JP2015138432A - Performance prediction device, performance model generation method and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To attain a performance prediction device for more suitably predicting a performance of a virtualization system.SOLUTION: A performance prediction device comprises: first performance model generation means for generating a performance model of the first virtual machine respectively on the basis of information about an application of each of first virtual machines of a first application unit capable of obtaining information about an application; second performance model generation means for letting a second application unit including a second virtual machine unable to obtain information about an application running on a virtual machine and hardware unit as a measurement object unit, and generating a performance model of a measurement object unit on the basis of a result of measuring the measurement object unit as the measurement object; and synthesis means for generating a performance model of s system by synthesis processing of using an output value in the performance model of one unit as an input value in the performance model of the other unit with respect to the performance model of each first virtual machine and the performance model of the measurement object unit.

Description

  The present invention relates to a performance prediction apparatus, a performance model generation method, and a program.

  In the computer virtualization technology, one or a plurality of virtual machines (virtual machines) are operated using a physical machine. A system having an environment for performing such an operation is called a virtualized system.

  When such a virtualization system is used, resource conflicts and virtual machine monitor emulation occur in hardware resources constituting 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. The virtual machine monitor emulation means that the virtual machine monitor simultaneously emulates virtual hardware of a plurality of virtual machines.

  In Non-Patent Document 1, regarding emulation and resource contention, each time one virtual machine operating on a physical machine increases, the CPU performance deteriorates, and the deterioration affects the performance of a system executed on the virtual machine. Is described.

  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 hardware environment that constitutes the physical machine becomes complicated. If the operation of the virtual machine monitor becomes complicated, the performance of the virtual machine is considered to deteriorate. Therefore, in order to predict the performance of a virtualized system, it is necessary to consider overhead and resource contention and to predict the performance in a situation where a plurality of virtual machines operate.

  As a performance prediction method of such a virtual system, there is a method using a performance model. Here, the performance model is a model that predicts the performance of a virtualization system (target system) that is a target of performance prediction using an input to the virtualization system. Typical examples of performance models include queues and regression models. As described above, there are mainly two types of methods for generating a performance model used for performance prediction.

  The method includes (1) a method for generating a performance model imitating the operation of the target system, and (2) a statistical process based on an input value to the target system and an output value corresponding to the input value. This is a method for generating a performance model using machine learning.

  First, (1) a method for generating a performance model imitating the operation of the target system will be described. Hereinafter, this method of generating a performance model is referred to as a “white box approach”. The white box approach requires that details of the operation of the target system (for example, design information) be clear when generating a performance model. In other words, if the operation of the target system is clear, the white box approach can easily generate a performance model even when the target system is replaced with another system.

  Non-Patent Document 2 and Patent Documents 2 to 5 describe examples of a method for generating a performance model by a white box approach having such characteristics.

  Non-Patent Document 2 describes a method for generating a performance model considering overhead based on queuing theory.

  Patent document 2 estimates the average service demand for each class based on the measured response time for each workload in the computer system, and queues the computer system based on the estimated average service demand and the characteristics of the workload. A method for generating a model is described.

  Patent Document 3 discloses a method for performing performance evaluation by modeling a storage virtualization apparatus based on model information of each model using a model of each component of a storage system using storage virtualization technology. Have been described. That is, according to Patent Document 3, the modeled storage virtualization apparatus (performance model of the storage virtualization apparatus) is generated using a white box approach.

  Patent Document 4 discloses, based on a rule for calculating the performance of a physical system, information on a correspondence relationship between a data flow model and a virtual server model, and a virtual server model object and a physical server model object. A method for generating a performance evaluation model from correspondence information is described.

  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 the performance evaluation value generated for each layer is used as the performance evaluation value of the entire system. Is described. At this time, in Patent Document 5, an approximate system (performance model) used to calculate a performance evaluation value for each layer is generated using a white box approach.

  Next, (2) a method for generating a performance model using statistical processing and machine learning based on an input value to the target system and an output value corresponding to the input value will be described. Hereinafter, this method of generating a performance model is referred to as a “black box approach”. The black box approach generates a model based on actual measurements.

  Non-Patent Document 3 and Patent Document 6 describe examples of a method for generating a performance model by a black box approach.

  Non-Patent Document 3 describes that a single virtual machine generates a performance model using an actually measured value in a state where the virtual machine is operating in a virtual machine monitor and a linear model.

  Patent Document 6 describes a virtualization system that calculates the overhead of a virtual machine based on load data and configuration data collected in advance, and regenerates the load using the load data and overhead. That is, the virtualization system described in Patent Document 6 generates a load for performing performance evaluation. That is, in the technique of Patent Document 6, a virtual machine (performance model) having a load condition close to an actual operation environment is generated using actual load data.

  An example of a performance prediction method using a performance model generated using the two types of generation methods described above is described in Patent Document 1.

  According to Patent Document 1, a system resource usage amount of a transaction is predicted from system resource usage amounts of a plurality of software components measured in advance. In the system performance prediction method described in Patent Document 1, the predicted performance of the entire target system is predicted by inputting the predicted system resource usage into the system performance prediction model.

  Patent Document 7 discloses a technique that uses a performance prediction function or performance model including overhead in a performance predicted value calculation device of a virtualized system.

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

Ajishiro Ikudai, Tanaka Akihiro, "Evaluation of resource management overhead in virtual machine environment", IPSJ research report. EVA, [System Evaluation], Information Processing Society of Japan, June 2006, Vol. 2006, no. 66, p. 17-22 Febricio Benevenuto and 6 others, “Performance Models for Virtualized Applications”, Frontiers of High Performance Computing and Networking-ISPA 2006 Workshops, Springer Berlin Heidelberg, 2006, Vol. 4331, p. 427-439 Lei Lu and five others, “Untangling Mixed Informantion to Calibrate Resource Utilization in Virtual Machines”, Proceeding of the 8th ACM international conference on Autonomic computing (ICAC '11), ACM, 2011, p. 151-160

  In a virtualization system, physical machine hardware may be shared by a plurality of virtual machines. Among the plurality of virtual machines, for example, a virtual machine administrator may not be able to obtain virtual machine design information, for example, because it is different from the user who uses the virtualization system. There is sex.

  However, when predicting the performance of such a virtualization system, it is necessary to consider virtual machines for which design information and the like as described above cannot be obtained.

  Further, when the hardware resource of the physical machine is managed by a third party, there is a possibility that the information on the hardware resource (hardware resource information) cannot be obtained.

  In the technique of Patent Document 1, the performance data obtained by measurement is used as a parameter value of a queue to predict the performance of the entire system. However, in the technique of Patent Document 1, performance prediction of a virtualization system is not performed in consideration of a virtual machine for which design information or the like cannot be obtained.

  In the techniques described in Non-Patent Document 2 and Patent Documents 2 to 5, a performance model is created by the white box approach described above. With such a technique, it is not known what operation a virtual machine for which design information or the like cannot be obtained or a hardware resource for which hardware resource information cannot be obtained. Therefore, with the above technology, it is not possible to create a performance model of a virtualization system including a virtual machine for which design information or the like cannot be obtained.

  Further, in Patent Document 6, a situation where a plurality of virtual machines operate in a host OS (Operating System) is created, and performance prediction is performed by actually measuring an evaluation target application operating in the guest OS. However, in Patent Document 6, since performance prediction is performed based on actual measurement values, when the specifications of the virtual machine are changed, the evaluation target application must be actually measured for each change. Therefore, in the technique of Patent Document 6, if a virtual machine for which design information or the like cannot be obtained is included, or hardware resource information in that case cannot be obtained, an application of the added virtual machine can be executed alone. difficult. For this reason, it is necessary to re-measure all virtual machines for every change, which is troublesome.

  In the technique described in Patent Document 7, the performance prediction of the virtualization system in consideration of a virtual machine for which design information or the like cannot be obtained is not performed.

  The present invention has been made in view of the above problems, and an object thereof is to realize a performance prediction apparatus capable of predicting the performance of a virtualization system more appropriately.

  A performance prediction apparatus according to an aspect of the present invention is a performance prediction apparatus for predicting the performance of an information processing system, and the information processing system is capable of obtaining information on an application operating on a virtual machine. Alternatively, a first application unit including a plurality of first virtual machines, a second application unit including one or a plurality of second virtual machines for which information relating to an application running on the virtual machine cannot be obtained, and a hardware unit The performance prediction device generates a performance model of the first virtual machine based on information related to the application of each of the first virtual machines of the first application unit, respectively. First performance model generation means, the second application unit, and the hardware unit Based on the result measured as the measurement target unit, second performance model generation means for generating a performance model of the measurement target unit, the performance model of each first virtual machine of the first application unit, and the measurement The information in which the dependency relationship between the two parts is reflected by the synthesis process using the output value in the performance model of one part as the input value in the performance model of the other part with respect to the performance model of the target part System performance model synthesis means for generating a performance model of the processing system.

  A performance model generation method according to an aspect of the present invention is a performance model generation method in a performance prediction apparatus for predicting the performance of an information processing system, wherein the information processing system is information related to an application operating on a virtual machine. A first application unit that includes one or more first virtual machines, and a second application unit that includes one or more second virtual machines for which information about applications running on the virtual machines is not available And the hardware part, wherein the performance model generation method uses the first virtual machine based on information about the application of each of the first virtual machines of the first application part. Each performance model is generated, and the second application unit and the hardware are generated. A performance model of the measurement target unit is generated based on the measurement result of the measurement target unit as a measurement target, and the performance model of each first virtual machine of the first application unit And the dependency between the two parts is reflected by using the output value in the performance model of one part as the input value in the performance model of the other part with respect to the performance model of the measurement target part And generating a performance model of the information processing system.

  A program according to an aspect of the present invention is a program that is executed by a computer including a performance prediction device for predicting the performance of an information processing system, and the information processing system includes information related to an application that operates on a virtual machine. A first application unit that includes one or more first virtual machines that are available; a second application unit that includes one or more second virtual machines for which information about an application running on the virtual machine is not available; And a hardware unit, wherein the program determines a performance model of the first virtual machine based on information about the application of each of the first virtual machines of the first application unit. Each of the generated processes, the second application unit, and the A process that generates a performance model of the measurement target unit based on a result of measurement using the measurement target unit as a measurement target, and the first virtual machine of the first application unit. The dependency relationship between the two parts is obtained by a synthesis process in which the output value in the performance model of one part is used as the input value in the performance model of the other part of the performance model and the performance model of the measurement target part. The reflected processing for generating the performance model of the information processing system is executed by a computer.

  According to the present invention, the performance of the virtualization system can be predicted more appropriately.

It is a block diagram which shows an example of a function structure of the performance prediction apparatus for predicting the performance of a virtualization system based on the 1st Embodiment of this invention. It is a figure for demonstrating notionally the precondition of the virtualization system based on the 1st Embodiment of this invention. It is a figure for demonstrating notionally the synthetic | combination process which the performance prediction model synthetic | combination part of the performance prediction apparatus which concerns on 1st Embodiment performs. It is a figure which illustrates the performance parameter | index which the measurement object part performance model which concerns on 1st Embodiment outputs. It is a block diagram which shows an example of a function structure of the performance prediction apparatus for predicting the performance of a virtualization system based on the 2nd Embodiment of this invention. It is a flowchart which shows an example of the flow of the performance model production | generation process by the performance prediction apparatus which concerns on the 2nd Embodiment of this invention. In the 2nd Embodiment of this invention, it is a flowchart which shows an example of the flow of a production | generation process of a measurement object part performance model. In the 2nd Embodiment of this invention, it is a flowchart which shows an example of the flow of a synthetic | combination process of a system performance model synthetic | combination part. In the 2nd Embodiment of this invention, it is a figure for demonstrating notionally the synthetic | combination process of a system performance model synthetic | combination part. It is a figure which illustrates illustartively the hardware constitutions of the computer (information processing apparatus) which can implement | achieve each embodiment of this invention.

<First Embodiment>
A first embodiment of the present invention will be described in detail with reference to the drawings.

  FIG. 1 is a block diagram illustrating an example of a functional configuration of a performance prediction apparatus for predicting the performance of a virtualization system according to the present embodiment. As shown in FIG. 1, the performance prediction device 1 is connected to the virtualization system 2 so as to be communicable. The virtualization system 2 is a target for which the performance prediction device 1 predicts performance (hereinafter also referred to as a target system or a virtual environment). 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 the present embodiment, detailed description regarding such communication means is omitted.

  The virtualization system (virtualized environment) 2 according to the present embodiment will be described as an example of an information processing system constructed using a virtualization technology.

  The virtualization system 2 includes a plurality of virtual machines. The plurality of virtual machines include, for example, those in which design information of the virtual machine cannot be obtained due to reasons such as the administrator of the virtual machine being different from the user using the virtualization system 2. That is, the virtualization system according to the present embodiment includes one or more virtual machines in which an application with a clear operation is executed and one or more virtual machines in which an application with an unknown operation is executed. The configuration shares the same physical machine. Hereinafter, the former virtual machine is also referred to as a first virtual machine or a prediction target virtual machine. The latter virtual machine is also referred to as a second virtual machine, an operation unknown virtual machine, or a non-predicted virtual machine.

(Virtualization system 2)
Here, the concept of the virtualization system 2 according to the present embodiment will be described with reference to FIG. FIG. 2 is a diagram conceptually illustrating a general virtualization system according to the first embodiment of the present invention.

  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 is a host OS type in which the virtual machine monitor operates in a host OS (Operating System), or 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 present embodiment is similar to a general virtualization system in that a virtual machine (first virtual machine 201, second virtual machine 301), a virtual machine monitor (hypervisor) 202, Three layers of hardware 203 of the physical machine are formed. As shown in FIG. 2, the virtual machines (201, 301) are M (M is a natural number excluding 0) first virtual machines 201 and N (N is a natural number excluding 0) second virtual machines. The number of machines is not limited. Each virtual machine (201, 301) can provide a guest OS environment and execute one or more arbitrary applications (AP) as illustrated in FIG.

  Here, the first virtual machine 201 is a virtual machine on which an application with a clear operation is executed, and the second virtual machine 301 is a virtual machine on which an application with an unknown operation is executed. In other words, the first virtual machine 201 in the present embodiment is a virtual machine from which information relating to an application running on the first virtual machine 201 can be obtained. In the present embodiment, the first virtual machine 201 is also referred to as a virtual machine from which details such as virtual machine design information can be obtained. Similarly, the second virtual machine 301 is a virtual machine for which information relating to an application running on the second virtual machine 301 cannot be obtained, and the second virtual machine 301 is a virtual machine for which details cannot be obtained. It is said that it is a virtual machine owned by.

In this embodiment, the virtualization system (target system) 2 having such a three-layer structure is represented by the following two layers (a) and (b) (two layers) surrounded by a broken line in FIG. ). That is,
(A) a hardware layer 205 including a virtual machine monitor 202 and physical machine hardware 203;
(B) An application layer 204 configured by virtual machines (201, 301).

Among these, when the performance prediction device 1 predicts the performance of the virtualization system 2, when the virtual machine included in the application layer 204 is only the first virtual machine 201, the virtualization system 2 includes two copies. It is defined that it is configured. In this embodiment, the virtualization system 2 having the following structure is defined as three parts. That is,
(1) a hardware unit 300 (hardware layer 205) configured by a virtual machine monitor 202 and physical machine hardware 203;
(2) a first application unit 400 configured by one or more first virtual machines 201,
(3) A second application unit 500 including one or more second virtual machines 301.

  That is, when predicting the performance of the virtualization system 2 handled as a two-layer structure, the performance prediction apparatus 1 according to the present embodiment defines that the virtualization system 2 is configured by three parts. And the performance prediction apparatus 1 produces | generates the performance model for predicting the performance of the virtualization system 2 on the assumption of such a definition. Here, as a method of definition in the performance prediction apparatus 1, for example, the first virtual machine 201 is registered in association with the first application unit 400, and the second virtual machine 301 is associated with the second application unit 500. The virtual machine monitor 202 and the physical machine hardware 203 may be registered in association with the hardware unit 300.

  In addition, the hardware unit 300 and the second application unit 500 surrounded by a one-dot chain line in FIG. 2 are collectively referred to as a measurement target unit 501.

(Performance prediction device 1)
As shown in FIG. 1, the performance prediction apparatus 1 includes a measurement target unit performance model generation unit (second performance model generation unit) 3, a virtual machine performance model generation unit (first performance model generation unit) 4, a system And a performance model synthesis unit 5.

  The generation of the performance model in the present embodiment includes at least a white box approach for creating a performance model that imitates the details of the system operation, and a black box approach for constructing a performance model based on the actual measurement values of the performance prediction target system. Will be used. For the generation of the performance model, for example, a queuing theory or a regression analysis method can be employed. In addition, since the performance model generation method in this Embodiment employ | adopts a general method, detailed description is abbreviate | omitted.

  The measurement target unit performance model generation unit 3 performs the measurement target unit performance model for the measurement target unit 501 (FIG. 3) based on the measurement result (actual measurement value) of the virtualization system 2 when the second virtual machine 301 operates. 310). More specifically, the measurement target unit performance model generation unit 3 performs the black box approach described above based on the measurement result of the measurement target unit 501 including the second application unit 500 and the hardware unit 300 as a measurement target. Thus, a performance model of the measurement target unit 501 is generated.

  In the present embodiment, generation of a performance model by the black box approach (also referred to as model generation) is based on an input value to an arbitrary system and an output value corresponding to the input value as described above. This is a technique for generating a performance model by processing or machine learning. In the model generation method based on the black box approach, for example, regression analysis, neural network, or genetic algorithm can be used. The model generation method is not limited to this, and may be any method that can realize the same approach method.

  The measurement target unit performance model generation unit 3 may generate the measurement target unit performance model (310) monthly, daily, every predetermined time, or the like. In the present embodiment, the measurement target unit 501 is also referred to as a virtualization system influence unit.

  The measurement target unit performance model generation unit 3 supplies the generated performance model (measurement target unit performance model 310) to the system performance model synthesis unit 5.

  The virtual machine performance model generation unit 4 performs the performance model of each virtual machine by the above-described white box approach based on the information regarding the application that operates in each first virtual machine 201 of the first application unit 400 of the virtualization system 2. (Virtual machine performance model 410 in FIG. 3) is generated. That is, when the operation of the virtual machine application in the virtualization system 2 is clear, the virtual machine performance model generation unit 4 generates a performance model for each virtual machine.

  The application of the first virtual machine 201 is an application that is executed on the first virtual machine 201 in response to an operation of an operator or the like, and the details of the operation of the application can be obtained.

  In the present embodiment, the model generation by the white box approach is a method of generating a performance model imitating the operation of an arbitrary system based on the design information of the arbitrary system as described above. In the model generation method based on the white box approach, for example, queuing theory or Petri net can be used. The model generation method is not limited to this, and may be any method that can realize the same approach method.

  The virtual machine performance model generation unit 4 supplies the generated performance model (virtual machine performance model 410) to the system performance model synthesis unit 5.

  The system performance model synthesis unit 5 receives a performance model (measurement target unit performance model 310) from the measurement target unit performance model generation unit 3. Further, the system performance model synthesis unit 5 receives a performance model (virtual machine performance model 410) from the virtual machine performance model generation unit 4. Then, the system performance model combining unit 5 performs a combining process using the output value in one performance model as an input value to the other performance model in the performance model. Through this synthesis process, the performance model of the virtualization system 2 that reflects the dependency between the measurement target unit performance model (310) and the virtual machine performance model (410) for each virtual machine of the first application unit ( A system performance model 510) is generated. Hereinafter, in the present embodiment, the performance model generation processing considering the dependency between the measurement target unit performance model (310) and the performance model (410) for each virtual machine of the first application unit is combined ( (Combining process).

  The processing of the system performance model synthesis unit 5 will be described in more detail with reference to FIG.

  Generally, in a virtualization system, a virtual machine executes an application by using hardware resources of a physical machine. Therefore, there is a dependency relationship between the hardware unit 300 and the first virtual machine 201 and the second virtual machine 301. That is, in a state where the virtual machines (201, 301) and the hardware unit 300 are separately provided, the performance of the virtualization system 2 that is a target for which performance should be predicted cannot be predicted appropriately.

  Therefore, the system performance model synthesis unit 5 of the performance prediction apparatus 1 according to the present embodiment generates a performance model (measurement target unit performance model 310) generated by the measurement target unit performance model generation unit 3 and virtual machine performance model generation. Using the performance model (virtual machine performance model 410) of each first virtual machine 201 of the first application unit 400 generated by the unit 4, a performance model (system performance model 510) of the virtualization system 2 is generated To do. Thereby, the performance prediction of the virtualization system 2 is enabled.

  FIG. 3 is a diagram for conceptually explaining the combining process performed by the system performance model combining unit 5.

  First, the system performance model synthesis unit 5 sends a resource request 60 to the hardware resource to each virtual machine performance model 410 of the first application unit 400 in response to a request being input to the first virtual machine 201. Generate. Then, the system performance model synthesis unit 5 inputs the resource request 60 for the hardware resource of the hardware unit 300 generated by each virtual machine performance model 410 of the first application unit 400 to the measurement target unit performance model 310. To do. Here, the resource request 60 for the hardware resource includes information regarding how each virtual machine performance model 410 of the first application unit 400 uses the hardware resource.

  Then, the system performance model synthesis unit 5 responds to the resource request 60 input to the measurement target unit performance model 310, and a performance index (to be described later, FIG. 9) that can predict the performance of the measurement target unit performance model 310. Is equivalent to the performance index 604 shown in FIG.

  Here, the measurement target unit performance model 310 may employ, for example, the usage rate of hardware resources, TAT (Turn Around Time), throughput, and the like in units of month, day, and time as the performance index. The performance index is illustrated in FIG. FIG. 4 is a diagram illustrating a performance index calculated by the measurement target unit performance model 310 according to the present embodiment. In the example shown in FIG. 4, “TAT (ms: milliseconds)” “when the time is changed (every 2 hours from 8 o'clock, and from 18:00 to 18:00 as one unit)” It represents the average value of “CPU usage rate (%)” and “throughput (tps: number of transactions per second)”. Then, the system performance model synthesis unit 5 provides the performance index calculated by the measurement target unit performance model 310, for example, as shown in FIG. In addition, although the structure provided using the display means which is not shown in figure may be sufficient as the provision method to an operator etc., this invention is not limited to this.

  When the process according to the input resource request 60 is completed in the measurement target unit performance model 310, the system performance model synthesis unit 5 issues a process completion notification 61 corresponding to the resource request to the measurement target unit performance model 310. Let Then, the system performance model composition unit 5 inputs a processing completion notification 61 to each virtual machine performance model 410 of the first application unit 400. Note that the mutual exchange between the measurement target unit performance model 310 and the virtual machine performance model 410 continues until the command (processing process) input to the virtualization system 2 is completed.

  That is, the system performance model synthesis unit 5 of the performance prediction apparatus 1 according to the present embodiment is different from the virtual machine performance model 410 of the first application unit 400 and the measurement target unit performance model 310 of the measurement target unit 501. A synthesis process is performed in which the output value in the performance model of the part is used as the input value of the performance model of the other part. With such a configuration, the system performance model synthesizing unit 5 generates a system performance model 510 (for example, the above-described performance index) of the virtualization system 2 in which the dependency between the two parts is reflected.

(effect)
As described above, the virtualization system 2 according to the present embodiment includes a first application unit 400 including one or more first virtual machines 201 that can obtain information on an application running on a virtual machine, a virtual It has the 2nd application part 500 containing the 1 or several 2nd virtual machine 301 and the hardware part 300 which cannot obtain the information regarding the application which operate | moves on a machine.

  According to the performance prediction apparatus 1 for predicting the performance of the virtualization system 2 as described above, the performance of the virtualization system 2 can be predicted more preferably.

  This is because the performance prediction device 1 according to the present embodiment includes the measurement target unit performance model generation unit 3, the virtual machine performance model generation unit 4, and the system performance model synthesis unit 5.

  The virtual machine performance model generation unit 4 generates a performance model (virtual machine performance model 410) of the first virtual machine 201 based on information related to the application of each first virtual machine 201 of the first application unit 400. . The measurement target unit performance model generation unit 3 uses the second application unit 500 and the hardware unit 300 as the measurement target unit 501, and based on the measurement result of the measurement target unit 501 as the measurement target, the performance model of the measurement target unit 501 ( A measurement target part performance model 310) is generated. Further, the system performance model synthesis unit 5 uses the output value in the performance model of one part as the input value in the performance model of the other part for the virtual machine performance model 410 and the measurement target part performance model 310. Through the synthesis process, a performance model of the virtualization system 2 (system performance model 510) reflecting the dependency between the two parts is generated.

  Therefore, the performance prediction apparatus 1 according to the present embodiment can appropriately predict the performance of the virtualization system 2 using the system performance model 510.

  In addition, the performance prediction apparatus 1 according to the present embodiment adopts a white box approach, and uses the first virtual machine using application information executed by the first virtual machine 201 that operates in the first application unit 400. A virtual machine performance model 410 is generated for each 201. In the environment where a plurality of first virtual machines 201 operate in the first application unit 400, a case will be described in which there is a change in the configuration of the virtual machine, such as addition or deletion of the first virtual machine 201, for example. At this time, the virtual machine performance model generation unit 4 can create a performance model for the first application unit 400 by adding / deleting the virtual machine performance model 410 of the added / deleted first virtual machine 201. it can. Therefore, according to the present embodiment, there is no need for the trouble of re-measurement of the virtualization system 2 according to the number of the first virtual machines 201, which is necessary when the black box approach is adopted.

  Moreover, the performance prediction apparatus 1 in the present embodiment generates the measurement target part performance model 310 using the black box approach. For this reason, (a) when the operation or specification of each component of physical hardware with complex operations is unclear, (b) a virtual machine that is not subject to prediction or a virtual machine whose details cannot be obtained is on the physical machine. Or (c) the performance model generation unit to be measured, regardless of whether (c) information on how many virtual machines are shared by the third party is not available. 3 can generate the measurement target part performance model 310. In other words, by adopting the black box approach, the measurement target unit performance model generation unit 3 does not need to clearly grasp the operation of the system or application unlike the white box approach. It can generate suitably. Therefore, even if the emulation of the second virtual machine 301, the details of resource contention in the hardware 203 of the physical machine, or the details of the application executed on the second virtual machine 301 are not available, the measurement target unit performance model The generation unit 3 can generate the measurement target unit performance model 310 in which the influence on the performance exerted by these components is correctly considered without explicitly describing details.

  Then, the system performance model synthesis unit 5 in the present embodiment performs a synthesis process using the virtual machine performance model 410 of each first virtual machine 201 and the measurement target unit performance model 310 in the first application unit 400, A system performance model 510 is generated in consideration of the dependency between these parts. That is, according to the present embodiment, it is possible to suitably predict the performance of the virtualization system 2 using the system performance model 510 generated in this way.

<Second Embodiment>
Next, a second embodiment based on the performance prediction device 1 according to the first embodiment described above will be described. For convenience of explanation, members having the same functions as those included in the drawings described in the first embodiment described above are given the same reference numerals, and descriptions thereof are omitted. In this embodiment, the virtualization system for which the performance prediction apparatus 10 predicts performance is the same as the first embodiment in the three parts (hardware unit 300, first application, etc.) illustrated in FIG. And the second application unit 500).

(Performance prediction device 10)
FIG. 5 is a block diagram showing a functional configuration of the performance prediction apparatus 10 for predicting the performance of the virtualization system 2 according to the second embodiment of the present invention. As shown in FIG. 5, the performance prediction apparatus 10 includes an input content specification unit 30, an information collection unit 31, a measurement target unit performance model generation unit 32, a storage unit 41, a virtual machine performance model identification unit 42, a system And a performance model synthesis unit 50.

  In the present embodiment, the storage unit 41 is described as an example configured inside the performance prediction device 10, but the storage unit 41 is realized by a storage device outside the performance prediction device 10 or the like. May be. In this case, the virtual machine performance model specifying unit 42 may refer to the storage device.

  Further, the performance prediction apparatus 10 in the present embodiment is connected to the virtualization system 2 so as to be communicable as in the first embodiment. Further, the performance prediction apparatus 10 is configured to receive an instruction from an operator 9 or the like.

(Input content specifying unit 30)
The input content designation unit 30 has a function as a user interface (UI, man machine) for the operator 9 or the like. The operator or the like 9 can specify the content related to the measurement using the input content specifying unit 30 when measuring the actual measurement value related to the virtualization system 2. Therefore, by the input operation by the operator 9 or the like 9 using the input content specifying unit 30, the virtualization system 2 causes the setting values related to the measurement of the actual measurement values (usage rate of each resource, throughput, TAT, etc.) of the virtualization system 2. , The measurement period (monthly, date and time, time, etc.), and the contents related to the execution command such as measurement are specified.

  The input content specifying unit 30 has been described as an example of input (designation) by the operator 9 or the like, but the present invention is not limited to this. A command from an external device such as a server via a LAN or the Internet may be input to the input content specifying unit 30.

  Further, the input content specifying unit 30 inputs (sets) the specified content (input content) to the virtualization system 2. Specifically, the input content specifying unit 30 is a command indicating a setting value, a measurement period (monthly, date, time, etc.) related to measurement of the virtualization system 2 and a measurement instruction, which are input by an operation of the operator 9 or the like. Are set in the virtualization system 2. Thereby, the virtualization system 2 measures the virtualization system 2 based on the information input by the input content specifying unit 30.

  The actual measurement value output from the virtualization system 2 using the above input, and the actual measurement value obtained as a result of the measurement is generated by the measurement target unit performance model generation unit 32 by the measurement target unit performance model 310. Used for

  The measurement of the actual measurement value in the virtualization system 2 may be performed prior to synthesis processing by the system performance model synthesis unit 50 described later. The contents of the measurement include a state in which N second virtual machines 301 are operating on the hardware unit 300 (N is a natural number including 0), and the usage frequency of the system such as monthly, daily, or every time. Benchmark test using benchmark software for each unit on the time axis with different trends. This benchmark software is software (computer program) that performs processing that can load various hardware resources related to the virtualization system 2, such as a CPU, a storage device, a communication network, and the like. There may be. In this embodiment, arbitrary software (computer program) that can reproduce the load situation of a specific hardware resource is adopted as benchmark software. However, the present invention is not limited to this. Since the measurement method of the actual measurement value in the virtualization system 2 can adopt a general procedure, detailed description in this embodiment is omitted.

(Information collection unit 31)
The information collection unit 31 collects actual measurement values measured for the virtualization system 2. Here, the actual measurement value is information that makes a pair of an input value in a benchmark test related to the virtualization system 2 executed according to the input content from the input content specifying unit 30 and an output value corresponding to the input value. The information collection unit 31 collects an input value to the virtualization system 2 and an output value corresponding to the input value as the actual measurement values described above from the virtualization system 2.

  Here, the input value includes a request for each hardware resource constituting the virtualization system 2 (information including information on how to use the hardware resource).

  In addition, although the information collection part 31 demonstrated the structure which collects this input value from the virtualization system 2, this invention is not limited to this. The information collecting unit 31 may be configured to collect the input value from the input content specifying unit 30.

  The output value corresponding to the input value is, for example, the usage rate, throughput, TAT, etc. of each hardware resource.

  The information collection unit 31 adds the collected actual measurement values to the measurement target unit performance model 310 that is a performance model of the measurement target unit 501 including the hardware unit 300 and the second application unit 500 including the second virtual machine 301. It is possible to input.

  Further, the information collection unit 31 stores the collected measured values (that is, the input values to the virtualization system 2 and the input values) in a computer-readable storage medium (for example, the storage unit 41) such as an HDD (Hard Disk Drive). Output value corresponding to the value). Further, the information collection unit 31 supplies the collected actual measurement value to the measurement target unit performance model generation unit 32 when the collection of the information is completed (measurement specified by the input content specification unit 30 is completed). To do.

(Measurement target part performance model generation part 32)
The measurement target unit performance model generation unit 32 performs the same processing as the measurement target unit performance model generation unit 3 in the first embodiment. The measurement target unit performance model generation unit 32 generates the measurement target unit performance model 310 by the black box approach based on the information (the above-described actual measurement value) input from the information collection unit 31.

  As described above, the actual measurement value includes an input value to the virtualization system 2 and an output value corresponding to the input value. As described above, the input value is data including a resource request (information including information on how to use hardware resources). The output value is data representing the usage rate, throughput, TAT, and the like of each resource. These input / output values are data collected every month, every day, and every time. The measurement target part performance model generation unit 32 generates a measurement target part performance model 310 by using arithmetic operations such as a neural neural network, a genetic algorithm, a polynomial function, and regression analysis using these input / output values. However, in the present invention, the generation method of the measurement target part performance model 310 is not limited to these calculation processes, and any method may be used as long as the measurement target part performance model 310 is realized by a similar approach using actually measured values.

  Then, the measurement target unit performance model generation unit 32 supplies the generated measurement target unit performance model 310 to the system performance model synthesis unit 50.

(Storage unit 41)
The storage unit 41 is a storage unit realized by, for example, a hard disk device that is a computer-readable storage medium. The storage unit 41 stores a performance model group 40.

  The performance model group 40 stored in the storage unit 41 is generated using design information of an application executed in the first virtual machine 201 operating in the first application unit 400. The performance model group 40 is a candidate that can be specified as the virtual machine performance model 410. That is, the performance model group 40 is a set of models having a configuration imitating the behavior of an application operating in the first application unit 400 and the first virtual machine 201.

  In the present embodiment, the individual virtual machine performance models 410 forming the performance model group 40 are generated by the above-described white box approach, similarly to the virtual machine performance model 410 described in the first embodiment. In the present embodiment, for example, the performance prediction apparatus 10 uses, for example, a general method that uses a white box approach to a virtual machine performance model constituting the performance model group 40 prior to synthesis processing by the system performance model synthesis unit 50 described later. It may be obtained by a general procedure. In the following description, it is assumed that the virtual machine performance model 410 is generated using the white box approach.

  That is, the storage unit 41 stores one or a plurality of virtual machine performance models 410 of applications that may operate in the first virtual machine 201 of the first application unit 400 generated using the white box approach. Yes.

For the performance model group 40, for example, a performance model generated using a queue, a Petri net, or the like according to the design information of the application executed on the first virtual machine 201 can be adopted. In addition, the production | generation method of the performance model group 40 in this Embodiment is not limited above. As the performance model generation method, any method may be employed in which the output is determined by a predetermined procedure such as analysis calculation or simulation for the input.
(Virtual machine performance model specifying unit 42)
The virtual machine performance model specifying unit 42 has a function as a user interface (UI, man machine) for the operator 9 or the like. The operator or the like 9 can specify information related to an application operating on the first virtual machine 201 executed by the first application unit 400 using the virtual machine performance model specifying unit 42. By the input operation by the operator 9 using the virtual machine performance model specifying unit 42, the virtual machine performance model specifying unit 42 specifies the virtual machine performance model 410 corresponding to or matching the input content from the performance model group 40. ,get.

  Specifically, the virtual machine performance model specifying unit 42 refers to the storage unit 41 based on information related to an application operating on the first virtual machine 201 in accordance with an operation of the operator 9 or the like. Then, the virtual machine performance model specifying unit 42 specifies a specific virtual machine performance model 410 associated with the information related to the application from the performance model group 40. Then, the virtual machine performance model specifying unit 42 acquires the specified virtual machine performance model 410 from the storage unit 41.

  Further, the virtual machine performance model specifying unit 42 supplies the acquired virtual machine performance model 410 to the system performance model synthesis unit 50.

  Here, the information (input contents) instructed by the operator 9 or the like is information related to an application executed in each first virtual machine 201 executed by the first application unit 400.

  The virtual machine performance model 410 corresponding to the input content represents the application content for each first virtual machine 201 that is a target whose performance is to be predicted. The contents of this application differ for each first virtual machine 201. Therefore, when three first virtual machines 201 are operating in the first application unit 400, the virtual machine performance model specifying unit 42 matches virtual application information in the three first virtual machines 201. It is necessary to take out the machine performance model 410 from the performance model group 40 (storage unit 41).

  The virtual machine performance model 410 is specified by using a virtual machine performance model 410 generated in advance using application information having the same contents as the information related to a desired application that the operator 9 desires to operate in the first virtual machine 201. The operator 9 or the like may be selected from the performance model group 40 by, for example, a human operation.

Note that although description has been made by taking as an example that input (designation) by the operator 9 or the like is performed on the virtual machine performance model specifying unit 42, the present invention is not limited to this. Necessary information may be input to the virtual machine performance model specifying unit 42 from an external device such as a server via a LAN or the Internet.
(System performance model synthesis unit 50)
The system performance model combining unit 50 combines the measurement target unit performance model 310 generated by the measurement target unit performance model generating unit 32 and the virtual machine performance model 410 specified by the virtual machine performance model specifying unit 42. As a result, the system performance model 510 of the virtualization system 2 is generated. The method for generating the system performance model 510 in the present embodiment is the same as that in the first embodiment. For example, the system performance model synthesis unit 50 may execute the synthesis process in response to an operation of an operator or the like on the input content specification unit 30 and / or the virtual machine performance model identification unit 42.

  In other words, the performance prediction apparatus 10 uses the system performance model 510 generated by the system performance model synthesis unit 50 to perform the performance that is a result of predicting the performance of the virtualization system 2 based on the request contents to each virtual machine. An index can be calculated. The operator 9 can predict the performance of the virtualization system 2 by referring to such a performance index via, for example, a display.

(Performance model generation process)
Next, the performance model generation process performed by the performance prediction apparatus 10 will be described with reference to FIGS.

  FIG. 6 is a flowchart showing an outline of the performance model generation process by the performance prediction apparatus 10 according to the second exemplary embodiment of the present invention. As shown in FIG. 6, the performance prediction apparatus 10 performs the following operations from step S101 to step S103.

  Step S101: The measurement target unit performance model generation unit 32 generates a performance model (measurement target unit performance model 310) for the measurement target unit 501 that combines the hardware unit 300 and the second application unit 500 of the virtualization system 2. Process. The generation process of the measurement target part performance model 310 will be described with reference to different drawings.

  Step S102: The virtual machine performance model specifying unit 42 specifies the virtual machine performance model 410 from the performance model group 40 of the storage unit 41 and acquires it from the storage unit 41 in accordance with the operation of the operator 9 or the like. Specifically, first, the virtual machine performance model specifying unit 42 refers to the storage unit 41 using information related to an application running on the target first virtual machine 201. Here, the target first virtual machine 201 is a virtual machine (first virtual machine 201) that operates in the virtualization system 2 and for which details can be obtained. Then, the virtual machine performance model specifying unit 42 specifies a virtual machine performance model 410 corresponding to (matching) the above information from the performance model group 40 stored in the storage unit 41, and the specified virtual machine performance model 410 is acquired.

  Note that steps S101 and S102 may be executed simultaneously or in reverse order.

  Step S103: The system performance model combining unit 50 performs a combining process using the measurement target unit performance model 310 generated in Step S101 and the one or more virtual machine performance models 410 specified in Step S102. Run. Thereby, a performance model for predicting the performance of the virtualization system 2 is generated. The performance model (performance predicted value) obtained as a result of the synthesis process is an output value of the system performance model synthesis unit 50. The processing (synthesizing process) of the system performance model combining unit 50 will be described with reference to different drawings.

  FIG. 7 is a flowchart showing an example of the flow of processing for generating the measurement target part performance model 310 in step S101 described above. As shown in FIG. 7, in the generation process of the measurement target part performance model 310, the following processes from step S201 to step S203 are performed.

  Step S201: The input content specifying unit 30 causes the virtualization system 2 to start measurement in response to the operation of the operator 9 or the like. Then, the virtualization system 2 performs measurement of the virtualization system 2.

  Step S202: The information collection unit 31 collects actual measurement values output by the measurement of the virtualization system 2 performed in step S201. The actual measurement value includes the influence of the operation of the second virtual machine 301 in the hardware unit 300. For example, the hardware unit 300 such as the usage rate, throughput, and TAT of each resource for each month, date, and time. , And information indicating the state of each second virtual machine 301 of the second application unit 500. The information collection unit 31 supplies the collected measured values (the input value input to the virtualization system 2 and the output value corresponding to the input value in step S201) to the measurement target unit performance model generation unit 32.

  Step S203: The measurement target unit performance model generation unit 32 generates the measurement target unit performance model 310 by the black box approach using the actual measurement value supplied from the information collection unit 31. Thereby, the generation process of the measurement target part performance model 310 is ended.

  FIG. 8 is a flowchart showing an example of the flow of the synthesis process performed by the system performance model synthesis unit 50 in step S103 described above. FIG. 9 is a diagram for conceptually explaining the synthesis process performed by the system performance model synthesis unit 50. As shown in FIG. 8, in the synthesis process of the system performance model synthesis unit 50, the following processes from step S301 to step S307 are performed.

  Step S301: The system performance model composition unit 50 inputs the request (600, 601) shown in FIG. 9 to each virtual machine performance model 410 of the first application unit 400.

  Step S302: The system performance model composition unit 50 causes the virtual machine performance model 410 to generate resource requests (602, 603) according to the request input in step S301. In this embodiment, the resource request (602, 603) is, for example, data including information about which hardware resource is used by the process that configures the first virtual machine 201 and how it is used. . The resource request (602, 603) in the present embodiment is the same as the resource request 60 in the first embodiment described above.

  Step S303: The system performance model synthesis unit 50 inputs the resource request (602, 603) generated by the virtual machine performance model 410 in Step S302 to the measurement target unit performance model 310.

  Step S304: The system performance model synthesis unit 50 processes the resource request (602, 603) input in step S303 using the measurement target unit performance model 310.

  Step S305: The system performance model synthesis unit 50 inputs a process completion notification (605, 606) indicating that the process in step S304 is completed from the measurement target unit performance model 310 to the virtual machine performance model 410. When the processing completion notification (605, 606) is output from the measurement target unit performance model 310, the system performance model synthesis unit 50 calculates the performance index (for example, resource usage rate, TAT, etc.) calculated by the measurement target unit performance model 310. 604 is output from the measurement target part performance model 310. The output timing may be calculated at an arbitrary timing. Specifically, for example, the system performance model synthesis unit 50 may calculate the performance index 604 when outputting the processing completion notification (605, 606), or may calculate the performance index 604 every time determined by the operator 9 or the like. Also good.

  Step S306: The system performance model synthesis unit 50 checks whether there is an unprocessed resource request output by the first application unit 400. If an unprocessed resource request remains in first application unit 400 (YES in step S306), system performance model combining unit 50 returns the process to step S303. If processing of all resource requests has been completed (NO in step S306), system performance model composition unit 50 advances the process to step S307. That is, a loop including steps from step S303 to step S306 is repeated until there is no resource request in the first application unit 400.

  Step S307: If it is determined in step S306 that all resource requests have been processed, the system performance model composition unit 50 outputs processed requests (607, 608) from each virtual machine performance model 410.

  Thereby, the operator 9 or the like 9 uses the calculation results (performance index 604 and the like) of the virtual machine performance model 410 and the measurement target unit performance model 310 of the first application unit 400 calculated by the system performance model synthesis unit 50. The performance of the virtualization system 2 can be predicted.

  Note that the processed request output in step S307 may be output after the processing for the corresponding request is completed, that is, after step S305.

  FIG. 9 is a diagram conceptually showing a method of synthesizing the system performance model 510 in the system performance model synthesizing unit 50 shown in FIG. In the first application unit 400 in FIG. 9, requests (600, 601) are input to the virtual machine performance model 410 prepared for each first virtual machine 201. The input request is then processed using a queue and output as processed requests (607, 608). Here, the virtual machine performance model 410 is a model imitating the operation of the application executed in the first application unit 400.

  That is, the request (600, 601) is input to the virtual machine performance model 410 of the first application unit 400 by the operation of the operator 9 or the like. Here, the request (600, 601) is data created in a simulated manner for the simulation using each virtual machine performance model 410. The contents of this data represent the load on each software resource, the use timing of hardware resources, and the like.

  In the present embodiment, when a plurality of first virtual machines 201 are operating, as illustrated in FIG. 9, each of the virtual machine performance models 410 corresponding to each of the plurality of first virtual machines 201 is used. On the other hand, a request is input. Individual requests are input to each virtual machine performance model 410 at the same time or at a time difference designated by an operator 9 or the like, for example.

  In the case of the example illustrated in FIG. 9, when a request (600, 601) is input to the virtual machine performance model 410 of the first application unit 400, the virtual machine performance model 410 performs processing corresponding to the request, Processed requests (607, 608) are output as processing results.

  The system performance model synthesizing unit 50 uses the virtual machine performance model 410 of the first application unit 400 to generate a resource request (602, 603) necessary for processing required by the request from the input request (600, 601). Generate. In the present embodiment, for example, the resource request (602, 603) is a resource (hardware resource) constituting the hardware unit 300 necessary for processing the request input to each virtual machine performance model 410. On the other hand, it is an instruction for requesting a specific process. Here, the resources configuring the hardware unit 300 are, for example, a CPU, a storage device (HDD), a communication network (NIC: Network Interface Card), and the like.

  The system performance model synthesis unit 50 inputs the resource request (602, 603) to the hardware resource generated by the first application unit 400 to the measurement target unit 501. The resource request (602, 603) input to the measurement target unit 501 is processed by the measurement target unit performance model 310. The system performance model synthesis unit 50 outputs a process completion notification (605, 606) from the measurement target unit 501 to the first application unit 400 after the processing by the measurement target unit performance model 310 is completed. The processing completion notifications (605, 606) correspond to the processing completion notification 61 (FIG. 3) regarding the resource request in the first embodiment described above.

  The system performance model synthesizing unit 50 outputs, from the measurement target unit performance model 310, the performance index 604 calculated in the processing for the resource request (602, 603) using the measurement target unit performance model 310 of the measurement target unit 501. The performance index 604 may include information such as the usage rate of each resource and TAT, for example.

  After the first application unit 400 receives the processing completion notification (605, 606) output from the measurement target unit 501, the system performance model synthesis unit 50 sends the processed requests (607, 608) to the virtual machine performance models 410. Output from.

  Note that after the first application unit 400 outputs (issues) the resource request (602, 603) to the measurement target unit 501, a process completion notification (605, 606) is sent from the measurement target unit 501 to the first application unit 400. The time (period) until it is input to the virtual machine performance model 410 is defined as the processing time (processing required time) PT of the virtual machine performance model 410. Regarding the processing time PT, after the resource request (602, 603) output from the first application unit 400 is input to the measurement target unit 501, the measurement target unit 501 notifies the first application unit 400 of processing completion (605). 606) can be regarded as the time (period) until output.

  Using the performance index and the like calculated in the above steps, the system performance model synthesis unit 50 provides information related to performance prediction of the virtualization system 2 to the operator 9 and the like.

  That is, by replacing with the specific virtual machine performance model 410 operating in the first virtual machine 201 of the first application unit 400, the operator 9 and the like 9 The performance of the system 2 can be predicted. Therefore, the operator 9 predicts the performance of the entire virtualization system 2 by using the relationship between the request (600, 601) that is an input value to the virtual machine performance model 410 and the processing time PT for the request. can do.

  Further, the operator 9 and the like 9 can predict the performance of the measurement target unit 501 based on the relationship between the input and output to the measurement target unit performance model 310 in which the measurement target unit 501 is replaced. For this reason, the operator 9 uses the relationship between the resource request (602, 603) that is an input value to the measurement target unit performance model 310 and the performance index 604 that is an output value, and the performance of the measurement target unit 501. Can be predicted.

  Thereby, the operator 9 can predict the performance of the entire virtualization system 2.

As described with reference to FIGS. 8 and 9, values (information) obtained in the synthesis process of the system performance model synthesis unit 50 are as follows. That is,
(A) Request (600, 601) input to each virtual machine performance model 410,
(B) Resource request (602, 603) to the measurement target unit performance model 310 generated by each virtual machine performance model 410,
(C) a performance index 604 calculated by the measurement target part performance model 310;
(D) Processing ownership time.

  The system performance model combining unit 50 may be configured to output at least one of (c) and (d) as the system performance model 510. The system performance model synthesis unit 50 may be configured to output (a) and / or (b) in addition to the above (c) and / or (d).

  That is, by using the measurement target unit performance model 310, the system performance model synthesis unit 50 can calculate the performance index 604 based on the content of the request to the virtualization system 2 that is the target for which the performance is to be predicted. it can. Accordingly, the operator 9 and the like 9 can predict the performance of the virtualization system 2 using the system performance model 510 by referring to the calculated values described above.

  In addition, the performance prediction apparatus 10 of this Embodiment demonstrated the structure provided with the virtual machine performance model specific | specification part 42 instead of the virtual machine performance model production | generation part 4 in 1st Embodiment. However, the performance prediction apparatus 10 according to the present exemplary embodiment may be configured to further include the virtual machine performance model generation unit 4 in the first exemplary embodiment in addition to the configuration illustrated in FIG.

(effect)
As described above, the performance prediction apparatus 10 according to the present embodiment defines the configuration of the virtualization system 2 as the hardware unit 300, the first application unit 400, and the second application unit 500. Then, the performance prediction apparatus 10 generates a measurement target unit performance model 310 of the measurement target unit 501 including the hardware unit 300 and the second application unit 500 and a virtual machine performance model 410 of the first application unit 400. Furthermore, the performance prediction apparatus 10 generates a system performance model 510 of the virtualization system 2 in which the dependency relationship between the respective units is reflected. Thereby, the performance of the virtualization system 2 can be predicted. Therefore, the performance prediction apparatus 10 according to the present embodiment suitably predicts the performance of the virtualization system 2 even when a plurality of virtual machines whose details are not available are operating in the virtualization system 2. Is possible.

<Example of hardware configuration>
Here, a configuration example of hardware capable of realizing the performance prediction apparatus according to each embodiment described above 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. 10 is a diagram illustrating a hardware configuration of a computer (information processing apparatus) capable of realizing each embodiment of the present invention.

  The hardware of the information processing apparatus (computer) 100 shown in FIG. 10 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 according to the above-described embodiments (FIGS. 1 and 5) to communicate with an external device via the communication network 200. In the hardware configuration, the CPU 11 governs the overall operation of the information processing apparatus 100 that realizes the performance prediction apparatuses (1, 10) according to the embodiments.

  The present invention described by taking each of the above-described embodiments as an example supplies, for example, a program (computer program) capable of realizing the processing described in each of the above-described embodiments to the information processing apparatus 100 illustrated in FIG. Then, the program is achieved by reading the program to the CPU 11 and executing it. Note that the program is, for example, the various processes described in the flowcharts (FIGS. 6, 7, and 8) referred to in the description of the above embodiments, or the block diagrams shown in FIGS. It may be a program capable of realizing each unit (each block) shown in the apparatus.

  The program supplied to 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 function of each unit shown in the performance prediction devices (1, 10) in the above-described embodiments, for example. Further, the various types of stored information 17B are, for example, the performance model group 40 of the first application unit 400 and the actual measurement values related to the virtualization system 2 in the above-described embodiments. 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. 5), and those skilled in the art can implement the program unit. You may select suitably.

  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 (200). In such a case, the present invention can be understood to be configured by a code (program group 17A) 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-described embodiments and examples. It will be apparent to those skilled in the art that various modifications and improvements can be made to the embodiment. 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.

  In addition, a part or all of the above-described embodiments can be described as the following supplementary notes. However, the present invention described by way of example with the above-described embodiments is not limited to the following.

  (Supplementary Note 1) A performance prediction apparatus for predicting the performance of an information processing system, wherein the information processing system is capable of obtaining information related to an application operating on a virtual machine, or one or a plurality of first virtual machines A virtualization system having a first application unit including a second application unit including one or a plurality of second virtual machines and a hardware unit in which information on an application operating on the virtual machine cannot be obtained. The performance prediction device includes first performance model generation means for generating a performance model of the first virtual machine based on information about the application of each of the first virtual machines of the first application unit, respectively. The second application unit and the hardware unit are measured as measurement target units. Based on the results, a second performance model generation means for generating a performance model of the measurement target unit, a performance model of each of the first virtual machines of the first application unit, and a performance model of the measurement target unit, On the other hand, the performance model of the information processing system in which the dependency between the two parts is reflected by the synthesis process using the output value in the performance model of one part as the input value in the performance model of the other part. And a system performance model synthesizing unit for generating the performance prediction apparatus.

  (Additional remark 2) The said system performance model synthetic | combination means is a said hardware part contained in the said measurement object part according to a request being input into the performance model of each said 1st virtual machine of the said 1st application part. A resource request including information on how to use the hardware resource that constitutes the first virtual machine is issued to the performance model of the first virtual machine, and the performance model of the measurement target unit performs processing according to the resource request In response to the completion of the process, a process completion notification is issued to the performance model of each first virtual machine, and a synthesis process using the resource request and the process completion notification is performed. The performance prediction apparatus according to Supplementary Note 1.

  (Additional remark 3) The said system performance model synthetic | combination means is a performance index calculated by the performance model of the said measurement object part as a performance model of the said information processing system, and the said resource request | requirement with respect to the performance model of the said measurement object part The performance prediction apparatus according to appendix 2, wherein at least one of the possession time from when the process is issued to when the process completion notification is issued from the performance model of the measurement target unit is presented.

  (Additional remark 4) The said hardware part contains the virtual machine monitor and the hardware of a physical machine, The performance prediction apparatus in any one of Additional remark 1 to 3 characterized by the above-mentioned.

  (Supplementary Note 5) The second virtual machine includes at least one of a virtual machine for which details of an application in the second virtual machine cannot be obtained and a virtual machine owned by a third party. The performance prediction apparatus according to any one of appendices 1 to 4.

  (Additional remark 6) The said 2nd performance model production | generation means measures the said information processing system, and is based on the input value used for the said measurement, and the output value obtained by the said measurement corresponding to the said input value The performance prediction apparatus according to any one of appendices 1 to 5, wherein a performance model of the measurement target unit is generated.

  (Supplementary note 7) The performance prediction apparatus according to any one of supplementary notes 1 to 6, wherein the measurement with respect to the measurement target unit is performed for each unit on a time axis having a different usage frequency tendency.

  (Supplementary Note 8) Instead of the first performance model generation means or in addition to the first performance model generation means, it is generated based on design information of an application operating on the first virtual machine. The apparatus according to any one of appendices 1 to 7, further comprising performance model specifying means for selecting a specific performance model from storage means storing one or more performance models of the first application unit. Performance prediction device.

  (Additional remark 9) When the said 2nd virtual machine operate | moves, the setting value regarding the measurement of the said information processing system, the input content designation | designated means which can designate the execution command of the said measurement, The setting value of the said information processing system Information collecting means for collecting the input value and the output value corresponding to the input value, wherein the second performance model generating means is based on the input value and the output value, The performance prediction apparatus according to any one of appendices 1 to 8, wherein a performance model of the measurement target unit is generated.

  (Supplementary Note 10) A performance model generation method in a performance prediction apparatus for predicting the performance of an information processing system, wherein the information processing system is capable of obtaining information on an application operating on a virtual machine. A first application unit including a first virtual machine; a second application unit including one or more second virtual machines for which information relating to an application running on the virtual machine cannot be obtained; and a hardware unit. In the virtualization system, the performance model generation method generates a performance model of the first virtual machine based on information on the application of each of the first virtual machines of the first application unit, and The second application part and the hardware part are the measurement target parts, and Based on the measurement result of the target unit as a measurement target, a performance model of the measurement target unit is generated, a performance model of each of the first virtual machines of the first application unit, and a performance model of the measurement target unit, On the other hand, by using the output value in the performance model of one part as the input value in the performance model of the other part, the performance model of the information processing system that reflects the dependency between the two parts is generated. A performance model generation method characterized by:

  (Supplementary Note 11) A program to be executed by a computer including a performance prediction device for predicting the performance of an information processing system, wherein the information processing system can obtain information on an application operating on a virtual machine, A first application unit including a plurality of first virtual machines, a second application unit including one or a plurality of second virtual machines for which information relating to an application running on the virtual machine cannot be obtained, a hardware unit, A process for generating a performance model of the first virtual machine based on information on the application of each of the first virtual machines in the first application unit, respectively. Measure the second application part and the hardware part And a process for generating a performance model of the measurement target unit based on a measurement result of the measurement target unit as a measurement target, a performance model of each first virtual machine of the first application unit, and the measurement The information in which the dependency relationship between the two parts is reflected by the synthesis process using the output value in the performance model of one part as the input value in the performance model of the other part with respect to the performance model of the target part A program for causing a computer to execute processing for generating a performance model of a processing system.

  (Supplementary note 12) A computer-readable recording medium characterized by storing the program according to Supplementary note 11.

DESCRIPTION OF SYMBOLS 1 Performance prediction apparatus 2 Virtualization system 3 Measurement object part performance model generation part 4 Virtual machine performance model generation part 5 System performance model synthesis part 10 Performance prediction apparatus 30 Input content designation part 31 Information collection part 32 Measurement object part Performance model generation part 40 Performance Model Group 41 Storage Unit 42 Virtual Machine Performance Model Specification Unit 50 System Performance Model Synthesis Unit 60 Resource Request 61 Processing Completion Notification 201 First Virtual Machine 202 Virtual Machine Monitor 203 Physical Machine Hardware 300 Hardware Unit 301 Second Virtual machine 310 measurement target part performance model 400 first application part 410 virtual machine performance model 500 second application part 501 measurement target part 510 system performance model 604 performance index

Claims (10)

A performance prediction device for predicting the performance of an information processing system,
The information processing system cannot obtain information on an application running on a virtual machine, a first application unit including one or more first virtual machines that can obtain information on an application running on the virtual machine, A virtualization system including a second application unit including one or a plurality of second virtual machines, and a hardware unit;
The performance prediction apparatus is
First performance model generation means for generating a performance model of the first virtual machine based on information on the application of each of the first virtual machines of the first application unit;
A second performance model generating unit configured to generate a performance model of the measurement target unit based on a result of measurement using the second application unit and the hardware unit as a measurement target unit and the measurement target unit as a measurement target;
For the performance model of each first virtual machine of the first application unit and the performance model of the measurement target unit, an output value in the performance model of one unit is used as an input value in the performance model of the other unit A performance prediction apparatus comprising: a system performance model synthesis unit that generates a performance model of the information processing system in which a dependency relationship between the two parts is reflected by a synthesis process to be used. The system performance model synthesis means includes:
Information on how to use hardware resources constituting the hardware unit included in the measurement target unit in response to a request input to the performance model of each first virtual machine of the first application unit A resource request including the request is issued to the performance model of the first virtual machine,
While allowing the performance model of the measurement target unit to perform processing according to the resource request, in response to completion of the processing, to issue a processing completion notification to the performance model of each first virtual machine,
The performance prediction apparatus according to claim 1, wherein a synthesis process using the resource request and the process completion notification is performed. The system performance model synthesizing means, as a performance model of the information processing system,
A performance index calculated by a performance model of the measurement target unit;
Presenting at least one of the possession time from when the resource request is issued to the performance model of the measurement target unit to when the processing completion notification is issued from the performance model of the measurement target unit The performance prediction apparatus according to claim 2, wherein the performance prediction apparatus is characterized.   The said hardware part contains the virtual machine monitor and the hardware of a physical machine, The performance prediction apparatus of any one of Claim 1 to 3 characterized by the above-mentioned.   The second performance model generation means measures the information processing system, and based on the input value used for the measurement and the output value obtained by the measurement corresponding to the input value, 5. The performance prediction apparatus according to claim 1, wherein a performance model of a part is generated.   6. The performance prediction apparatus according to claim 1, wherein the measurement of the measurement target unit is performed for each unit on a time axis having a different usage frequency tendency.   Instead of the first performance model generation means, or in addition to the first performance model generation means, the first performance model is generated based on design information of an application operating in the first virtual machine. 7. The apparatus according to claim 1, further comprising a performance model specifying unit that selects a specific performance model from a storage unit that stores one or more performance models of the application unit. 8. Performance prediction device. A setting value related to the measurement of the information processing system when the second virtual machine operates, and an input content specifying means capable of specifying an execution instruction of the measurement;
An information collection means for collecting an input value that is a setting value of the information processing system and an output value corresponding to the input value;
The said 2nd performance model production | generation means produces | generates the performance model of the said measurement object part based on the said input value and the said output value, The any one of Claim 1 to 7 characterized by the above-mentioned. The performance prediction apparatus described. A performance model generation method in a performance prediction device for predicting the performance of an information processing system,
The information processing system cannot obtain information on an application running on a virtual machine, a first application unit including one or more first virtual machines that can obtain information on an application running on the virtual machine, A virtualization system including a second application unit including one or a plurality of second virtual machines, and a hardware unit;
The performance model generation method includes:
Based on the information related to the application of each of the first virtual machines in the first application unit, respectively, generate a performance model of the first virtual machine,
Based on the measurement result of the second application unit and the hardware unit as a measurement target unit and the measurement target unit as a measurement target, a performance model of the measurement target unit is generated,
For the performance model of each first virtual machine of the first application unit and the performance model of the measurement target unit, an output value in the performance model of one unit is used as an input value in the performance model of the other unit A performance model generation method characterized by generating a performance model of the information processing system in which the dependency relationship between the two parts is reflected by using. A program to be executed by a computer including a performance prediction device for predicting the performance of an information processing system,
The information processing system cannot obtain information on an application running on a virtual machine, a first application unit including one or more first virtual machines that can obtain information on an application running on the virtual machine, A virtualization system including a second application unit including one or a plurality of second virtual machines, and a hardware unit;
The program is
Processing for generating a performance model of the first virtual machine based on information on the application of each of the first virtual machines in the first application unit;
A process of generating a performance model of the measurement target unit based on a result of measurement using the second application unit and the hardware unit as a measurement target unit and the measurement target unit as a measurement target;
For the performance model of each first virtual machine of the first application unit and the performance model of the measurement target unit, an output value in the performance model of one unit is used as an input value in the performance model of the other unit A program for causing a computer to execute a process of generating a performance model of the information processing system in which a dependency relationship between the two parts is reflected by a combining process to be used.

Images