JP2014132419A - Performance prediction device of virtualization system, performance prediction method and computer program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a performance prediction device or the like for easily predicting the performance of a virtualization system even if a virtual machine operating in one physical machine is a single machine or a plurality of machines.SOLUTION: A performance prediction device 1 of a virtualization system (VS) has a first PPM generation part 101 for generating a performance prediction model (PPM) of an AP layer on the basis of information about an AP, operating in a virtual machine executed by a VS comprising an application (AP) layer and a hardware (HW) layer, a second PPM generation part 102 for generating a PPM of the HW layer on the basis of a result obtained by measuring the VS, and a performance prediction model synthesis part 103 for generating a PPM of the VS with a dependence relation between the AP layer and the HW layer reflected thereon by synthesis processing using an output value in the PPM of one layer as an input value in the PPM of the other layer with the PPM of the AP layer and the PPM of the HW layer as objects.

Description

  The present invention relates to a technical field for predicting the performance of a virtualized system (virtualized environment) by using a computer.

  Computer virtualization technology allows multiple virtual machines to run on a single physical machine.

  However, when using the virtualization technology, overhead and resource contention occur in the hardware resources of the physical machine. Here, the hardware resources are, for example, a CPU (Central Processing Unit), a storage device, a memory, and the like that the physical machine has. The resource contention is, for example, hardware resource contention caused by simultaneous execution of a plurality of transactions, or software contention caused when the virtual machine monitor simultaneously emulates virtual hardware of a plurality of virtual machines.

  Regarding such overhead and resource contention, Non-Patent Document 1 points out that the CPU performance deteriorates every time one virtual machine operating on a physical machine increases, and the performance of the system in which the deterioration is executed in the virtual machine. It has an influence on.

  In general, if the number of virtual machines operating in a physical machine increases, it is assumed that the operation of a virtual machine monitor that emulates the operation of the hardware constituting the physical machine becomes complicated. If the operation of the virtual machine monitor becomes complicated, it is assumed that the performance of the virtual machine monitor also deteriorates.

  Considering the possibility of these performance degradations, overhead and resource contention are taken into account in the performance prediction of a virtualized system, and the performance can be easily predicted even when multiple virtual machines are operating. is necessary. Here, the virtualized system is a system having an environment in which a single physical machine or a plurality of virtual machines can operate on one physical machine by a virtualization technique.

  As a system performance prediction method, there is a method using a performance prediction model. Here, the performance prediction model is a model that predicts the performance of the target system (virtualized system that is the target of performance prediction) according to the input to the system (for example, as a typical example, a queue or regression analysis). ). The performance prediction method using such a performance prediction model includes mainly two types of methods described below.

  First, a method for generating a performance prediction model that imitates the operation of the system will be described. Hereinafter, this model generation method is referred to as a “white box approach” in the present application. In the white box approach, in order to generate a model that can mimic the behavior of the target system, details of the behavior of the target system (for example, design information, etc.) must be clear when generating the performance prediction model. . That is, according to the white box approach, if the operation of the target system is clear, the performance prediction model can be easily generated even when the target system is replaced with another system.

  Examples of performing performance prediction of a target system using a performance prediction model generated by a white-box approach having such characteristics include Non-Patent Document 2 and Patent Documents 2 to 5.

  In Non-Patent Document 2, a performance prediction model is generated based on queuing theory. More specifically, in Non-Patent Document 2, when creating a performance prediction model for a virtual machine monitor whose operation content is complicated, an overhead value calculated by comparing the operation of a physical machine between the virtualized system and the non-virtualized system Is used. Thereby, in Non-Patent Document 2, it is possible to create a performance prediction model that also considers overhead.

  Patent Document 2 discloses a method for generating a queue model of a computer system. More specifically, in Patent Document 2, when generating a queuing model, the response time for each workload is measured, and the average service for each class is important based on the measured response time and the queuing model. Estimate the degree. Furthermore, in Patent Document 2, a queue for the entire system is generated by using the estimated average service demand and workload characteristics.

  In Patent Document 3, a storage performance prediction model is generated by using a white box approach using model information for each component of a storage virtualization apparatus.

  In Patent Document 4, a performance prediction model of a virtualization system is generated by using a white box approach using design information of each component. In Patent Document 4, an overhead is measured for each type of virtual machine monitor stored in a database, and the measured overhead is reflected in a performance calculation formula to obtain a performance prediction model in consideration of the overhead. .

  In Patent Document 5, a system having hierarchical competitive resources is divided into an upper layer (software part) and a lower layer (hardware part), and a performance prediction model generated for each layer is used as the performance of the entire system. Treat as a prediction result. At that time, in Patent Document 5, a white box approach is used as a method for generating a performance prediction model for both the upper layer and the lower layer.

  Next, a method for generating a performance prediction model by using statistical processing and machine learning based on an input value to the system and an output value corresponding to the input value will be described. Hereinafter, in the present application, such a model generation method is referred to as a “black box approach”. In the black box approach, a model based on actual measurement values is generated, so that it is not necessary for an operator or the like to clearly recognize the operation of the target system. That is, according to the black box approach, it is possible to generate a performance prediction model of the target system even if the operation of the target system is complicated.

  Non-Patent Document 3 and Patent Document 6 are examples of performing performance prediction using a performance prediction model generated by a black box approach.

  In Non-Patent Document 3, a performance prediction model is generated by using a measured value in a state where a single virtual machine is operating in a virtual machine monitor and a linear model. That is, in Non-Patent Document 3, when a plurality of virtual machines operate on a virtual machine monitor, the performance prediction model is a linear sum of linear models in a state where a single virtual machine is operating.

  In Patent Document 6, data representing overhead based on load data and measured values is reproduced by a load reproduction function of a host OS (Operating System) to reproduce a virtualization system in which a plurality of virtual machines operate. . And in patent document 6, the performance evaluation of the said virtualization system is performed by utilizing the measured value obtained by reproduction | regeneration of the evaluation object application of guest OS in the condition where the reproduced some said virtual machine operate | moves.

  Patent Document 7 is an approach that uses both of the two types of performance prediction model generation methods described above. Patent Document 7 discloses a method for predicting the performance of a target system composed of a plurality of software components. Specifically, in Patent Document 7, the system resource usage obtained by actual measurement during the operation of each software component is stored in a database as performance data. Next, Patent Document 7 extracts the system resource usage of each software component constituting the target system from the accumulated performance data, and is necessary for transaction processing based on the sum of the extracted results. Predict system resource usage. Patent Document 7 predicts the performance of the entire target system based on the predicted resource usage of transaction processing and queuing theory.

JP 2010-009160 A JP 2011-086295 A JP 2006-146354 A JP 2012-146015 A JP 2000-029753 A JP 2010-128866 A JP 2004-220453 A

Ajishiro Ikudai, Tanaka Yasuhiro, Evaluation of resource management overhead in virtual machine environment, Information Processing Society of Japan (EVA), 17-22, 2006 Performance Models For Virtualized Applications, Febricio Benevento, Cesar Fernandez, Matheus Santos, Virgilio Almeida, Jussara Almeida. (John) Janakiraman, and Jose Renato Santos, ISPA, 2006 International Society for the Performing Arts 2006 Workshops, Lecture Notes in Computer 39, 43/39 Untangling Mixed Informantion To Calibrate Resource Utilization In Virtual Machines, Lei Lu, Hui Zhang, Guofei Jiang, Haifeng Chen, Kenji Yoshihira, Evgenia Smirni, International Conference on Autonomic Computing 2011, Proceeding of the 8th ACM International Conference on Autonomic computing, pp151-160ICAC , 2011

  The technique disclosed in Patent Document 4 which is a performance prediction method for a virtualized system stores overhead measured for each type of virtual machine monitor in a database, and reflects these overheads in a performance calculation formula. Thereby, the method disclosed in Patent Document 4 realizes a performance prediction method considering such overhead.

  Patent Document 1 discloses a technique that uses a performance prediction function and a performance prediction model including overhead in a performance prediction value calculation device of a virtualization system.

  However, Patent Documents 1 and 4 only disclose a performance prediction method in which a virtual machine operates alone, and does not consider the occurrence of resource contention due to the operation of a plurality of virtual machines.

  On the other hand, in Patent Document 5, a situation in which a plurality of virtual machines operate in the host OS is created, and performance prediction is performed by actually measuring an evaluation target application operating in the guest OS in the situation. In Patent Document 5, overhead in a situation where a plurality of virtual machines operate is considered. However, in Patent Document 5, since performance prediction is performed based on actual measurement values, when the specification of the virtual machine is changed, the application to be evaluated must be actually measured every time the change is made. It's a problem.

  Further, in the method of predicting the performance of the virtual environment using a model, as described above with respect to Non-Patent Document 1 and Patent Documents 3 and 4, a white-box approach is performed based on details regarding the components of the target system. Generates a performance prediction model using The details of the operation of the target system need to be clear for model generation by such a white box approach. In particular, in the white box approach, the operation of the virtual machine monitor in a virtual environment is complicated, and therefore, the model generation by the white box approach including a virtual machine requires complicated work.

  Accordingly, the present invention mainly provides a performance prediction apparatus and the like that can easily predict the performance of a virtualization system even when the number of virtual machines operating on one physical machine is single or multiple. And

  In order to achieve the above object, a performance prediction apparatus for a virtualization system according to the present invention comprises the following configuration.

That is, the performance prediction device of the virtualization system according to the present invention is
A first performance prediction model generation unit that generates a performance prediction model of the application layer based on information related to the application that operates in a virtual machine executed in a virtualization system including an application layer and a hardware layer When,
A second performance prediction model generation unit that generates a performance prediction model of the hardware layer based on the measurement result of the virtualization system;
By combining the performance prediction model of the application layer and the performance prediction model of the hardware layer, the output value in the performance prediction model of one layer is used as the input value in the performance prediction model of the other layer. , A performance prediction model synthesis unit that generates a performance prediction model of the virtualization system in which the dependency relationship between the two layers is reflected;
It is characterized by providing.

  In order to achieve the object, the virtualization system performance prediction method according to the present invention is characterized by having the following configuration.

That is, the performance prediction method of the virtualization system according to the present invention is:
Generate a performance prediction model of the application layer based on information about the application that operates in a virtual machine executed in a virtualization system composed of an application layer and a hardware layer,
Based on the measurement result of the virtualization system, generate a performance prediction model of the hardware layer,
Targeting the performance prediction model of the application layer and the performance prediction model of the hardware layer, by using the output value in the performance prediction model of one layer as an input value in the performance prediction model of the other layer, A performance prediction model of the virtual system reflecting the dependency relationship between the two layers is generated.

  Further, the same object is achieved by a computer program for realizing a performance prediction apparatus for a virtualization system having the above-described configuration and a corresponding method by a computer, and a computer-readable storage medium storing the computer program. Is also achieved.

  According to the present invention described above, the performance of the virtualization system can be easily predicted even when the number of virtual machines operating on one physical machine is single or multiple.

It is a block diagram which shows the function structure of the performance prediction apparatus of the virtualization system based on the 1st Embodiment of this invention. It is a figure which illustrates notionally the precondition of the performance prediction apparatus of the virtualization system based on the 1st Embodiment of this invention. It is a block diagram explaining the synthetic | combination processing method by the performance prediction model synthetic | combination part 303 of the virtual environment shown in FIG. FIG. 3 is a diagram conceptually illustrating a performance prediction process (synthesis process) performed by a performance prediction model synthesis unit 103 in a virtual environment in the first embodiment. It is a figure which illustrates the performance parameter | index which the performance prediction model 302 of a hardware layer outputs in 1st Embodiment. FIG. 6 is a block diagram showing a functional configuration of the performance prediction apparatus according to the second embodiment of the present invention. It is a flowchart which shows the outline | summary of the performance prediction process by the performance prediction apparatus 10 which concerns on the 2nd Embodiment of this invention. 10 is a flowchart illustrating a processing procedure for generating a performance prediction model 302 by a hardware layer performance prediction model generation unit 102 according to the second embodiment. It is explanatory drawing which represented notionally the synthesis | combining method of the performance prediction model 303 of a virtual environment in the performance prediction model synthetic | combination part 103 of the virtual environment shown in FIG. 10 is a flowchart illustrating a procedure of a synthesis process (performance prediction) performed by a performance prediction model synthesis unit 103 in a virtual environment in the second embodiment. It is a figure which illustrates notionally the precondition of the performance prediction apparatus of the virtualization system which concerns on the modification of the 1st and 2nd embodiment of this invention. It is a figure which illustrates illustratively the hardware constitutions of the computer (information processing apparatus) which can implement | achieve each embodiment and its modification of this invention.

  Next, embodiments of the present invention will be described in detail with reference to the drawings.

<First Embodiment>
FIG. 1 is a block diagram showing a functional configuration of a performance prediction apparatus for a virtualization system according to the first embodiment of the present invention.

  In FIG. 1, the performance prediction device 1 is connected to a virtualization system 2 so as to be communicable. The virtualization system 2 is a target (hereinafter also referred to as “target system” or “virtualized environment”) for which the performance prediction apparatus 1 should predict performance. As a connection method between the performance prediction apparatus 1 and the virtualization system 2, for example, general communication means such as the Internet or a LAN (Local Area Network) can be employed. Therefore, in this embodiment, detailed description regarding the communication means is omitted.

  In the present embodiment, the performance prediction device 1 of the virtualization system 2 includes an application layer performance prediction model generation unit (first performance prediction model generation unit) 101 and a hardware layer performance prediction model generation unit (second A performance prediction model generation unit) 102 and a performance prediction model synthesis unit 103 of a virtual environment.

  That is, the performance prediction model generation unit 101 of the application layer is based on information about the application that operates in the virtual machine executed by the virtualization system 2 including the application layer and the hardware layer. A performance prediction model (301) is generated.

  The performance prediction model generation unit 102 of the hardware layer generates a performance prediction model (302) of the hardware layer based on the measurement result of the virtualization system 2.

  The performance prediction model synthesis unit 103 of the virtual environment outputs the output value in the performance prediction model of one layer for the performance prediction model (301) of the application layer and the performance prediction model (302) of the hardware layer. Is used as an input value in the performance prediction model of the other layer to generate a performance prediction model of the virtualization system 2 in which the dependency relationship between the two layers is reflected.

  Hereinafter, a more specific configuration example of the performance prediction apparatus 1 having the above-described configuration will be described in detail.

  The performance prediction apparatus 1 can predict the performance of each layer by generating a model reflecting the operation characteristics of the two layers, that is, the application layer and the hardware layer. For generation of the model in the present embodiment, at least a second method for constructing a model based on an actual measurement value of a performance prediction target system and a first method for creating a model that imitates the details of the operation of the system are used. I will do it. Specifically, in the present embodiment, for example, queuing theory or regression analysis method can be employed for generating the model. As these model generation methods themselves, a general method can be adopted at present, and a detailed description in this embodiment will be omitted. The model as the final product generated by the performance prediction apparatus 1 is referred to as a “performance prediction model” (the same applies to the following embodiments).

  A hardware configuration (hardware resource) capable of realizing the performance prediction apparatus 1 will be described later with reference to FIG.

  Next, preconditions in the present embodiment will be described with reference to FIG. FIG. 2 is a diagram conceptually illustrating the preconditions of the performance prediction apparatus for the virtualization system according to the first embodiment of the present invention. In this embodiment, when the performance prediction device 1 predicts the performance of the virtualization system (target system) 2, it is defined that the configuration of the virtualization system has a hierarchical configuration.

  In FIG. 2, the virtualization system 2 is constructed as a monolithic hypervisor type as an example. However, the configuration method of the virtualization system 2 includes a host OS type in which the virtual machine monitor operates in a host OS (Operating System), and a microkernel having a minimum function for the virtual machine monitor to provide a virtual environment. -A hypervisor type may be used. That is, the configuration method of the virtualization system 2 can be adopted by any method as long as it can realize an environment in which a virtual machine can operate in one physical machine.

  The virtualization system 2 in the first embodiment has three layers: a virtual machine 201, a virtual machine monitor (hypervisor) 202, and hardware of a physical machine 203, as in a general virtualization system. As shown in FIG. 2, there are n virtual machines 201 (where n is a natural number including 0), and it is not limited whether there are a plurality of virtual machines.

Each virtual machine 201 can provide a guest OS environment and execute one or more arbitrary applications (AP) as illustrated in FIG. In this embodiment, the virtualization system (target system) 2 having the three-layer structure is defined as two layers (two layers) as follows. That is,
An application layer 204 configured by the virtual machine 201,
A hardware layer 205 composed of a virtual machine monitor 202 and physical machine hardware 203.

  In other words, when predicting the performance of the virtualization system 2 having such a three-layer structure, the performance prediction apparatus 1 handles the two layers. Then, the performance prediction apparatus 1 predicts the performance of the virtualization system 2 on the assumption of the definition. Here, as a method of definition in the performance prediction apparatus 1, for example, the virtual machine 201 is registered in association with (associated with) the application layer 204, and the virtual machine monitor 202 and the physical machine hardware 203 are connected to the hardware. Registration may be performed in association with the layer 205.

  FIG. 3 is a block diagram illustrating a synthesis processing method performed by the performance prediction model synthesis unit 303 in the virtual environment illustrated in FIG. In this embodiment, the performance prediction model synthesis unit 103 uses a performance prediction model 301 of the application layer (204) and a performance prediction model 302 of the hardware layer (205) as shown in FIG. The performance prediction model 303 is synthesized.

  More specifically, the performance prediction model generation unit 101 of the application layer is based on the information about the application that operates in the virtual machine 201 executed in the virtualization system 2 in accordance with an operation of an operator or the like. A layer performance prediction model 301 is generated. In the present embodiment, the performance prediction model generation unit 101 generates the performance prediction model 301 by the white box approach (first method) described above in the “background art” column.

  That is, the model generation by the white box approach is a model generation method that imitates the operation of an arbitrary system based on design information of the arbitrary system. Specifically, for example, queuing theory or Petri net can be used as a model generation method based on the white box approach. However, as long as the same approach method can be realized, the method for generating the performance prediction model of the application layer in the present embodiment is not limited. For example, a method similar to the method disclosed in Patent Document 4 that generates a queue using design information of each component of the target system may be adopted.

  On the other hand, the performance prediction model generation unit 102 of the hardware layer uses the black box approach (second method) described above in the “Background Technology” column based on the actual measurement value obtained by the measurement of the virtualization system 2. A hardware layer performance prediction model 302 is generated. In other words, in this embodiment, model generation by a black box approach is a method of generating a model by statistical processing or machine learning based on an input value to an arbitrary system and an output value corresponding to the input value. is there. Specifically, for example, a regression analysis, a neural network, or a genetic algorithm can be used as a model generation method based on such a black box approach. However, the method for generating the performance prediction model of the hardware layer in the present embodiment is not limited as long as the same approach method can be realized. For example, a model generation method disclosed in Non-Patent Document 3 using an actual measurement value in a state where a single virtual machine is operating in a virtual machine monitor and a linear model may be employed.

  Next, the performance prediction model synthesis unit 103 of the virtual environment will be described.

  Generally, in a virtualized system (virtualized environment), a virtual machine executed in an application layer executes a desired application (processing) by using hardware resources constituting the hardware layer. Therefore, there is a dependency between the application layer and the hardware layer. In other words, in a state where the application layer and the hardware layer are separately provided, it is impossible to accurately predict the performance of the virtualization system, which is a target for which the performance is to be predicted.

  On the other hand, the performance prediction model composition unit 103 of the virtual environment according to the present embodiment can perform performance prediction in consideration of the dependency relationship existing between the application layer 204 and the hardware layer 205. That is, the performance prediction model synthesis unit 103 is generated by the performance prediction model 301 of the application layer and the performance prediction model 102 of the hardware layer, which is generated by the performance prediction model generation unit 101 of the application layer. By generating the performance prediction model 303 of the virtualization system 2 using the performance prediction model 302 of the hardware layer, it is possible to predict the performance of the entire virtualization system. The procedure of this performance prediction will be specifically described below with reference to FIG.

  FIG. 4 is a diagram conceptually illustrating the performance prediction process (synthesis process) performed by the performance prediction model synthesis unit 103 in the virtual environment in the first embodiment.

  The performance prediction model composition unit 103 of the virtual environment inputs the resource request 808 for the hardware resource (physical machine) calculated by the performance prediction model 301 of the application layer to the performance prediction model 302 of the hardware layer. Here, the resource request 808 for the hardware resource includes information regarding how the performance prediction model 301 of the application layer uses the hardware resource.

  In response to the input of the resource request 808, the hardware layer performance prediction model 302 can predict the performance of the entire target system 2 (a performance index 809 (FIG. 9) described later). Can be provided to an operator or the like. Here, the performance index is, for example, a usage rate of hardware resources, TAT (Turn Around Time), throughput, and the like. Such a performance index is provided to an operator or the like as exemplified in FIG. When the processing according to the input resource request 808 is completed, the performance prediction model composition unit 103 of the virtual environment inputs a processing completion notification 810 regarding the resource request to the performance prediction model 301 of the application layer.

  FIG. 5 is a diagram illustrating a performance index output by the hardware layer performance prediction model 302 in the first embodiment. More specifically, in the example illustrated in FIG. 5, “CPU usage rate (%) with respect to the maximum number of threads when the maximum number of threads is increased (100, 300,..., 1900) for the same system. ”,“ Throughput (tps: number of transactions per second) ”, and“ TAT (ms: milliseconds) ”.

  Further, as shown in FIG. 4, the performance prediction model synthesis unit 103 of the virtual environment inputs a process completion notification 810 regarding the resource request in the hardware layer performance prediction model 302 to the performance prediction model 301 of the application layer. To do.

  That is, according to the performance prediction model synthesis unit 103 of the virtual environment according to the present embodiment, the output in the performance prediction model of one layer is targeted for the performance prediction model of the application layer and the performance prediction model of the hardware layer. The value is used as an input value in the performance prediction model of the other layer. According to the performance prediction model 303 of the target system generated by such a processing configuration, performance prediction considering the dependency relationship between the two layers is realized. In the present embodiment, the performance prediction process using the dependency relationship between the two layers is also referred to as synthesis (synthesis process).

  The effect of this embodiment will be described below. In the present embodiment, the white box approach (first method) is used to generate the performance prediction model of the application layer 204 in which the details of the operation of the virtual machine 201 are clear. In other words, the performance prediction model generated using the white box approach generally has a clear system configuration for each virtual machine. Thus, in the present embodiment, the performance prediction model generation unit 101 of the application layer simply adds a virtual machine to be added even when a plurality of virtual machines 201 need to operate in the application layer 204. The performance prediction model 301 of the application layer can be easily generated. Therefore, according to the present embodiment, the application layer 204 does not require the trouble (cost) of re-measurement according to the number of virtual machines as in the case where the black box approach (second method) is adopted. The virtual machine 201 can be added easily.

  On the other hand, in this embodiment, a black box approach (second method) is used to generate the performance prediction model 302 of the hardware layer. In other words, when the performance prediction model related to the hardware layer 205 is set as the generation specification by the white box approach, the specifications of each component (virtual machine monitor 202, physical machine 203, etc.) having complicated operations are clarified prior to generation of the model. Must be set to On the other hand, the performance prediction model generated using the black box approach generates the model using the actual measurement value, so even if the operation is unclear for each component that has a complex operation, the hardware layer The performance prediction model 302 can be generated. That is, unlike the white box approach, the black box approach does not require a clear understanding of the system operation, and thus the performance prediction model 302 can be easily generated. Therefore, according to the performance prediction model generation unit 102 of the hardware layer, the performance exerted by the virtual machine monitor 202 can be achieved without explicitly describing details such as emulation in the virtual machine monitor 202 and resource competition in the hardware 203 of the physical machine. Thus, the performance prediction model 302 in which the influence of the above is correctly considered can be generated.

  Then, according to the performance prediction model synthesis unit 103 of the virtual environment according to the present embodiment, each of these is performed by the synthesis process using the performance prediction model 301 of the application layer and the performance prediction model 302 of the hardware layer. It is possible to generate a performance prediction model 303 of a virtual environment in which inter-layer dependencies are taken into consideration. That is, according to the performance prediction model 303, the performance of the entire virtualization system (target system) 2 can be accurately predicted.

  Thus, according to the present embodiment, a performance prediction method that takes into consideration the characteristics of the operation of the virtualization system 2 in the application layer 204 and the hardware layer 205 is employed. Therefore, according to the performance prediction apparatus 1 according to the present embodiment, the performance of the virtualization system 2 can be easily and accurately obtained even when the operator or the like does not clearly understand the details of overhead and resource competition. It is possible to predict. The overhead in the embodiment is assumed to be, for example, an overhead caused by the hardware 203 of the physical machine or an overhead caused by emulation in the virtual machine monitor 202. The resource conflict is assumed to be, for example, a hardware resource conflict caused by simultaneous execution of a plurality of transactions, or a software conflict caused by the virtual machine monitor 202 emulating virtual hardware of a plurality of virtual machines at the same time. .

<Second Embodiment>
Next, a second embodiment based on the performance prediction device 1 according to the first embodiment described above will be described. In the following description, the characteristic part according to the present embodiment will be mainly described. At this time, the same reference numerals are assigned to the same configurations as those in the first embodiment described above, and the duplicate description is omitted. Also in this embodiment, the performance prediction apparatus 10 is assumed to configure the two layers (that is, the application layer 204 and the hardware layer 205) described with reference to FIG. 2 in the first embodiment. To do.

  A performance prediction apparatus 10 as a second embodiment will be described with reference to FIG. FIG. 6 is a block diagram showing a functional configuration of the performance prediction apparatus according to the second embodiment of the present invention.

  The performance prediction apparatus 10 according to the present embodiment includes an application layer performance prediction model identification unit 104, a hardware layer performance prediction model generation unit 102, a virtual environment performance prediction model synthesis unit 103, and an input content specification unit 401. And an information collecting unit 402 and a storage unit 3.

  Note that the storage unit 3 may be provided in a database or the like outside the performance prediction apparatus 1. In this case, the performance prediction model specifying unit 104 of the application layer may refer to such a database. Also in the present embodiment, the performance prediction apparatus 10 may include the application layer performance prediction model generation unit 101 described above in the first embodiment together with the application layer performance prediction model specifying unit 104. Good.

  Hereinafter, each part of the performance prediction apparatus 10 which has such an apparatus structure is demonstrated in detail. A hardware configuration (hardware resource) capable of realizing the performance prediction apparatus 10 will be described later with reference to FIG.

  Unlike the first embodiment, the performance prediction model specifying unit 104 in the application layer obtains a specific performance prediction model (301) from the storage unit 3. That is, the storage unit 3 stores the performance prediction model group 110 of the application layer 204 generated using a white box approach (first method). The performance prediction model specifying unit 104 refers to the storage unit 3 based on information related to an application operating on the virtual machine 201 of the virtualization system 2 in accordance with an operation of an operator or the like. Thereby, the performance prediction model specification part 104 can specify the specific performance prediction model 301 linked | related with the information regarding the said application from the performance prediction model group 110 which concerns.

  In the present embodiment, as an example, the device configuration includes the storage unit 3 inside the performance prediction device 10. However, the present invention described by taking this embodiment as an example is not limited to such a device configuration, and as in the first embodiment, a device that can refer to a storage unit (database) provided outside the own device. It may be a configuration.

  Similar to the first embodiment, the performance prediction model generation unit 302 of the hardware layer 205 generates the performance prediction model 302 of the hardware layer 205 using a black box approach (second method).

  The input content specifying unit 401 is used by an operator (user) when measuring an actual measurement value related to the virtualization system (target system) 2 to be used for generation of the performance prediction model 302 by the performance prediction model generation unit 102 of the hardware layer. 20 has a function as a user interface (UI, man-machine interface) capable of designating input contents.

  In the present embodiment, the individual performance prediction models that form the application layer performance prediction model group 110 are the white boxes described above in the “Background Technology” column, as with the application layer performance prediction model 301 according to the first embodiment. Generated by a dynamic approach (first method). That is, the one or more performance prediction models 301 constituting the performance prediction model group 110 are, for example, the performance prediction apparatus 10 having a white-box approach prior to the synthesis processing by the performance prediction model synthesis unit 103 in the virtual environment. What is necessary is just to obtain | require by the general procedure using.

  The storage unit 3 is, for example, a hard disk drive (HDD) that is a computer-readable storage medium, and is generated using design information of an application that operates on the virtual machine 201 of the virtualization system 2 The performance prediction model group 110 thus obtained can be accumulated. The performance prediction model 110 group is a candidate group that can be identified as the performance prediction model 301 of the application layer, and needs to have a configuration imitating the behavior of an application operating in the virtual machine 201. In the present embodiment, the performance prediction model group 110 may employ a performance prediction model generated using a queue, a Petri net, or the like according to design information of an application running on the virtual machine 201. However, when the performance prediction model is generated, the generation method is not limited as long as it is a performance prediction model in which the output is determined by a predetermined procedure by analysis calculation or simulation for the input.

  The application layer performance prediction model specifying unit 104 acquires the specific performance prediction model 301 from the application layer performance prediction model group 110 stored in the storage unit 3. Specifically, the performance prediction model specifying unit 104 specifies a specific performance prediction model corresponding to the input content in response to the operator 20 specifying the input content such as information on a desired application operating in the virtual machine 201. 301 is obtained from the performance prediction model group 110 of the application layer. As an input method to the performance prediction model specifying unit 104, an instruction from an external device such as a server via a LAN or the Internet may be used in addition to the operator himself. For this reason, in this embodiment, it is expressed as an operator 20 as shown in FIG.

  Further, the performance prediction model specifying unit 104 of the application layer converts the specific performance prediction model 301 for the application layer 204 corresponding to the input content instructed by the operator 20 or the like into the performance prediction model combining unit 103 of the virtual environment. It also has a function of inputting (providing) to. Here, the input content is application information in each virtual machine 201 executed in the application layer 204. The specific performance prediction model 301 corresponding to the input content represents the system content of each virtual machine 201 that is a target for which performance is to be predicted. Since this system content is different for each virtual machine 201, the corresponding performance prediction model 301 is also different for each system (for each virtual machine 201). For example, if three virtual machines 201 operate in the application layer 204, the performance prediction model 301 that matches the application information related to the virtual machines 201 corresponding to the three virtual machines 201 is stored in the performance prediction model group 110 (storage unit) of the application layer. It is necessary to take out from 3).

  In the input content designation unit 401, an input operation by the operator 20 or the like is related to a setting value related to measurement of the actual measurement value (for example, usage rate, throughput, TAT, etc. of each resource) of the virtualization system 2 or a measurement execution command. Input content is specified. The input operation method for the input content specifying unit 401 is not limited to the input by the operator itself, as in the input method for the performance prediction model specifying unit 104 described above, and may be an instruction from an external device. The input content specifying unit 401 has a function of inputting (providing) the input content specified as described above to the virtualization system 2.

  For example, the performance prediction apparatus 10 may measure the actual measurement values related to the virtualization system 2 prior to the synthesis processing by the performance prediction model synthesis unit 103 in the virtual environment. The content of the measurement includes a benchmark test using benchmark software in the virtual machine 201. For example, the benchmark software includes software (computer program) that can load various hardware resources related to the target system 2, such as a CPU, a storage device, and a communication network. In the present embodiment, any program can be adopted as the benchmark software as long as the program can reproduce the load status of a specific hardware resource.

  In addition, since the measurement procedure itself of the actual measurement value regarding the virtualization system 2 can employ | adopt a general procedure, detailed description in this embodiment is abbreviate | omitted.

  The information collection unit 402 can collect actual measurement values measured for the target system 2. Here, the actual measurement value is a pair of an input value in a benchmark test related to the target system 2 executed in accordance with the input content from the input content specifying unit 401 and an output value corresponding to the input value. Information. Furthermore, the information collection unit 402 can input (provide) the collected actual measurement values to the performance prediction model 302 of the hardware layer 205.

  More specifically, the information collection unit 402 collects an input value to the target system 2 and an output value corresponding to the input value as the actual measurement values described above. The input value is a resource request to each hardware resource (information including information on how to use the hardware resource), and corresponds to the resource request 808 (FIG. 3) described in the first embodiment. To do. The information collection unit 402 may obtain the input value from the target system 2 or the input content designation unit 401. The output value corresponding to the input value is the usage rate or TAT of each hardware resource.

  Then, the information collection unit 402 stores the collected measured values (that is, the input values to the target system 2 and the output values corresponding to the input values) in a computer-readable storage medium such as an HDD (for example, the storage unit 3). ) Is stored. In addition, the information collection unit 402 has a function of outputting the actual measurement value to the performance prediction model generation unit 102 of the hardware layer in accordance with the operation of the operator 20 or the like.

  The hardware layer performance prediction model generation unit 102 uses a black box approach based on information input from the information collection unit 402 (the above-described actual measurement value) triggered by an operation of an operator 20 or the like, for example. 205 performance prediction models 302 are generated.

  And the performance prediction model production | generation part 102 also has a function which inputs the produced | generated performance prediction model 302 to the performance prediction model synthetic | combination part 103 of a virtual environment triggered by operation of operators 20 etc., for example.

  The performance prediction model synthesis unit 103 of the virtual environment includes a performance prediction model 301 specified by the performance prediction model specification unit 104 of the application layer and a performance prediction model generated by the performance prediction model generation unit 102 of the hardware layer. The performance prediction model 303 of the target system 2 is created by synthesizing with 302. Also in the present embodiment, the method for generating (combining) the performance prediction model 303 of the target system 2 is the same as that in the first embodiment. The performance prediction model synthesis unit 103 in the virtual environment may execute the synthesis process triggered by an operation of an operator or the like, for example.

  That is, according to the performance prediction apparatus 10 according to the present embodiment, the content of the request to the virtualization system (target system) 2 by using the performance prediction model 303 generated by the performance prediction model synthesis unit 103 of the virtual environment. Based on the above, it is possible to calculate a performance index (performance predicted value: usage rate of each resource, TAT, throughput, etc.) as a performance prediction result of the entire target system 2. The operator 20 can predict the performance of the target system 2 by referring to the performance index, for example, via a display or the like (not shown: equivalent to an input / output user interface 13 shown in FIG. 12 described later). .

(Performance prediction process)
Next, the performance prediction processing performed by the performance prediction apparatus 10 in order to realize the above-described series of performance prediction operations will be described more specifically with reference to FIGS.

  FIG. 7 is a flowchart showing an outline of the performance prediction process by the performance prediction apparatus 10 according to the second embodiment of the present invention.

  First, in step S101, the performance prediction model generation unit 102 of the hardware layer generates a performance prediction model 302 of the hardware layer 205 in response to an operation of an operator 20 or the like, for example, and generates the generated performance prediction model 302 as a virtual Is input to the performance prediction model composition unit 103 of the computer environment. Details of step S101 will be described later with reference to FIG.

  Next, in step S102, the performance prediction model specifying unit 101 of the application layer uses, as an input value (search key), information related to an application that operates on the virtual machine 201 of the virtualization system 2 in response to an operation of the operator 20 or the like. The specific performance prediction model 301 corresponding to the input value is acquired by referring to the performance prediction model group 110 of the application layer stored in the storage unit 3.

  That is, step S102 is a step of acquiring a specific performance prediction model (application layer performance prediction model) 301 that matches the application information related to the virtual machine 201 executed in the application layer 204 from the application layer performance prediction model group 110. It is. In other words, in step S102, the performance prediction model of the application layer 204 generated in advance using information about a desired application that the operator 20 desires to operate in the virtual machine 201 and application information having the same content is obtained. In this step, the operator 20 selects the performance prediction model group 110 by, for example, a human operation. The selected performance prediction model is a specific performance prediction model 301 to be used for performance prediction executed in step S103 described later.

  In step S103, the performance prediction model composition unit 103 of the virtual environment receives the performance prediction model 302 of the hardware layer 205 input in step S101 and the application layer specified in step S102. The overall performance of the virtualization system 2 is predicted by executing the above-described synthesis process using the performance prediction model 301. The performance prediction value obtained as a result of this performance prediction is the output of the performance prediction model synthesis unit 103 in the virtual environment. Step S103 is executed, for example, triggered by the operation of the operator 21 or the like. Details of step S103 will be described later with reference to FIG.

  FIG. 8 is a flowchart showing a processing procedure in which the performance prediction model generation unit 102 of the hardware layer generates the performance prediction model 302 in the second embodiment, and shows details of step S101 described above.

  First, in step S201, the performance prediction model generation unit 102 of the hardware layer instructs the input content specification unit 401 to perform measurement related to the target system 2 in response to an operation of the operator 20 or the like. At this time, the input content designating unit 401 receives the setting value related to the measurement content of the actual measurement value (measurement value) related to the target system 2, the command (request) instructing the execution of the measurement, etc. Set to System 2. The virtualization system 2 starts measurement of the virtualization system 2 in response to the input of the information specified in step S201 by the input content specifying unit 401.

  Next, in step S202, the information collection unit 402 collects actual measurement values (measurement values) output according to the measurement of the virtualization system 2 started in step S201. The actual measurement value is information representing the state of the hardware layer 205 such as the usage rate, throughput, and TAT of each resource. The information collection unit 402 inputs the collected actual measurement values (the input value input in step S201 and the output value corresponding to the input value) to the performance prediction model generation unit 102 of the hardware layer.

  Then, in step S203, the hardware layer performance prediction model generation unit 302 uses a measured value input by the information collection unit 402 in response to an operation of the operator 20 or the like, for example, by a black box approach. A performance prediction model 302 of the layer 205 is generated. For the generation of the performance prediction model 302, actual values (input values and output values corresponding to the input values) collected by the information collection unit 402 in step S202 are used.

  The input value is data including each resource request (information on how to use the hardware resource). The output value is data representing the usage rate, throughput, TAT, and the like of each resource. The hardware layer performance prediction model generation unit 302 uses these input / output values to generate the performance prediction model 302 by arithmetic processing such as neural network, genetic algorithm, polynomial function, regression analysis, and the like. However, the present invention described by taking this embodiment as an example is not limited to these arithmetic processes, and it is sufficient that the performance prediction model 302 can be realized by a similar approach using actually measured values.

  FIG. 9 is an explanatory diagram conceptually showing a method of synthesizing the performance prediction model 303 of the virtual environment in the performance prediction model synthesis unit 103 of the virtual environment shown in FIG. In the application layer 204 shown in FIG. 9, a request (805, 806) is input to the performance prediction model 801 of the application layer prepared for each virtual machine, and then processed using a queue, and then processed. This represents a situation in which the request (812, 813) is output. Here, the performance prediction model 801 is a model imitating the operation of an application executed in the application layer.

  That is, requests 805 and 806 are input to the performance prediction model 801 of the application layer by the operation of the operator 20 or the like. Here, the requests 805 and 806 are data created in a simulated manner for simulation using each performance prediction model 801. The contents of this data represent the load on each software resource, the use timing of hardware resources, and the like.

  When there are a plurality of virtual machines (201) operating in the application layer 204, requests to the respective virtual machines are input to the respective virtual machines simultaneously or with a time difference designated by the operator 20 or the like. . In the case of the example illustrated in FIG. 9, when requests 805 and 806 are input to the performance prediction model 801 of the application layer, the performance prediction model 801 performs processing corresponding to the request, and then processes the processed requests 812 and 813. Is output as a processing result related to the input request.

  Using the application layer performance prediction model 801, the virtual environment performance prediction model composition unit 103 calculates resource requests 807 and 808 necessary for the processing requested by the request from the input requests 805 and 805. The resource requests 807 and 808 are instructions for requesting a specific process for the resources constituting the hardware layer 205 necessary for realizing the process corresponding to the request input to each performance prediction model 801. Suppose that Here, the resources configuring the hardware layer 205 are, for example, a CPU, a storage device (HDD), a communication network (NIC: Network Interface Card), and the like.

  The performance prediction model composition unit 103 of the virtual environment inputs resource requests 807 and 808 for hardware resources calculated by the application layer 204 to the hardware layer 205. The resource requests 807 and 808 input to the hardware layer 205 are processed by the performance prediction model 802 of the hardware layer. The performance prediction model composition unit 103 in the virtual environment outputs processing completion notifications 810 and 811 from the hardware layer 205 to the application layer 204 after the processing by the performance prediction model 802 is completed. The processing completion notifications 810 and 811 correspond to the processing completion notification 810 (FIG. 4) related to the resource request in the first embodiment described above.

  Using the hardware layer performance prediction model 802, the virtual environment performance prediction model composition unit 103 outputs the performance index 809 calculated in the processing for the resource requests 807 and 808 from the hardware layer performance prediction model 802. . The performance index 809 is information such as the usage rate and TAT of each resource, for example.

  The performance prediction model composition unit 103 of the virtual environment outputs processed requests 812 and 813 from each performance prediction model 802 after the application layer 204 receives the processing completion notifications 810 and 811 output from the hardware layer 205. .

  Note that after the application layer 204 outputs (issues) the resource requests 807 and 808 to the hardware layer 205, processing completion notifications 810 and 811 output (issued) from the hardware layer 205 are input to the application layer 204. The time (period) until the processing is performed is defined as the processing time (processing required time) PT of the performance prediction model 801 of the application layer. The processing time PT is from when the resource requests 807 and 808 output from the application layer 204 are input to the hardware layer 205 until the hardware layer 205 outputs processing completion notifications 810 and 811 to the application layer 204. It can also be understood as a time (period).

  Using the performance index calculated in each of the above steps, the performance prediction model synthesis unit 103 of the virtual environment provides information that can predict the performance of the virtualization system 2 to the operator 20.

  In other words, the operator 20 predicts the performance of the virtualization system 2 based on the input / output relationship of the performance prediction model 801 of the application layer 204 replaced as an application operating in the virtual machine 201 of the application layer 204. be able to. Specifically, the operator 20 uses the relationship between the requests 805 and 806, which are input values to the performance prediction model 801 of the application layer, and the processing time PT for the request to determine the performance of the entire virtualization system 2. Can be predicted.

  The operator 20 can predict the performance of the hardware layer 205 based on the relationship between the input and output to the performance prediction model 802 in which the hardware layer 205 is replaced. Specifically, the operator 20 predicts the performance of the hardware layer 205 using the relationship between the resource requests 807 and 808 that are input values to the performance prediction model 802 and the performance index 809 that is an output value. be able to.

  Thereby, the operator 20 can predict the performance of the entire virtualization system 2.

  FIG. 10 is a flowchart showing the procedure of the synthesis process (performance prediction) performed by the performance prediction model synthesis unit 103 of the virtual environment in the second embodiment, and shows details of step S103 described above.

  Step S301: The performance prediction model composition unit 103 of the virtual environment inputs requests 805 and 806 to the performance prediction model 801 of the application layer in response to the operation of the operator 20 or the like. The request is data created in a simulated manner for simulation using each performance prediction model 301 of the application layer. That is, when a plurality of performance prediction models 801 are operating as illustrated in FIG. 9, a request is input to each performance prediction model. The contents of such data are the load on each software resource, the use timing of hardware resources, and the like. The individual requests to each virtual machine are input at the same time or at a timing with a time difference designated by the operator 20 or the like.

  Step S302: The performance prediction model composition unit 103 of the virtual environment calculates resource requests 807 and 808 according to the performance prediction model 801 of the application layer according to the request input in step S301. The resource request is data including information on which hardware resource is used by the process constituting the virtual machine and how.

  Step S303: The performance prediction model composition unit 103 of the virtual environment inputs the resource requests 807 and 808 calculated in Step S302 to the performance prediction model 802 of the hardware layer.

  Step S304: The performance prediction model composition unit 103 of the virtual environment processes the resource requests 807 and 808 input in step S303 by the hardware layer performance prediction model 802.

  Step S305: The performance prediction model composition unit 103 of the virtual environment inputs process completion notifications 810 and 811 indicating that the process in step S304 is completed from the hardware layer 205 to the application layer 204. At the timing when the processing completion notification is output from the hardware layer 205, the performance prediction model synthesis unit 103 uses the performance index (resource usage rate, TAT, etc.) 809 calculated by the hardware layer performance prediction model 802 as the hardware Output from layer 205. The timing for calculating the performance index 809 may be any timing. Specifically, the timing for calculating the performance index may be, for example, when the processing completion notifications 810 and 811 are output, or may be every time determined by the operator 20 or the like.

  Step S306: The virtual environment performance prediction model synthesis unit 103 checks the hardware layer 205 for the presence of an unprocessed resource request output by the application layer 204. If an unprocessed resource request remains in application layer 204 (YES in step S306), performance prediction model synthesis unit 103 returns the process to step S303. On the other hand, when all resource request processing has been completed (NO in step S306), performance prediction model combining unit 103 advances the process to step S307. That is, a loop composed of steps S303 to S306 is repeated until there is no resource request in the application layer 204.

  Step S307: When the performance prediction model composition unit 103 of the virtual environment determines that the processing has been completed for all resource requests in Step S306, it outputs processed requests 812 and 813 from each performance prediction model 801.

  Step S308: Using the calculation results (performance index 809 and the like) of the application layer 204 and the hardware layer 205 calculated by the performance prediction model synthesis unit 103 in the process so far, the operator and the like 20 Perform performance prediction.

Values (information) obtained in the processing steps from step S301 to step S307 described above are as follows. That is,
Requests (805, 806) input to the performance prediction model 801 of the application layer,
A resource request (807, 808) to the hardware layer 205 calculated by the performance prediction model 801 of the application layer,
-Performance index (809: usage rate of each resource, TAT, etc.) calculated by the hardware layer performance prediction model 802,
Processing time (processing time) PT.

  That is, by using the performance prediction model 802, the performance prediction model composition unit 103 of the virtualization environment uses the performance index (809: 809 :) based on the content of the request to the virtualization system 2 that is the target of performance prediction. The usage rate, TAT, and throughput of each resource can be calculated. Accordingly, the operator 20 can easily predict the performance of the entire virtualization system 2 as a performance prediction using the performance prediction model (303) of the virtual environment by referring to the calculated values described above. .

  The performance prediction apparatus 10 according to the present embodiment described above defines the configuration of the virtualization system 2 as the application layer 204 and the hardware layer 205, generates performance prediction models 801 and 802 for these layers, The performance prediction is performed by generating the performance prediction model 303 of the virtualization system 2 in which the dependency relationship is reflected. Thereby, even when the number of virtual machines (201) operating in the application layer 204 is single or plural, it is possible to easily predict the performance of the virtualization system 2.

(Modification of the first and second embodiments)
Next, modified examples of the first and second embodiments described above will be described. FIG. 11 is a diagram conceptually illustrating the preconditions of the performance prediction apparatus for the virtualization system according to the modification of the first and second embodiments of the present invention. In the present embodiment, when the performance of the virtualization system (target system) 2 is predicted, the configuration of the virtualization system is hierarchically composed of three layers: an application layer 204, a virtual machine monitor layer 206, and a hardware layer 205A. Defined as a simple configuration. This configuration is different from the precondition (two-layer structure) shown in FIG.

  That is, in this modification, it is also possible to generate performance prediction models in these three layers and perform performance prediction of the target system 2 using the generated performance prediction model. Even when defined in such three layers, it is necessary to generate a performance prediction model using an approach that can reflect the characteristics of the system configuration of each layer in the performance prediction model. Also by this modification, it is possible to achieve substantially the same effects as those of the above-described embodiments.

  According to each embodiment and the modification described above, since the behavior (behavior) in the real environment is reflected in the performance prediction model of the virtualization system 2, it can be adapted to various performance prediction applications. It is. That is, according to the above-described performance prediction apparatus (1, 10), when it is desired to confirm the behavior of an information processing system constructed using a virtualization technology prior to actual operation, the performance prediction apparatus (1, 10) Accurate performance prediction is realized.

<Example of hardware configuration>
Here, a configuration example of hardware capable of realizing the performance prediction apparatus according to each of the above-described embodiments and modifications thereof will be described. The performance prediction devices (1, 10) described above may be realized as a dedicated device, but may be realized using a computer (information processing device).

  FIG. 12 is a diagram for exemplarily explaining a hardware configuration of a computer (information processing apparatus) capable of realizing each embodiment of the present invention and its modification.

  The hardware of the information processing apparatus (computer) 100 shown in FIG. 12 includes a CPU 11, a communication interface (I / F) 12, an input / output user interface 13, a ROM (Read Only Memory) 14, a RAM (Random Access Memory) 15, A storage device 17 and a drive device 18 of a computer-readable storage medium 19 are provided, and these are connected via a bus 16. The input / output user interface 13 is a man-machine interface such as a keyboard which is an example of an input device and a display as an output device. The communication interface 13 is a general communication means for the devices (FIGS. 1 and 6) according to the above-described embodiments to communicate with an external device via the communication network 300. In the hardware configuration, the CPU 11 governs the overall operation of the information processing apparatus 100 as the performance prediction apparatus (1, 10) according to each embodiment.

  The present invention described with reference to each of the above-described embodiments and the modifications thereof is based on the functions of the flowcharts (FIGS. 7, 8, and 10) referred to in the description of the embodiments, or FIGS. In the block diagram shown, a program (computer program) capable of realizing each unit (each block) shown in the apparatus is supplied to the information processing apparatus 100 shown in FIG. 12, and then the program is read to the CPU 11 To achieve. The program supplied in the information processing apparatus 100 may be stored in a readable / writable temporary storage memory (15) or a non-volatile storage device (17) such as a hard disk drive. That is, in the storage device 17, the program group 17 </ b> A is a program that can realize the functions of the respective units shown in the performance prediction devices (1, 10) in the above-described embodiments, for example. Further, the various kinds of stored information 17B are, for example, the performance prediction model group 110 of the application layer in each of the above-described embodiments, the actual measurement values related to the target system 2, and the like. However, when the program is installed in the information processing apparatus 100, the structural unit of each program module is not limited to the division of each block shown in the block diagrams (FIG. 1 and FIG. 6), and those skilled in the art can implement the program. This is a matter that can be selected as appropriate.

  In the above case, the program can be supplied to the apparatus by a method of installing in the apparatus via various computer-readable recording media (19) such as CD-ROM and flash memory, and the Internet. Currently, a general procedure can be adopted, such as a method of downloading from the outside via the communication line (300). In such a case, the present invention can be understood to be configured by a code (program group 107A) constituting the computer program or a storage medium (19) in which the code is stored.

  In the above, this invention was demonstrated as an example applied to exemplary embodiment mentioned above and its Example. However, the technical scope of the present invention is not limited to the scope described in the above embodiments and examples. It will be apparent to those skilled in the art that various modifications and improvements can be made to such embodiments. In such a case, new embodiments to which such changes or improvements are added can also be included in the technical scope of the present invention. This is clear from the matters described in the claims.

  Note that a part or all of the above-described embodiment and its modifications can be described as the following supplementary notes. However, the present invention described by way of example with the above-described embodiment and its modifications is not limited to the following.

That is,
(Appendix 1)
A first performance prediction model generation unit that generates a performance prediction model of the application layer based on information related to the application that operates in a virtual machine executed in a virtualization system including an application layer and a hardware layer When,
A second performance prediction model generation unit that generates a performance prediction model of the hardware layer based on the measurement result of the virtualization system;
By combining the performance prediction model of the application layer and the performance prediction model of the hardware layer, the output value in the performance prediction model of one layer is used as the input value in the performance prediction model of the other layer. , A performance prediction model synthesis unit that generates a performance prediction model of the virtualization system in which the dependency relationship between the two layers is reflected;
A performance prediction apparatus for a virtualization system, comprising:

(Appendix 2)
The first performance prediction model generation unit includes:
In response to a request being input to the performance prediction model of the application layer, a resource request including information on how to use hardware resources constituting the hardware layer is issued to the hardware layer,
The second performance prediction model generation unit includes:
The hardware layer performance prediction model performs processing according to the resource request, and upon completion of the processing, issues a processing completion notification to the application layer,
The performance prediction model synthesis unit
The virtual system performance prediction apparatus according to appendix 1, wherein the resource request and the process completion notification are used in the synthesis process.

(Appendix 3)
The performance prediction model synthesis unit is a performance prediction model of the virtualization system,
The request;
The resource request;
A performance index calculated by the performance prediction model of the hardware layer; and
Appendix 2 wherein at least one of the required time from when the resource request is issued to the hardware layer to when the processing completion notification is issued from the hardware layer is presented The performance prediction apparatus of a virtualization system as described.

(Appendix 4)
The performance index is
4. The virtual system performance prediction apparatus according to appendix 3, wherein the performance prediction apparatus includes at least one of a usage rate of hardware resources constituting the hardware layer, TAT (Turn Around Time), and throughput.

(Appendix 5)
The hardware layer is
The virtual machine system performance prediction apparatus according to any one of appendix 1 to appendix 4, including a virtual machine monitor and physical machine hardware.

(Appendix 6)
The second performance prediction model generation unit includes:
Appendices 1 to 1, wherein the virtualization system is measured, and a performance prediction model of the hardware layer is generated based on an input value obtained as a result of the measurement and an output value corresponding to the input value. The virtualization system performance prediction apparatus according to any one of Appendix 5.

(Appendix 7)
Instead of the first performance prediction model generation unit or in addition to the first performance prediction model generation unit,
A performance prediction model specifying unit that selects a specific performance prediction model from a storage unit that stores one or more performance prediction models of the application layer generated based on design information of an application that operates on the virtual machine The performance prediction device for a virtualization system according to any one of appendix 1 to appendix 6, wherein

(Appendix 8)
The virtualization system performance prediction apparatus according to any one of appendix 1 to appendix 7, further comprising the storage unit.

(Appendix 9)
An input content designation unit capable of designating a setting value related to measurement of the virtualization system and an execution instruction of the measurement;
An information collection unit that collects an input value that is a measurement value of the virtualization system and an output value corresponding to the input value;
The virtual system performance prediction apparatus according to any one of Supplementary Note 1 to Supplementary Note 8, further comprising:

(Appendix 10)
Generate a performance prediction model of the application layer based on information about the application that operates in a virtual machine executed in a virtualization system composed of an application layer and a hardware layer,
Based on the measurement result of the virtualization system, generate a performance prediction model of the hardware layer,
Targeting the performance prediction model of the application layer and the performance prediction model of the hardware layer, by using the output value in the performance prediction model of one layer as an input value in the performance prediction model of the other layer, A virtualization system performance prediction method, wherein the virtualization system performance prediction model reflecting the dependency between the two layers is generated.

(Appendix 11)
The virtual system performance prediction method according to appendix 10, wherein the definition of the hardware layer includes a virtual machine monitor and hardware of a physical machine.

(Appendix 12)
A computer program for controlling the operation of a performance prediction device of a virtualization system, the computer program,
A first performance prediction model generation function that generates a performance prediction model of the application layer based on information related to the application that operates in a virtual machine executed in a virtualization system including an application layer and a hardware layer When,
A second performance prediction model generation function for generating a performance prediction model of the hardware layer based on the measurement result of the virtualization system;
By combining the performance prediction model of the application layer and the performance prediction model of the hardware layer, the output value in the performance prediction model of one layer is used as the input value in the performance prediction model of the other layer. A performance prediction model synthesis function for generating a performance prediction model of the virtualization system, in which the dependency relationship between the two layers is reflected,
A computer program characterized by being realized by a computer.

DESCRIPTION OF SYMBOLS 1 Performance prediction apparatus 2 (Target system) Virtualization system 3 Memory | storage part 10 Performance prediction apparatus 11 CPU
12 Communication interface 13 Input / output user interface 14 ROM
15 RAM
16 Bus 17 Storage device 18 Drive device 19 Storage medium 20 Operator etc. 101 Application layer performance prediction model generation unit 102 Hardware layer performance prediction model generation unit 103 Virtualization environment performance prediction model synthesis unit 104 Application layer performance prediction model Specific unit 110 Application layer performance prediction model group 201 Virtual machine 202 Virtual machine monitor 203 Physical machine hardware 204 Application layer 205 Hardware layer 205A Hardware layer 206 Virtual machine monitor layer 300 Communication network 301 Application layer performance prediction model 302 Performance prediction model of hardware layer 303 Performance prediction model of virtual environment 401 Input content specification unit 402 Information collection unit 801 Application layer performance prediction model Model 802 Performance prediction model of hardware layer 805 request 806 request 807 resource request to hardware resource 808 resource request to hardware resource 809 performance index 810 processing completion notification related to resource request 811 processing completion notification related to resource request 812 processed request 813 Request processed

Claims (10)

A first performance prediction model generation unit that generates a performance prediction model of the application layer based on information related to the application that operates in a virtual machine executed in a virtualization system including an application layer and a hardware layer When,
A second performance prediction model generation unit that generates a performance prediction model of the hardware layer based on the measurement result of the virtualization system;
By combining the performance prediction model of the application layer and the performance prediction model of the hardware layer, the output value in the performance prediction model of one layer is used as the input value in the performance prediction model of the other layer. , A performance prediction model synthesis unit that generates a performance prediction model of the virtualization system in which the dependency relationship between the two layers is reflected;
A performance prediction apparatus for a virtualization system, comprising: The first performance prediction model generation unit includes:
In response to a request being input to the performance prediction model of the application layer, a resource request including information on how to use hardware resources constituting the hardware layer is issued to the hardware layer,
The second performance prediction model generation unit includes:
The hardware layer performance prediction model performs processing according to the resource request, and upon completion of the processing, issues a processing completion notification to the application layer,
The performance prediction model synthesis unit
The virtualization system performance prediction apparatus according to claim 1, wherein the resource request and the process completion notification are used in the synthesis process. The performance prediction model synthesis unit is a performance prediction model of the virtualization system,
The request;
The resource request;
A performance index calculated by the performance prediction model of the hardware layer; and
The present invention presents at least one of a required time from when the resource request is issued to the hardware layer to when the processing completion notification is issued from the hardware layer. 3. The performance prediction apparatus for a virtualization system according to 2. The hardware layer is
The virtual machine performance prediction apparatus according to any one of claims 1 to 3, further comprising a virtual machine monitor and hardware of a physical machine. The second performance prediction model generation unit includes:
The hardware layer performance prediction model is generated based on an input value obtained by measuring the virtualization system and an output value corresponding to the input value. The performance prediction apparatus of the virtualization system according to any one of claims 1 to 4. Instead of the first performance prediction model generation unit or in addition to the first performance prediction model generation unit,
A performance prediction model specifying unit that selects a specific performance prediction model from a storage unit that stores one or more performance prediction models of the application layer generated based on design information of an application that operates on the virtual machine The virtual system performance prediction apparatus according to any one of claims 1 to 5, further comprising: An input content designation unit capable of designating a setting value related to measurement of the virtualization system and an execution instruction of the measurement;
An information collection unit that collects an input value that is a measurement value of the virtualization system and an output value corresponding to the input value;
The virtual system performance prediction apparatus according to claim 1, further comprising: Generate a performance prediction model of the application layer based on information about the application that operates in a virtual machine executed in a virtualization system composed of an application layer and a hardware layer,
Based on the measurement result of the virtualization system, generate a performance prediction model of the hardware layer,
Targeting the performance prediction model of the application layer and the performance prediction model of the hardware layer, by using the output value in the performance prediction model of one layer as an input value in the performance prediction model of the other layer, A virtualization system performance prediction method, wherein the virtualization system performance prediction model reflecting the dependency between the two layers is generated. The virtual system performance prediction method according to claim 8, wherein the definition of the hardware layer includes a virtual machine monitor and hardware of a physical machine. A computer program for controlling the operation of a performance prediction device of a virtualization system, the computer program,
A first performance prediction model generation function that generates a performance prediction model of the application layer based on information related to the application that operates in a virtual machine executed in a virtualization system including an application layer and a hardware layer When,
A second performance prediction model generation function for generating a performance prediction model of the hardware layer based on the measurement result of the virtualization system;
By combining the performance prediction model of the application layer and the performance prediction model of the hardware layer, the output value in the performance prediction model of one layer is used as the input value in the performance prediction model of the other layer. A performance prediction model synthesis function for generating a performance prediction model of the virtualization system, in which the dependency relationship between the two layers is reflected,
A computer program characterized by being realized by a computer.

Images