, system configuration candidates deemed unnecessary by the user can be deleted at the user's discretion.

  The data processing apparatus 100 may include both of the system configuration candidate narrowing down means 135 in the second embodiment and the system configuration candidate presenting means 136 in the third embodiment. In the case of such a configuration, after the system configuration candidate narrowing-down means 135 performs the process of step A5-a in the second embodiment, the system configuration candidate presenting means 136 performs step A5- in the third embodiment. The process of b may be performed, and then the processes after step A6 may be performed. Alternatively, after the system configuration candidate presentation unit 136 performs the process of step A5-b in the third embodiment, the system configuration candidate narrowing unit 135 performs the process of step A5-a in the second embodiment. Thereafter, the processing after step A6 may be performed.

  In each embodiment from the first embodiment to the third embodiment, the system configuration candidate enumeration unit 131 temporarily stores the derived system configuration candidates in the storage device, and the effective range calculation unit 132 The system configuration may be read from the storage device. FIG. 12 is a block diagram illustrating an example of a system configuration candidate derivation device that temporarily stores system configuration candidates in a storage device. Components similar to those in the first embodiment are denoted by the same reference numerals as those in FIG. In addition to the inter-element correlation deriving unit 110, the performance profile unit 120, the system configuration candidate listing unit 131, the effective range calculation unit 132, the trigger generation unit 133, the operation sequence generation unit 134, and the rule generation unit 130. Further, system configuration candidate storage means 141 is provided.

  The system configuration candidate storage unit 141 is a storage device that stores the system configuration candidates derived by the system configuration candidate listing unit 131. The system configuration candidate listing unit 131 stores the derived system configuration candidates in the system configuration candidate storage unit 141. At this time, the system configuration candidate storage unit 141 also stores in the system configuration candidate storage unit 141 information on the correlation of the metric derived by the inter-element correlation deriving unit 110. Then, the effective range calculation unit 132 reads the information on the correlation between each system configuration candidate and the metric from the system configuration candidate storage unit 141 and performs the process of step A5.

  Note that a performance profile is also used in the process of step A5. As in each of the embodiments already described, the performance profile unit 120 may pass the performance profile to the effective range calculation unit 132. Alternatively, the performance profile unit 120 may store the performance profile in advance, and the effective range calculation unit 132 may read the performance profile from the performance profile unit 120. Further, the system configuration candidate listing unit 131 stores each performance profile generated by the performance profile unit 120 (or each performance profile stored in advance by the performance profile unit 120) in the system configuration candidate storage unit 141 together with the system configuration candidates. It may be memorized. Then, the effective range calculation unit 132 may read the system configuration candidate and the performance profile from the system configuration candidate storage unit 141.

  Also in the configuration as shown in FIG. 12, the data processing apparatus 100 may include system configuration candidate narrowing means 135 (see FIG. 6) and system configuration candidate presenting means 136 (see FIG. 9).

  Further, the system configuration candidate derivation device of the present invention does not include the effective range calculation unit 132, the trigger generation unit 133, the operation sequence 134, and the rule generation unit 130, and performs the processing from step A1 to A4. Good. Even in such a configuration, the system configuration candidate enumeration unit 131 can derive system configuration candidates based on the correlation of metrics and the performance profile. Combinations that are neither too much nor too few can be derived as system configuration candidates.

  Next, an example of a specific device for realizing the system configuration candidate derivation device of the present invention will be described. FIG. 13 is an explanatory diagram illustrating an example of a computer that implements the system configuration candidate derivation device according to the present invention. The computer 1000 includes a CPU 1001, an external storage device 1003, a memory 1002, an input device 1004, and an output device 1005 such as a display device. The memory 1002 stores a program, and the CPU 1001 according to the program, the inter-element correlation deriving unit 110, the profile unit 120, the system configuration candidate listing unit 131, the effective range calculating unit 132, the trigger generating unit 133, and the operation sequence generating unit 134. It operates as the rule generation means 130. In the case of the second embodiment or the third embodiment, the system configuration candidate narrowing means 135 and the system configuration candidate presentation means 136 are also operated.

  The external storage device 1003 stores the derived system configuration candidate and the generated system configuration change rule. After generating the system configuration change rule, the CPU 1001 may store the system configuration change rule in the external storage device 1003. The external storage device 1003 may store information indicating the number of servers for each layer that are prepared in advance and are physically connected to the communication network. This information is referred to as the upper limit value of the system component of each layer that can be included in the system configuration candidate.

  The input device 1004 is an input device such as a mouse, a keyboard, or a touch panel. An instruction from the user (for example, an instruction to start or end the operation of the present invention) is input via the input device 1004.

  Examples of the present invention will be described below. The configuration of the target system 160 is assumed to be the configuration shown in FIG. However, the first layer 210, the second layer 220, and the third layer 230 are a Web layer, an AP layer, and a DB layer, respectively. Hereinafter, the servers 211, 221, and 231 of each layer are referred to as a Web server 211, an AP server 221, and a DB server 231, respectively. When the DB layer 210 is accessed from outside the target system 160, the Web server 211, the AP server 221 and the DB server 231 operate in cooperation and return a response to the access.

  Such operation information for the operation target element of the target system 160 is described in a format such as a shell script in UNIX (registered trademark) or a workflow language such as WS-BPEL, and stored in the system operation storage unit 140 in advance. ing.

  The inter-element correlation deriving means 110 collects a metric representing the state of the system component (each server) by receiving the execution log generated by each server 211, 221, 231 (step A1). The inter-element correlation deriving unit 110 may receive a metric from the monitoring system 170 that monitors the target system 160.

  Next, the inter-element correlation deriving means 110 identifies a combination of metrics having correlation among the collected metrics, and derives the correlation between the metrics (step A2). A regression analysis method may be applied to this process.

  In this embodiment, it is assumed that there is a correlation between the number of transactions executed by the AP server 221, the number of requests processed by the Web server 211, and the number of queries executed by the DB server 231. The number of transactions executed by the AP server 221 is twice the number of requests processed by the Web server 211, and the number of queries executed by the DB server 231 is 1/2 of the number of transactions executed by the AP server 221. Assume that there is a correlation. That is, it is assumed that there is the following correlation.

Number of transactions = 2 x number of requests Query execution number = 0.5 x number of transactions

  Note that the inter-element correlation deriving means 110 may derive combinations of metrics having correlations or correlations between metrics by analyzing programs installed in the servers 211, 221, and 231 of each layer. Good. For example, it is assumed that the program installed in the Web server 211 describes a code that causes the AP server 221 to execute two transactions when the Web server 211 receives one request. In addition, it is assumed that a code for issuing a query to the DB server 231 is described in one of the two transactions performed by the AP server 221 in the program installed in the AP server 221. The inter-element correlation deriving unit 110 may derive the correlation of the above example from these codes.

  Further, the performance profile unit 120 generates a performance profile (step A3). The performance profile means 120 collects metrics representing the state of system components (each server) through the execution log generated by each server, the server or communication network monitoring system 170, and the system configuration (target system) at that time. 160, the configuration of the target system, such as the number of servers operable, is also acquired. The performance profile unit 120 compares the system configuration and metrics, and generates a performance profile such as how many requests and how many requests the Web server can handle.

  In addition, the relationship between the number of system components and the processing amount can be obtained by taking out the system components operated in the target system 160 from the target system in advance and increasing or decreasing the number of system components in the test environment. You may measure. And the structure which the performance profile means 120 has memorize | stored as a performance profile as the measurement result may be sufficient.

  FIG. 14 is an explanatory diagram illustrating an example of a performance profile generated by the performance profile unit 120. FIG. 14A shows the relationship between the number of Web servers 211 and the number of requests per unit time that can be processed by each Web server 211. More specifically, the number of requests that can be processed by one Web server per unit time is 70, which indicates that the number of Web servers and the number of requests that can be processed per unit time are in a proportional relationship. FIG. 14B shows the relationship between the number of AP servers 221 and the number of transactions per unit time that can be processed by each AP server 221. Specifically, the number of transactions that can be processed by one AP server per unit time is 100, which indicates that the number of AP servers and the number of transactions that can be processed per unit time are in a proportional relationship. FIG. 14C shows the relationship between the number of DB servers 231 and the number of queries per unit time that can be processed by each DB server 231. Specifically, the number of DB servers and the number of queries that can be processed per unit time are in a proportional relationship, but if the number of DB servers becomes four or more, queries that can be processed in unit time even if the number of DB servers is increased. The number does not increase.

  The system configuration candidate listing unit 131 derives and lists system configuration candidates (step A4). In this example, as described above, there is a correlation of “number of transactions = 2 × number of requests” and “number of queries executed = 0.5 × number of transactions”. That is, there is a correlation that “the number of requests in the Web layer: the number of transactions in the AP layer: the number of query executions in the DB layer is 1: 2: 1”.

  The system configuration candidate listing unit 131 derives system configuration candidates by the first method or the second method already described.

  FIG. 15 is an explanatory diagram illustrating an example of system configuration candidates derived by the system configuration candidate listing unit 131. The number on the left side of each column represents the number of servers. For example, candidate 1 is a system configuration candidate “one Web server, one AP server, and one DB server”. The numerical value in parentheses on the right side of each column represents the number of processes that can be processed by the left number of servers. For example, in Candidate 1, 70 requests can be processed per unit time by one Web server, 50 transactions can be executed per unit time by one AP server, and 25 queries per unit time can be executed by one DB server. It can be executed.

  In this example, as shown in FIG. 14C, when the number of DB servers 4 becomes four or more, the processing capacity of the DB layer is saturated, so the number of servers included in the system configuration candidates is increased. When the system configuration candidate is derived so that the number of DB servers is four, the process of step A4 is terminated. In addition, the number of servers prepared for each layer that are prepared in advance and physically connected to the communication network in the target system is stored in advance, and the step is performed when that number is reached in any layer in the system configuration candidate. You may complete | finish the process of A4.

  The effective range calculation unit 132 calculates the effective range (the load range that can be processed by each server in each layer) (step A5). FIG. 16 is an explanatory diagram illustrating an example of the effective range obtained by the effective range calculation unit 132. As shown in FIG. 15, in candidate 1, the Web server can process 70 requests per unit time, but “the number of requests in the Web layer: the number of transactions in the AP layer: the number of queries executed in the DB layer is 1: 2: 1 ”. Therefore, since the number of transactions that can be processed by the AP server per unit time is 50, as shown in FIG. 16, in the candidate 1, the Web server processes 0 to 25 requests per unit time. Even if the Web server tries to process more requests, the performance of the entire target system will be saturated because it cannot be processed by the AP layer.

  Further, when the system configuration candidate derivation device includes the system configuration candidate narrowing means 135 (see FIG. 6), the system configuration candidate narrowing means 135 determines that the system configuration candidate when the overlapping range between the effective ranges is large. One of the configuration candidates is excluded from the system configuration candidates (step A5-a). For example, the effective ranges of candidate 3 and candidate 4 illustrated in FIG. 16 are compared. The valid range for the number of requests is “0 to 70” for candidate 3 and “0 to 75” for candidate 4. The overlapping portion “0 to 70” of the two effective ranges occupies 90% or more of the effective range “0 to 75” of the candidate 4. The same applies to the valid ranges related to the number of transactions and the number of query executions. In this case, the system configuration candidate narrowing down means 135 deletes either candidate 3 or candidate 4. It is preferable to delete candidate 3 while leaving candidate 4 with a wide effective range.

  Next, the trigger generation means 133 generates a trigger based on the boundary of the effective range (step A6). The generation of a trigger for changing the system configuration from candidate 1 to candidate 2 will be described as an example. The boundary values of the valid ranges of the metric of candidate 1 and candidate 2 are 25 for the number of requests in the Web layer, 50 for the number of transactions in the AP layer, and 25 for the number of query executions in the DB layer. The trigger generation means 133 uses a value subtracted from the value of each boundary as a trigger by a ratio designated by the operation manager or a preset ratio such as 10%. For example, a trigger is generated that “the number of requests in the Web layer exceeds 20, or the number of transactions in the AP layer exceeds 40, or the number of query executions in the DB layer exceeds 20.” This trigger is a condition for changing the system configuration from candidate 1 to candidate 2.

  The operation sequence generation unit 134 generates an operation sequence from the difference between the system configuration candidates (step A7). For example, candidate 2 has one more AP server and one DB server than candidate 1 (see FIG. 15). Therefore, one AP server and one DB server are the differences between candidates 1 and 2. The operation sequence generation unit 134 extracts the operation information for adding the AP server and the operation information for adding the DB server from the system operation storage unit 140, and sets the information as one operation sequence. If the operation information cannot be read from the system operation storage unit 140, the operation sequence generation unit 134 outputs a message indicating that the operation information should be prepared and stored in the system operation storage unit 140.

  The rule generation means 130 combines the trigger and the operation sequence into a system configuration change rule.

  The present invention is suitable as a system configuration candidate derivation device for deriving a combination of system components included in a system to be managed and in which some components are not excessively large or small. Applied.

It is a block diagram showing an example of a system configuration candidate derivation device of a 1st embodiment of the present invention. It is explanatory drawing which shows the example of an object system. It is explanatory drawing which shows the example of a performance profile. It is explanatory drawing which shows the example of a performance profile. It is a flowchart which shows the example of a process progress of the system configuration candidate derivation device of a 1st embodiment. It is a block diagram which shows the example of the system configuration candidate derivation device of the 2nd embodiment of the present invention. It is a flowchart which shows the example of process progress of the system configuration candidate derivation device of a 2nd embodiment. It is explanatory drawing which shows the example of the effective range of two system configuration candidates. It is a block diagram which shows the example of the system configuration candidate derivation device of the 3rd Embodiment of this invention. It is a flowchart which shows the example of process progress of the system configuration candidate derivation device of a 3rd embodiment. It is explanatory drawing which shows the example of the display screen of the list of system configuration candidates. It is a block diagram which shows the example of the system configuration candidate derivation apparatus which memorize | stores a system configuration candidate in a memory | storage device once. It is explanatory drawing which shows the example of the computer which implement | achieves the system structure candidate derivation | leading-out apparatus of this invention. It is explanatory drawing which illustrated the example of the performance profile. It is explanatory drawing which shows the example of a system configuration candidate. It is explanatory drawing which shows the example of an effective range. It is explanatory drawing which shows the example of the system used as management object.

Explanation of symbols

DESCRIPTION OF SYMBOLS 100 Data processor 110 Inter-element correlation derivation means 120 Performance profile means 130 Rule generation means 131 System configuration candidate enumeration means 132 Effective range calculation means 133 Trigger generation means 134 Operation sequence generation means 140 System operation storage means

Claims (10)

A system configuration candidate derivation device for deriving a system configuration candidate that is a candidate for configuration of a managed system that includes a plurality of layers that are a set of system components that are loaded with the same type of software and operate according to the software,
An inter-element correlation deriving means for measuring a metric of each layer indicating the load status of the managed system as a value and deriving a correlation of values between metrics in a set of metrics measured from different layers ;
System configuration candidates for deriving system configuration candidates based on performance profiles that are information that can derive the relationship between the number of system components and the processing amount of the system components, and the correlation of values between metrics A system configuration candidate derivation device comprising: enumeration means.   The system configuration candidate derivation device according to claim 1, further comprising performance profile means for providing a performance profile. The inter-element correlation deriving means continuously measures the metric as time passes, substitutes the metric at each time for an expression including multiplication of the variable and the coefficient, using the metric as a variable, and calculates the coefficient. The system configuration candidate derivation device according to claim 1 or 2, wherein if the calculated variance of coefficients is equal to or less than a threshold value, the correlation of values between metrics is derived based on the coefficients. The system configuration candidate enumeration means obtains the processing amount in the number of units in the reference layer based on the performance profile for each number of system components in one reference layer, and the value between the processing amount and the metric The system configuration candidate is derived by calculating the processing amount in the other layer based on the correlation between the two, and deriving from the performance profile the number of units capable of processing the processing amount in the other layer. The system configuration candidate derivation device according to any one of the three. The system configuration candidate enumeration means obtains the throughput ratio of each layer from the correlation of values between metrics, sets the number of system components in each layer to the initial value, obtains the throughput ratio of each layer from the performance profile, and from the performance profile Identify the system component of the layer corresponding to the term with the smallest division result by dividing each term of the obtained throughput ratio by each corresponding term of the throughput ratio obtained from the correlation of values between metrics. The system configuration candidate derivation device according to any one of claims 1 to 3, wherein a system configuration candidate is derived by repeating the addition of the number of system component elements in a layer. A system configuration candidate derivation program installed on a computer for deriving a system configuration candidate that is a candidate for configuration of a managed system including a plurality of layers that are a set of system components that operate with the same type of software. ,
In the computer,
Inter-element correlation derivation processing that measures the metrics of each layer showing the load status of the managed system as values and derives the correlation of values between metrics in the metric pairs measured from different layers , and the number of system components To execute system configuration candidate enumeration processing that derives system configuration candidates based on the performance profile, which is information that can derive the relationship between the number of system components and the amount of processing of the system components, and the correlation of values between metrics System configuration candidate derivation program. On the computer,
In the inter-element correlation derivation process, the metric is continuously measured over time, the metric at each time is substituted into an equation including multiplication of the variable and the coefficient as the metric, and the coefficient is calculated. The system configuration candidate derivation program according to claim 6, wherein if the variance of the calculated coefficient is equal to or less than a threshold value, a process of deriving a correlation of values between metrics based on the coefficient is executed. On the computer,
In the system configuration candidate enumeration process, for each number of system components in one reference layer, the processing amount in the reference layer is calculated based on the performance profile, and the value between the processing amount and the metric And calculating a processing amount in another layer based on the correlation, and deriving a system configuration candidate by deriving from the performance profile the number of the processing amount in the other layer that can be processed. The system configuration candidate derivation program according to claim 6 or claim 7. On the computer,
In the system configuration candidate enumeration process, calculate the throughput ratio of each layer from the correlation of values between metrics, determine the number of system components in each layer as the initial value, determine the throughput ratio of each layer from the performance profile, and from the performance profile Identify the system component of the layer corresponding to the term with the smallest division result by dividing each term of the obtained throughput ratio by each corresponding term of the throughput ratio obtained from the correlation of values between metrics. The system configuration candidate derivation program according to claim 6 or 7, wherein a process for deriving a system configuration candidate is executed by repeatedly adding the number of system component elements of a layer. A system configuration candidate derivation method for deriving a system configuration candidate that is a candidate for the configuration of a managed system including a plurality of layers that are a set of system components that are loaded with the same kind of software and operate according to the software,
The inter-element correlation deriving means measures the metric of each layer indicating the load status of the managed system as a value , derives the correlation of the value between the metrics in the metric set measured from different layers ,
The system configuration candidate enumeration means is configured based on the performance profile, which is information that can derive the relationship between the number of system components and the processing amount of the system components, and the correlation of values between metrics. A system configuration candidate derivation method characterized by deriving candidates.

Images