WO2013141018A1

WO2013141018A1 - Device for supporting optimal system design

Info

Publication number: WO2013141018A1
Application number: PCT/JP2013/056014
Authority: WO
Inventors: さやか伊豆倉
Original assignee: 日本電気株式会社
Priority date: 2012-03-21
Filing date: 2013-03-05
Publication date: 2013-09-26
Also published as: JPWO2013141018A1

Abstract

The invention is provided with a simulator for simulating the operation of a system on the basis of system design information and calculating a predictive value of a service level for the system; a service level decision unit for deciding whether or not the service level predictive value has achieved a preconfigured service level objective; and a parameter set generator for generating, on the basis of the simulation result, a parameter set capable of achieving the service level objective.

Description

[Name of invention determined by ISA based on Rule 37.2] Optimal system design support device

The present invention relates to an optimum system design support apparatus, an optimum system design support method, and a program.

When developing a computer system, an SLA (Service Level Agreement), which is a contract that guarantees a certain level of service quality regarding performance, availability, etc., is generally concluded between the contracted organization of the development and the contracted customer. It has become. The SLA is composed of a plurality of SLOs (Service Level Objective) representing more specific service level target values.

Evaluation indexes that indicate system performance include response time that represents the time from when a processing request is sent to the system until the processing is completed, the usage rate of each resource that makes up the system, and the transfer rate of the process. The maximum throughput to be expressed is often used. As the SLO, it is often set that the response time and the resource usage rate are within an allowable range, and that the maximum throughput is a certain value or more.

Usually, whether or not SLO for such performance is achieved is determined by using a model based on queuing theory, a regression model constructed based on actual measurement data, or knowledge of experts in the technical field. Judgment is made by predicting system performance.

For example, Patent Document 1 uses queuing theory to estimate the response time from the number of requests arriving at the system, the specifications of the servers that make up the system, etc., and determine whether the value falls within a predetermined range. The technology is described.

In Patent Document 2, a performance prediction method based on queuing theory is applied, and the response time is expressed as a function of a parameter set that affects the performance. Deriving a system configuration parameter set is described.

In Patent Document 3, the amount of work (the load on the server) and the hardware configuration have been changed by measuring the operating status of the system and using the least squares method to calculate the work amount and response time. Techniques for predicting response times in cases are described.

Further, in Patent Document 4, as a method for calculating the processing capacity of an application, in addition to a calculation method based on an open queuing theory, interpolation of a plurality of sample data obtained by measuring response times and server usage rates is performed. The calculation method is described.

Further, Patent Document 5 describes a technique for monitoring the system operating status, determining whether the service level is equal to or higher than a certain level from the CPU usage rate of the server, and increasing or decreasing the number of servers provided to the user. Yes.

WO2004 / 092971 JP 2004-030292 A JP 05-324358 A JP 2006-301852 A Japanese Patent No. 4292663

However, in general, it is difficult to accurately formulate response time and maximum throughput values in a system modeled by a queue. For example, in the case of a model including parallel processing and synchronous processing, values of response time and maximum throughput cannot be obtained as a closed expression. In other words, the response time can be formulated as in Patent Documents 1 and 2 is limited to a few systems that perform only limited processing.

Also, as in

Patent Documents

3 and 4, when the behavior of the system is recursively obtained based on the actually measured values and the response time is expressed by an approximate expression, an error occurs.

Patent Document 5 deals only with the CPU usage rate that allows simple proportional calculation as SLO and the upper and lower limits of the number of servers, and does not consider SLO for response time and maximum throughput.

Therefore, an object of the present invention is to automatically derive an inexpensive system configuration that can guarantee to achieve SLO for performance values that cannot be formulated in various forms of systems.

An optimal system design support apparatus according to the present invention executes a simulation of the operation of the system based on system design information, calculates a service level prediction value in the system, and the service level prediction value is A service level determination unit that determines whether a preset service level target is achieved, and a parameter set generation that generates a parameter set that may achieve the service level target based on the result of the simulation Part.

An optimal system design support method according to the present invention includes a step of executing a simulation of the operation of the system based on system design information to calculate a predicted value of a service level in the system, and the predicted value of the service level Determining whether or not a preset service level target is achieved, and generating a parameter set that can achieve the service level target based on the result of the simulation. Is.

The program according to the present invention is a computer that executes a simulation of the operation of the system based on system design information, calculates a service level prediction value in the system, and the service level prediction value includes: A service level determination unit that determines whether a preset service level target is achieved, and a parameter set generation unit that generates a parameter set that can achieve the service level target based on the simulation result , To function as.

According to the present invention, it is possible to guarantee that an SLO for a performance value that cannot be formulated can be achieved in various types of systems, and an inexpensive system configuration can be automatically derived.

The block diagram showing the structure of the optimal system design assistance apparatus by Embodiment 1 of this invention. The figure which shows the specific example of the setting parameter appended to the system model and the system model by Embodiment 1 of this invention. The flowchart of operation | movement of the optimal system design support apparatus by Embodiment 1 of this invention. The block diagram showing the structure of the optimal system design assistance apparatus by Embodiment 2 of this invention. The flowchart of operation | movement of the optimal system design assistance apparatus by Embodiment 2 of this invention. The figure explaining operation | movement of the optimal system design assistance apparatus by Embodiment 2 of this invention. The figure explaining operation | movement of the optimal system design assistance apparatus by Embodiment 2 of this invention. The figure explaining operation | movement of the optimal system design assistance apparatus by Embodiment 2 of this invention.

Embodiment 1 FIG.
Next, embodiments for carrying out the present invention will be described in detail with reference to the drawings.
FIG. 1 is a block diagram showing the configuration of an optimum system design support apparatus 10 according to the first embodiment of the present invention. Specifically, the optimum system design support apparatus 10 is realized by an information processing apparatus such as a computer that operates according to a program. The optimum system design support apparatus 10 may be configured by a single computer or may be configured by a plurality of computers connected to each other via a communication line.

As shown in FIG. 1, the optimal system design support apparatus 10 includes a simulator 11, a simulation result storage unit 12, a service level determination unit 13, an optimal possibility determination unit 14, a parameter set generation unit 15, and an optimal parameter set output unit 16. It has.

The simulator 11, the service level determination unit 13, the optimal possibility determination unit 14, the parameter set generation unit 15, and the optimal parameter set output unit 16 are realized by the CPU executing a predetermined program stored in a ROM or the like. Corresponds to the function module. The simulation result storage unit 12 is implemented by an external storage device.

When a model (system model) representing design information of a system described in a formal language, that is, a machine-processable format is input, the simulator 11 simulates the operation of the system represented by the model, The predicted performance value of the system is calculated. FIG. 2 shows a specific example of the system model and the setting parameters appended to the system model.

The simulation result storage unit 12 is realized by a storage device such as a magnetic disk device or an optical disk device, and stores the predicted performance value of the system calculated by the simulator 11.

The service level determination unit 13 determines whether the predicted response time of the system included in the simulation result has achieved SLO (service level target).

The optimal possibility determination unit 14 compares the price of a system composed of a certain parameter set and another parameter set for the parameter set determined to achieve the SLO in the service level determination unit 13, and the parameter set with a lower price. Is determined to be optimal. Here, the price of the system is represented by a total value of individual prices set in advance for each server, components such as the mounted CPU, and network devices such as routers.

Also, the optimal possibility determination unit 14 determines that the parameter set determined to achieve the SLO after the sufficient number of trials designated in advance is optimal.

Further, the parameter set generation unit 15 refers to the value of the parameter set determined to achieve the SLO in the service level determination unit 13 and the simulation result, and newly sets a parameter set that achieves the SLO and has a lower cost. Generate.

Specifically, the parameter set generation unit 15 specifies a resource that is estimated to be least loaded from the simulation result, and reduces the resource in the parameter set to obtain a new parameter set.

For example, in the system model shown in FIG. 2, when the CPU usage rate of the Web / AP server is the lowest, it can be estimated that even if the Web / AP server is reduced, the response time of the entire system is less affected. Therefore, a parameter set is generated by reducing the number of Web / AP servers by one.

In addition, the parameter set generation unit 15 refers to the value of the parameter set determined by the service level determination unit 14 as not to achieve SLO, and sets any one of the settings from the values of various setting parameters included in the parameter set. Generate a parameter set with large parameters.

For example, in the system model shown in FIG. 2, in the case of a parameter set in which the number of Web / AP servers and DB servers is two, if the response time does not achieve SLO, a combination of a smaller number of units, That is, SLO is not achieved with the parameter set {Web / AP server, DB server} = {1, 1}, {1, 2}, {2, 1}. Therefore, the parameter set generation unit 15 generates a parameter set in which the number of either one or both servers is increased by one or more.

Also, the parameter set generation unit 15 refers to the price of the system with the parameter set determined to be optimal by the optimal possibility determination unit 14, and generates a parameter set whose price is lower than that.

The optimum parameter set output unit 16 refers to the value of the parameter set determined not to achieve the SLO in the service level determination unit 13 and the simulation result, and newly sets a parameter set that is estimated to be highly likely to achieve the SLO. To generate. Specifically, from the simulation result, a resource that is estimated to be the most loaded is identified, and the resource is increased in the parameter set to obtain a new parameter set.

For example, in the system model shown in FIG. 2, when the CPU usage rate of the DB server is the highest, it can be estimated that the DB server is a bottleneck for processing. Generate an increased parameter set.

Also, the optimum parameter set output unit 16 outputs the parameter set that is finally determined to be optimum as the optimum parameter set. Specifically, the optimum parameter set is presented to the user by writing out the file in an arbitrary format or displaying the file on a display device such as a display.

Next, the operation of the optimum system design support apparatus 10 will be described with reference to FIG.
First, in step A1, the optimum system design support apparatus 10 sets the number of trials (numTrial), and an integer value i representing the number of trials (hereinafter referred to as trial count; i <numTrial) is set as an initial value. Set 1 As the value of numTrial, an integer value that is considered to be sufficient to cover the entire parameter set space is set based on the scale of the system described in the system model and the number of types of setting parameters. Generally, when there are many types of large-scale systems and setting parameters, it is necessary to set a large value. In addition, a large value that is impossible for the price of the system is set as the initial value of the initial value (costMin) of the lowest price of the system that is a reference for determining the optimum possibility of each parameter set.

Subsequently, a parameter set (paramSet) is randomly generated (step A2), and a predicted value of performance in the system model of the parameter set is calculated by the simulator 11 (step A3). Further, the service level determination unit 13 determines whether or not the predicted performance value has achieved SLO (step A4).

When the SLO has not been achieved (No), the parameter set generation unit 15 newly generates a parameter set that is considered highly likely to achieve the SLO. Specifically, the simulation result is referred to, a resource that is estimated to be most loaded is specified, and a parameter set (numMaxSvr) with the increased resource is newly generated (step A5).
Steps A3 to A5 are repeated until a parameter set that achieves SLO is found.

On the other hand, when the SLO has been achieved in Step A4 (Yes), or when the process of Steps A3, A4, and A5 is repeated to reach a parameter set that achieves the SLO, the value of the parameter set is set as paramSetTmp. Save (step A6).

Subsequently, the parameter set generation unit 15 newly generates a parameter set having a lower price configuration. Specifically, the simulation result is referred to, a resource that is estimated to be least loaded is specified, and a parameter set (numMinSvr) in which the resource is reduced is newly generated (step A7). This is repeated as long as the predicted performance value achieves SLO (steps A8 and A9).

If the SLO is not achieved in step A9, the price (cost) when the system is configured with the parameter set before reducing the resource by one is calculated (step A10).

Subsequently, in step A11, the optimal possibility determination unit 14 compares the values of costMin and cost. If cost is greater than costMin, it is determined that the optimum parameter set has not been updated in the current trial, the trial count i is incremented by 1 in step A12, and the processing from step A2 is repeated.

On the other hand, if cost is smaller than costMin, there is a possibility that the parameter set is an optimal parameter set. Therefore, paramSet is assigned to the optimal parameter set (optParam), and the cost value is assigned to costMin (step A13). ).

The above processing is repeated for the number of trials specified in advance (step A14). When i reaches numTrial, optParam is output and the processing is terminated (step 15).

On the other hand, if i has not reached numTrial, the value of i is increased by 1 (step A12), and the processing from step A2 is repeated. However, in the trial after the next time, when the paramSet is generated in step A3, the past simulation result is referred to, and the value of one or more setting parameters among the setting parameters included in the parameter set that does not achieve SLO. The parameter set is generated so as to have a configuration with a lower price than CostMin.

As described above, according to the present embodiment, when the SLO for the performance value that cannot be formulated is set, the performance improvement rate for each setting parameter is estimated using the simulation result. A parameter set that achieves SLO can be efficiently extracted. In addition, by using the simulation result, by excluding parameter set candidates that cannot be the minimum system configuration from the search range, it is possible to efficiently derive a parameter set that achieves SLO and has the minimum price. it can.

Embodiment 2. FIG.
FIG. 4 is a block diagram showing the configuration of the optimum system design support apparatus 40 according to Embodiment 2 of the present invention. The optimum system design support apparatus 40 according to the second embodiment includes a search range limiting unit 41 in addition to the configuration of the optimum system design support apparatus 10 according to the first embodiment.

The search range limiting unit 41 checks the upper limit and lower limit of each setting parameter set in the system model, and obtains a parameter set that is an intermediate value thereof. Then, a simulation is performed using the parameter set, and the upper limit or lower limit of the setting parameter is updated with reference to the simulation result.

That is, the search range limiting unit 41 further limits the range where the optimal parameter set exists. Specifically, when the predicted performance value in the intermediate parameter set has achieved SLO, the resource with the least load is identified from the simulation result, and the upper limit of the setting parameter corresponding to the resource is set to the parameter. Replace with set value.

On the other hand, if the SLO has not been achieved, the resource that is most heavily loaded, that is, the bottleneck of processing, is identified from the simulation result, and the lower limit of the setting parameter corresponding to the resource is set as the parameter. Replace with set value. The above processing is repeated until the parameter range is sufficiently narrowed down.

The operation of the optimum system design support apparatus 40 will be described with reference to FIG. Here, in the system model shown in FIG. 2, a case where the binary of the number of Web / AP servers and the number of DB servers is used as a setting parameter will be described as an example.

First, in step B1, an upper limit (paramMax) and a lower limit (paramMin) of the parameter set are set. These are one-dimensional arrays in which the number of elements is equal to the number of setting parameters. As these initial values, values preset in the system model can be used or arbitrarily set. In the example of FIG. 2, the number of elements in the parameter set is 2. Further, the upper limit of the number of servers is 10 and the lower limit is 1. That is, the initial values of the upper limit and the lower limit of the parameter set are paramMax = {10, 10} and parammin = {1, 1} (FIG. 6). Here, the array representing each parameter set is expressed in the order {number of Web / AP servers, number of DB servers}.

Next, a simulation is performed at the intermediate point (parammid) of the initial value set in step B1 (step B2). Here, processing is performed by rounding up the value so that the value of parammid becomes an integer value. That is, in the example of FIG. 2, parammid = {6, 6} (FIG. 6).

Next, it is determined whether or not the predicted performance value in parammid has achieved SLO (step B3). When SLO is achieved, the simulation result is referred to, and the upper limit value (paramMax [svrMin]) of the setting parameter corresponding to the resource (svrMin) that is considered to be least loaded is an intermediate value during the simulation. Replace with the value of the point (step B4). For example, when the simulation result that the load of the Web / AP server in FIG. 2 is low is obtained, the upper limit value of the number of Web / AP servers is changed from 10 to 6. That is, paramMax = {6, 10} (FIG. 7).

On the other hand, when the SLO is not achieved, the simulation result is referred to and the lower limit value (paramMin [svrMax]) of the setting parameter corresponding to the resource (svrMax) considered to be the most loaded is simulated. The intermediate value is replaced (step B5). For example, when the simulation result that the load of the DB server in FIG. 2 is high is obtained, the lower limit value of the number of DB servers is changed from 1 to 6. That is, paraMmin = {1, 6} (FIG. 8).

The above processing is repeated until it can be determined that further narrowing down is not possible for all setting parameters, that is, until the difference between the upper limit value and the lower limit value is 2 or less (step B6), and the processing from step A1. To continue. However, the parameter set generated in step A3 is limited to that included in the range narrowed down as described above.

As described above, according to this embodiment, the range in which the boundary surface between the region where the predicted performance value achieves SLO and the region where the predicted performance value does not achieve is limited, and the simulation is performed using the parameter set included in the vicinity of the boundary surface. Start. As a result, parameter set candidates can be searched in the vicinity of the boundary surface, so that an optimum parameter set can be derived more efficiently with a small number of simulations.

The optimum system design support device of the present invention is not limited to the configuration of the above embodiment, and various modifications and changes from the configuration of the above embodiment are also included in the scope of the present invention. It is.

Further, although the present invention has been described by taking the response time as an example of SLO, the present invention can be applied by replacing with other evaluation indexes such as maximum throughput.

This application claims priority based on Japanese Patent Application No. 2012-63561 filed on Mar. 21, 2012, the entire disclosure of which is incorporated herein.

The present invention has been described above with reference to the embodiments, but the present invention is not limited to the above embodiments. Various changes that can be understood by those skilled in the art can be made to the configuration and details of the present invention within the scope of the present invention.

A part or all of the above embodiment can be described as the following supplementary notes, but is not limited thereto.
(Supplementary note 1) Based on the design information of the system, a simulator for executing an operation simulation of the system and calculating a predicted value of the service level in the system, and a service level in which the predicted value of the service level is set in advance Optimal comprising: a service level determination unit that determines whether or not the target has been achieved; and a parameter set generation unit that generates a parameter set that may achieve the service level target based on the result of the simulation System design support device.

(Supplementary note 2) The optimum system design support apparatus according to supplementary note 1, wherein when the service level determination unit determines that the service level target is not achieved, the parameter set generation unit An optimal system design support device that generates a parameter set in which the amount of resources that is most loaded is increased based on the result of simulation.

(Supplementary note 3) The optimum system design support apparatus according to supplementary note 1 or 2, wherein the parameter set generation unit determines that the service level determination unit achieves the service level target. An optimal system design support device that generates a parameter set with the least loaded resources reduced based on the simulation result.

(Supplementary note 4) The optimum system design support apparatus according to any one of supplementary notes 1 to 3, wherein the parameter set generation unit is a parameter that does not achieve the service level target based on a result of the simulation. An optimal system design support device that generates a parameter set that has a larger amount of resources than the set.

(Supplementary note 5) The optimum system design support apparatus according to any one of supplementary notes 1 to 4, wherein a price of a new system with a parameter set that achieves the service level target is calculated, and a previous simulation is performed. If the price of the new system is lower than the price of the previous system with the parameter set that achieves the service level target obtained, the optimal parameter set candidate is updated with the new parameter set. An optimal system design support apparatus comprising a possibility determination unit, wherein the parameter set generation unit generates only a parameter set whose price is lower than that of the optimal parameter set candidate.

(Supplementary note 6) The optimum system design support apparatus according to any one of supplementary notes 1 to 5, wherein an optimum parameter set that is stored when a preset number of trials have been completed is presented to the user An optimum system design support device having a parameter set output unit.

(Supplementary note 7) The optimum system design support apparatus according to any one of supplementary notes 1 to 6, wherein a minimum system configuration may be obtained from the simulation result and the characteristic of an evaluation function indicating a service level of the system. An optimal system design support apparatus, comprising: a search range limiting unit that limits a range of a certain parameter set, wherein the parameter set generation unit generates only a parameter set included in the range.

(Supplementary note 8) The optimum system design support apparatus according to supplementary note 7, wherein the search range limiting unit obtains an upper limit and a lower limit of each setting parameter set in the system, and performs a simulation using an intermediate value thereof And the upper limit value of the setting parameter corresponding to the least loaded resource and the lower limit value of the setting parameter corresponding to the most loaded resource are replaced with the intermediate value based on the result of the simulation. Optimal system design support device.

(Supplementary Note 9) Based on system design information, executing a simulation of the operation of the system to calculate a predicted value of a service level in the system;
Determining whether the predicted value of the service level achieves a preset service level target;
And generating a parameter set that may achieve the service level target based on the result of the simulation.

(Appendix 10)
Based on system design information, a simulator that performs a simulation of the operation of the system and calculates a predicted value of a service level in the system;
A service level determination unit for determining whether the predicted value of the service level has achieved a preset service level target;
The program for functioning as a parameter set production | generation part which produces | generates the parameter set which may achieve the said service level target based on the result of the said simulation.

The present invention is suitable for automatically deriving an inexpensive system configuration that can guarantee to achieve SLO for performance values that cannot be formulated in various types of systems.

10, 40 Optimal system design support device, 11 simulator, 12 simulation result storage unit, 13 service level determination unit, 14 optimal possibility determination unit, 15 parameter set generation unit, 16 optimal parameter set output unit, 41 search range limiting unit

Claims

Based on system design information, a simulator that performs a simulation of the operation of the system and calculates a predicted value of a service level in the system;
A service level determination unit for determining whether the predicted value of the service level has achieved a preset service level target;
An optimal system design support apparatus, comprising: a parameter set generation unit that generates a parameter set that may achieve the service level target based on the result of the simulation.
The optimal system design support device according to claim 1,
When the service level determination unit determines that the service level target is not achieved, the parameter set generation unit increases the amount of resources that are most loaded based on the simulation result. Optimal system design support device that generates parameter sets.
The optimum system design support device according to claim 1 or 2,
When the service level determination unit determines that the service level target is achieved, the parameter set generation unit is configured to reduce the least loaded resource based on the simulation result. Optimal system design support device that generates sets.
The optimal system design support device according to any one of claims 1 to 3,
The optimal system design support device, wherein the parameter set generation unit generates a parameter set having a larger amount of resources than a parameter set that does not achieve the service level target based on the simulation result.
The optimal system design support apparatus according to any one of claims 1 to 4,
Calculate the price of the new system with the parameter set that achieves the service level target, and compare the price of the new system with the parameter set that achieves the service level target obtained in the previous simulation, and When the price of the system is cheaper, it has an optimal possibility determination unit that updates the optimal parameter set candidate with a new parameter set,
The optimal system design support device, wherein the parameter set generation unit generates only a parameter set whose price is lower than that of the optimal parameter set candidate.
The optimal system design support apparatus according to any one of claims 1 to 5,
An optimum system design support apparatus comprising an optimum parameter set output unit for presenting to a user an optimum parameter set stored at the end of a predetermined number of trials.
The optimal system design support apparatus according to any one of claims 1 to 6,
From the simulation result and the characteristic of the evaluation function indicating the service level of the system, a search range limiting unit that limits the range of the parameter set that may be the minimum system configuration,
The parameter set generation unit is an optimum system design support device that generates only a parameter set included in the range.
The optimal system design support device according to claim 7,
The search range limiting unit acquires an upper limit and a lower limit of each setting parameter set in the system, performs a simulation using an intermediate value thereof, and based on a result of the simulation, a resource that is least loaded An optimum system design support apparatus that replaces the upper limit value of the setting parameter corresponding to the above and the lower limit value of the setting parameter corresponding to the most heavily loaded resource with the intermediate value.
Executing a simulation of the operation of the system based on the design information of the system and calculating a predicted value of a service level in the system;
Determining whether the predicted value of the service level achieves a preset service level target;
And generating a parameter set that may achieve the service level target based on the result of the simulation.
Computer
Based on system design information, a simulator that performs a simulation of the operation of the system and calculates a predicted value of a service level in the system;
A service level determination unit for determining whether the predicted value of the service level has achieved a preset service level target;
The program for functioning as a parameter set production | generation part which produces | generates the parameter set which may achieve the said service level target based on the result of the said simulation.