JP5691529B2 - Performance evaluation system, performance evaluation method and performance evaluation program - Google Patents

Description

  The present invention relates to a performance evaluation system, a performance evaluation method, and a performance evaluation program for evaluating the performance of a virtualized system.

  When designing an IT (Information Technology) system, it is necessary to evaluate and verify whether the design satisfies the required performance and availability. Methods for modeling the system to be evaluated and evaluating performance, reliability, availability, etc. have been studied for a long time. Recently, a method for evaluating performance and reliability by describing the configuration and behavior of a system in UML (Unified Modeling Language), SysML (Systems Modeling Language), or a similar modeling language has been proposed.

  For example, the performance evaluation model generation method described in Patent Document 1, the UML design method described in Patent Document 2, the method of converting to the probabilistic performance evaluation model described in Patent Document 3, and the Patent Document 4 All of the described performance evaluation methods are methods for evaluating performance on a UML basis.

  In the method of evaluating performance based on the UML, a performance model (typically a queuing model) is generated from the behavior of the system described by the UML sequence diagram and activity diagram, and analysis and simulation are performed. UML employs graphic description as a model notation, and it is easy for general system engineers (SEs) to understand and model.

  On the other hand, hardware virtualization technology using VMWare (registered trademark), Xen (registered trademark), etc. can flexibly and efficiently arrange servers, so that the required performance and It is possible to meet availability.

  Patent Document 5 describes a performance value calculation device that calculates a performance value representing performance according to the amount of resources allocated to a virtual machine. In the performance value calculating apparatus described in Patent Document 5, the CPU usage rate is calculated based on the CPU execution time tcpu and the disk execution time tdisk according to the upper limit value α1 of the CPU usage rate, the CPU execution time tcpu, and the disk execution time tdisk. And the second performance value calculated based on the upper limit value α1 of the CPU usage rate and the CPU execution time tcpu is selected as the performance value.

  Patent Document 6 describes a computer design support system. In the system described in Patent Literature 6, a know-how database stores a plurality of system configuration diagrams showing the system configuration for each model and the load status collected for each business type, and a necessary system based on this load status Predict performance and scale.

JP 2001-318812 A JP 2005-327094 A JP 2007-179165 A JP 2007-188179 A JP 2010-9160 A JP 2002-222227 A

  However, even if a general method is used, there is a problem that it is difficult to estimate the performance of a virtualized system at the application design stage.

  In the method for evaluating performance based on UML, the performance is estimated based on design information describing the flow of processing between servers. However, in this method, only the control flow between hardware devices is modeled, and the overhead due to the use of the virtualization technology is not considered.

  Further, in the computer design support system described in Patent Document 6, information related to a specific system configuration such as a network configuration and load distribution is accumulated as know-how data. However, the computer design support system described in Patent Document 6 does not consider the overhead caused by using the virtualization technology.

  On the other hand, in the performance value calculation device described in Patent Document 5, a performance evaluation function and model that takes into account the overhead due to the use of the virtualization technology is given, so it is possible to estimate the performance of the virtualized system. is there. However, when a general SE performs performance evaluation using these functions and models, it is necessary to calculate the overhead of virtualization based on the design of the system, which may require troublesome work. Therefore, it is desirable that the performance of a virtual system can be easily estimated even when a general SE performs performance evaluation.

  Therefore, an object of the present invention is to provide a performance evaluation system, a performance evaluation method, and a performance evaluation program that can easily estimate the performance of a virtual system.

The performance evaluation system according to the present invention includes an object on the path and a connection relationship between the objects as elements of a data flow that is a path of data transmitted and received between processes defined on the IT system. Process data flow model, which is a model in which attributes indicating the nature of data flow are set, virtual server model describing the connection relationship between objects in a virtualized system, and connection relationship between objects in a physical system The physical server model that describes the virtual server model, virtual realization relationship information that indicates the correspondence between the elements of the process data flow model and the virtual server model object, and the correspondence between the virtual server model object and the physical server model object Physics realization related information that is information indicating And the system model storage unit that, based on the rules for calculating the performance of the physical system, the attributes set in the elements of the process data flow model specified by the virtual realization relation information and physical realization relationship information, virtualized And a performance evaluation model generating means for generating a performance evaluation model for evaluating the performance of the system, and a performance evaluation means for evaluating the performance of the virtualized system based on the performance evaluation model.

The performance evaluation method according to the present invention includes, as elements of a data flow, which is a path of data transmitted and received between processes defined on the IT system, and objects on the path and connection relationships between the objects. Information indicating the correspondence between the element in the process data flow model, which is a model in which an attribute indicating the nature of the data flow is set in the element, and the object in the virtual server model describing the connection relationship between the object in the virtualized system Specified by the virtual realization relationship information and the physical realization relationship information, which is information indicating the correspondence between the virtual server model object and the physical server model object describing the connection relationship between the objects in the physical system. to the elements of the process data flow model From the specified attributes, a performance evaluation model that evaluates the performance of the virtualized system is generated based on the rules for calculating the performance of the physical system, and the performance of the virtualized system based on the performance evaluation model is generated. It is characterized by performing evaluation.

The performance evaluation program according to the present invention includes, as elements of a data flow, which is a data path transmitted / received between processes defined on the IT system, and objects on the path and a connection relationship between the objects. Process data flow model, which is a model in which attributes indicating the characteristics of data flow are set in elements, virtual server model that describes the connection relationship between objects in a virtualized system, and connections between objects in a physical system Correspondence between the physical server model describing the relationship, the virtual realization relationship information that is the information indicating the correspondence between the elements of the process data flow model and the object of the virtual server model, and the object of the virtual server model and the object of the physical server model Physical realization relationship information that is information indicating the relationship A performance evaluation program that is applied to a computer having a system model storage means for storing, and is specified by the virtual realization relation information and the physical realization relation information based on the rules for calculating the physical system performance in the computer. Performance evaluation model generation processing for generating a performance evaluation model for evaluating the performance of a virtualized system from attributes set in the elements of the process data flow model to be processed, and a system virtualized based on the performance evaluation model The performance evaluation process for performing the performance evaluation is executed.

  According to the present invention, the performance of a virtual system can be easily estimated.

It is a block diagram which shows the example of the performance evaluation system in the 1st Embodiment of this invention. It is explanatory drawing which shows the example at the time of describing the process data flow model using a sequence diagram. It is explanatory drawing which shows an example of the attribute with respect to each element of a process data flow model. It is explanatory drawing which shows an example of a virtual server model. It is explanatory drawing which shows an example of virtual realization relationship. It is explanatory drawing which shows an example of a physical server model. It is explanatory drawing which shows an example of physical realization relationship. It is explanatory drawing which shows an example of virtualization performance DB. It is explanatory drawing which shows an example of physical performance DB. It is a flowchart which shows the example of operation | movement of the performance evaluation system in this embodiment. It is explanatory drawing which shows the example which converted the data flow in a process data flow model into the data flow in the virtual server model. It is explanatory drawing which shows the example of the correspondence of the data flow in a process data flow model, and a virtual data flow. It is explanatory drawing which shows the example which converted the virtual data flow into the physical data flow. It is explanatory drawing which shows the example of the correspondence of a virtual data flow and a physical data flow. It is explanatory drawing which shows the example which calculates the load of each element in a physical data flow. It is a block diagram which shows the example of the performance evaluation system in the 2nd Embodiment of this invention. It is explanatory drawing which shows the example of a dynamic physical realization relationship. It is a block diagram which shows the example of the minimum structure of the performance evaluation system by this invention.

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

Embodiment 1. FIG.
FIG. 1 is a block diagram showing an example of a performance evaluation system in the first embodiment of the present invention. The performance evaluation system in this embodiment includes a system model storage unit 10, a performance evaluation model generation unit 20, a virtualization performance storage unit 30, a physical performance storage unit 40, and a performance evaluation unit 50.

  The system model storage unit 10 stores a process data flow model 11, virtual realization relationship information 12, a virtual server model 13, physical realization relationship information 14, and a physical server model 15. The process data flow model 11, the virtual realization relation information 12, the virtual server model 13, the physical realization relation information 14, and the physical server model 15 are stored in advance in the system model storage unit 10.

  The process data flow model 11 is a model that describes data exchange (flow) between processes on the IT system. Specifically, the process data flow model 11 is an element of a data path transmitted and received between processes defined on the IT system (hereinafter, referred to as a data flow) as an element between objects on the path. It is a model that includes connection relationships. Further, an attribute indicating the performance (property) of the data flow is set in each element of the process data flow model 11.

  The virtual server model 13 is a model that describes the configuration of a virtual system. Specifically, the virtual server model 13 is a model representing a connection relationship between virtualized objects.

  The physical server model 15 is a model that describes the configuration of a physical system. Specifically, the physical server model 15 is a model that represents a connection relationship between physical objects (for example, a server device and a client terminal).

  The virtual realization relationship information 12 is information describing the correspondence between each element of the process data flow model 11 and each object of the virtual server model 13.

  The physical realization relationship information 14 is information describing the correspondence between each object of the virtual server model 13 and each object of the physical server model 15.

  Hereinafter, a specific example of information stored in the system model storage unit 10 will be described by taking as an example a case where performance evaluation of a Web three-layer system is performed.

  FIG. 2 is an explanatory diagram showing an example of the process data flow model 11. The process data flow model 11 can be described by a diagram representing data behavior such as a UML sequence diagram and an activity diagram. FIG. 2 is an explanatory diagram showing an example when the process data flow model 11 is described using a sequence diagram.

  In the example shown in FIG. 2, there are four processes “: client”, “: web”, “: ap”, and “: db” on the system, “: client” becomes “: web”, and “: web”. “web” is sent to “: ap”, “: ap” is sent to “: db”, and the messages m1, m2, and m3 are sent in this order, and then “: db” is sent to “: ap” and “: ap” is sent. ": Web" and ": web" to ": client" indicate that messages m4, m5, and m6 are returned in this order. “: Client” corresponds to a Web client, “: web” corresponds to a Web server, “: ap” corresponds to a Web application server, and “: db” corresponds to a database server. Each process “: web”, “: ap”, The processes performed at “: db” are e1, e2, and e3, respectively.

  As described above, attributes representing the data flow in detail can be set for each element of the process data flow model 11. Specifically, information indicating the system load is set in the attribute of each element of the data flow. FIG. 3 is an explanatory diagram illustrating an example of attributes set in each element of the process data flow model 11 illustrated in FIG. FIG. 3A shows an example of the average size of each message. The average size illustrated in FIG. 3A is the number and size of image files arranged on the Web server, the size of an HTML (HyperText Markup Language) file generated on the Web application server, and issued to the database. This is determined by the content of the query to be executed.

  As the average message size, for example, a value estimated by the SE from the specifications of the Web application may be used. Further, when a Web application is already installed, the actually measured message size may be used as the average message size.

  FIG. 3B shows an average time (CPU time) required for processing a request in each server. The example shown in FIG. 3B indicates that it takes 5 milliseconds for the Web server, 10 milliseconds for the Web application server, and 20 milliseconds for the database server to process one request. Similarly to the average message size, the average time may be the time estimated by the SE or the actually measured time.

  In the case of the average time (CPU time), the value is different between a high-speed computer and a low-speed computer even if the processing is the same, unlike the message size. Therefore, when a computer to be used as a reference is selected and the processing time actually measured on the computer is set as the average CPU time, the time corrected by the CPU benchmark value may be set as the average CPU time. For example, when the CPU benchmark value of the reference computer is 1.0 and the CPU benchmark value of the computer used for measurement is 0.5, the average of 5 milliseconds, which is a value obtained by multiplying the measurement value by 0.5 CPU time.

  Attributes for each element of the process data flow model 11 can be set in various forms. For example, when the process data flow model 11 is a UML sequence diagram, the attribute values shown in FIG. 3 may be set in each element of the process data flow model 11 using, for example, tagged values. As described above, the average message size and the average CPU time exemplified in FIG. 3 are set as information indicating the system load.

  FIG. 4 is an explanatory diagram illustrating an example of the virtual server model 13. The virtual server model 13 can be described by a diagram representing a connection relationship between objects such as a UML composite configuration diagram and a SysML internal block diagram. That is, examples of the elements of the virtual server model 13 include an object representing each server and a connection relationship. In the example shown in FIG. 4, the virtual system uses “: vCLI” as a client machine, “: vRouter” as a router, “: vWEB” as a Web server machine, “: vSW” as a switch, and “: vSW” as an application server machine. : VAP "is provided as a database server machine, and each object is connected with a connection relationship vc1, vc2, vc3, vc4, vc5.

  FIG. 5 is an explanatory diagram showing an example of the virtual realization relationship information 12. In the virtual realization relationship information 12 illustrated in FIG. 5, each element of the process data flow model 11 operates as which element (object) of the virtual server model 13 in a corresponding table format. The virtual realization relationship information 12 can be described by, for example, a UML deployment diagram or a SysML allocation. Note that SysML supports notation of allocation in a tabular format, and the virtual realization relationship information 12 can be described with almost the same notation as illustrated in FIG. In the example illustrated in FIG. 5, for example, the process “: web” is executed on the server machine “: vWEB”, and the process “: ap” is executed on another server machine “: vAP”.

  FIG. 6 is an explanatory diagram showing an example of the physical server model 15. The physical server model 15 can be described by a diagram representing a connection relationship between objects such as a UML composite configuration diagram and a SysML internal block diagram. In the example illustrated in FIG. 6, the physical system includes “: pCLI” as a client machine, “: pRouter” as a router, “: pSV” as a server machine, and “: pDB” as a database server machine, It shows that each machine has a connection relationship indicated by a solid line in the figure. Also, an attribute indicating the type of virtualization mechanism (hypervisor) is set in the server machine “: pSV” illustrated in FIG. 6, and this server represents that server virtualization is performed with Xen.

  FIG. 7 is an explanatory diagram showing an example of the physical realization relationship information 14. The physical realization relationship information 14 illustrated in FIG. 7 represents which element (object) of the physical server model 15 corresponds to each element (object) of the virtual server model 13. In the example illustrated in FIG. 7, for example, “: vWEB” and “: vAP” on the virtual server model 13 operate on the same physical server “: pSV”.

  The virtualization performance storage means 30 stores a calculation formula (hereinafter referred to as a virtualization overhead function) for calculating performance in consideration of overhead due to virtualization as a database. Hereinafter, this database is referred to as a virtualization performance DB. The virtualization overhead function is defined as a function having information indicating the system load set for each element of the process data flow model 11 as an argument.

  FIG. 8 is an explanatory diagram of an example of the virtualization performance DB. The virtualization performance DB illustrated in FIG. 8 is a database that holds a function (virtualization overhead function) that receives a nominal load and outputs a substantial load for each type of hypervisor. Here, the nominal value means a value indicating the system load set in each element of the process data flow model 11, and the real value means a value indicating the load of each object in the physical server model 15. .

  The example shown in FIG. 8 indicates that the virtualization overhead function for calculating the actual CPU time CPU_r is defined by the nominal CPU time CPU_n and the nominal message size MSIZE_n. Specifically, the actual CPU time is calculated by adding the sum of CPU_n and a value obtained by multiplying MSIZE_n by a certain coefficient. That is, in the virtualization overhead function illustrated in FIG. 8, processing required for network transmission is added as a CPU load on the physical server. Load information set for each element on the process data flow model 11 is used for the nominal CPU time CPU_n and the nominal message size MSIZE_n.

  The physical performance storage unit 40 stores physical performance specifications of each element in the physical server model 15 as a database. Hereinafter, this database is referred to as a physical performance DB. This performance specification is set for each element in the physical server model 15. FIG. 9 is an explanatory diagram showing an example of the physical performance DB. The example illustrated in FIG. 9 indicates that the throughput of the router pRouter and the CPU benchmark value of each server are included in the physical performance DB.

  Note that, unlike the system model, the virtualization performance DB and the physical performance DB have the same value regardless of the system configuration. Therefore, it is desirable that SEs are not set individually but are prepared in advance as a common database and shared by a plurality of SEs. Calculation of information to be stored in the virtualization performance DB and physical performance DB, and input of information to each DB requires specialized knowledge, so it is expected that a lot of man-hours will be required. Therefore, by preparing in advance as a common database, it is possible to reduce the man-hours that a general SE inputs individually.

  The performance evaluation model generation means 20 generates a performance evaluation model used for system performance evaluation based on a rule for calculating physical system performance from attributes in the process data flow model 11. Note that the performance evaluation model generation means 20 in the first embodiment includes a virtual data flow calculation means 21, a physical data flow calculation means 22, a load calculation means 23, and a usage rate calculation means 24.

  The virtual data flow calculation means 21 calculates the data flow (flow) in the virtual server model from the process data flow model 11, the virtual realization relationship information 12, and the virtual server model 13. Hereinafter, the data flow in the virtual server model is referred to as a virtual data flow. Specifically, the virtual data flow calculation unit 21 specifies the correspondence between the elements in the process data flow model 11 and the objects in the virtual server model 13 based on the virtual realization relationship information 12, and is defined in the process data flow model 11. A flow (virtual data flow) on the virtual server model 13 corresponding to the data flow to be calculated is calculated.

  The physical data flow calculation unit 22 calculates the data flow in the physical server model from the virtual data flow, the virtual server model 13, the physical realization relationship information 14, and the physical server model 15. Hereinafter, the data flow in the physical server model is referred to as a physical data flow. Specifically, the physical data flow calculation unit 22 specifies the correspondence between the object in the virtual server model 13 and the object in the physical server model 15 based on the physical realization relationship information 14 and is defined in the virtual server model 13. A flow (physical data flow) on the physical server model 15 corresponding to the virtual data flow is calculated.

  The load calculation unit 23 calculates the load applied to each element in the physical server model 15 from the physical flow and the virtualization performance DB. Specifically, the load calculating unit 23 determines the correspondence between the elements in the process data flow model 11 and the objects in the virtual server model 13 and the correspondence between the objects in the virtual server model 13 and the objects in the physical server model 15. The elements in the process data flow model 11 corresponding to the physical data flow are specified. Then, the load calculation unit 23 calculates the load of each physical data flow based on the load information set for each element in the process data flow model 11. Furthermore, the load calculation unit 23 calculates the load applied to each element in the physical server model 15 based on the calculated load of each physical data flow and the virtualization overhead function defined in the virtualization performance DB.

  The usage rate calculation unit 24 calculates the usage rate of each element in the physical server model 15 from the load calculated by the load calculation unit 23 and the physical performance DB. The result calculated by the usage rate calculating means 24 is hereinafter referred to as a performance evaluation model.

  The performance evaluation unit 50 performs system performance evaluation based on the performance evaluation model generated by the performance evaluation model generation unit 20. The performance evaluation unit 50 outputs a performance evaluation result 60. The performance evaluation unit 50 performs system performance evaluation using, for example, a queue model. However, the method by which the performance evaluation unit 50 performs the performance evaluation is not limited to the queue model. In addition, the performance evaluation unit 50 may perform performance evaluation from the maximum throughput of the entire system, the response time, the usage rate of each element, and the like.

  The virtual data flow calculation unit 21, the physical data flow calculation unit 22, the load calculation unit 23, the usage rate calculation unit 24, and the performance evaluation unit 50 are executed by a CPU of a computer that operates according to a program (performance evaluation program). Realized. For example, the program is stored in a storage unit (not shown) of a device that performs performance evaluation, and the CPU reads the program, and according to the program, a virtual data flow calculation unit 21, a physical data flow calculation unit 22, and a load calculation unit. 23, the usage rate calculation means 24, and the performance evaluation means 50 may be operated. Further, the virtual data flow calculation unit 21, the physical data flow calculation unit 22, the load calculation unit 23, the usage rate calculation unit 24, and the performance evaluation unit 50 may be realized by dedicated hardware. Good.

  Next, the operation will be described. FIG. 10 is a flowchart showing an example of the operation of the performance evaluation system in the present embodiment.

  First, the virtual data flow calculation means 21 converts the data flow in the process data flow model 11 into a data flow in the virtual server model 13 (step S101). Specifically, when the message m transmitted from the element a1 to the element a2 is defined in the process data flow model 11 and the elements a1 and a2 are realized by the elements b1 and b2 in the virtual server model 13, respectively, The virtual data flow calculation means 21 defines the shortest path between b1 and b2 as a data flow (virtual data flow) in the virtual server model 13 corresponding to the message m. The correspondence relationship between the elements in the process data flow model 11 and the objects in the virtual server model 13 is specified based on the virtual realization relationship information 12.

  FIG. 11 is an explanatory diagram illustrating an example in which the data flow in the process data flow model 11 illustrated in FIG. 2 is converted into a data flow in the virtual server model 13 illustrated in FIG. For example, the message m3 illustrated in FIG. 2 represents a data flow from “: ap” to “: db”. Elements on the virtual server model 13 that realize “: ap” and “: db” are “: vAP” and “: vDB”, respectively. Therefore, the data flow (virtual data flow) on the virtual server model 13 corresponding to the message m3 can be represented by a sequence [v6, u6, v7, u7, v8].

  FIG. 12 is an explanatory diagram illustrating an example of a correspondence relationship between a data flow and a virtual data flow in the process data flow model 11. By performing the process of step S101, the data flows on the virtual server model 13 corresponding to the data flows defined in the process data flow model 11 can be listed as a conversion table as illustrated in FIG.

  Next, the physical data flow calculation unit 22 converts the data flow in the virtual server model 13 into a data flow on the physical server model 15 (step S102). Specifically, it is assumed that elements p1 and p2 exist on the physical server model 15. Here, in the virtual server model 13, when the virtual data flow transmitted between the two connected elements v1 and v2 is vc, the elements v1 and v2 in the virtual server model 13 are elements p1 and p1, respectively. When p2 is realized, the physical data flow calculation unit 22 defines the shortest path between p1 and p2 as a data flow (physical data flow) in the physical server model 15 corresponding to the virtual data flow vc. The correspondence relationship between the object in the virtual server model 13 and the object in the physical server model 15 is specified based on the physical realization relationship information 14.

  FIG. 13 is an explanatory diagram illustrating an example in which the virtual data flow illustrated in FIG. 11 is converted into a physical data flow. FIG. 14 is an explanatory diagram illustrating an example of a correspondence relationship between a virtual data flow and a physical data flow. The conversion from the virtual data flow to the physical data flow can be expressed as a conversion table between the virtual data flow and the physical data flow, as illustrated in FIG.

  Next, the load calculation unit 23 calculates each load of the physical data flow on the physical server model 15. The load calculation means 23 first calculates the load corresponding to each element of the physical data flow from the load (see FIG. 3) set for each element of the process data flow model 11. The load corresponding to each element of the physical data flow reverses the conversion table of the virtual data flow and the physical data flow (see FIG. 14) and the conversion table of the process data flow and the virtual data flow (see FIG. 12). Can be obtained by:

  FIG. 15 is an explanatory diagram illustrating an example of calculating the load of each element in the physical data flow. For example, the elements of the virtual data flow corresponding to the physical data flow q2 are v3, v4, v5, v6, u3, u5 from FIG. 14, and the elements of the process data flow corresponding to these elements are from FIG. , M1, m2, m2, m3, e1, e2. Also, from FIG. 3, the message size is set for m1, m2, and m3, and the CPU time is set for e1 and e2. Therefore, when the message size and the CPU time are added, the load corresponding to q2 is 9 (= 1 + 3 + 3 + 2) KB, 15 (5 + 10) ms.

  Next, the load calculation unit 23 calculates the load applied to the elements (“: pCLI”, “: pRouter”, “: pSV”, “: pDB”) in the physical server model 15. The load calculation means 23 uses the virtualization overhead function shown in FIG. In the example shown in FIG. 8, the CPU time of an element is represented by the sum of the CPU time of flows in the element and the value obtained by multiplying the message size of flows in and out of the element by a coefficient. Is done.

  In this case, the CPU time of “: pSV” is calculated as 15 ms (q2 CPU time) + 0.0001 * 37 KB (total message size of p2, p3, p6, p7, q2, q3) = 18.7 ms. (Step S103).

  Next, the usage rate calculation unit 24 refers to the physical performance DB illustrated in FIG. 9 and calculates the usage rate per request from the load information of each element in the physical server model (step S4). For example, the CPU usage rate can be calculated by multiplying the CPU time by the CPU benchmark value and dividing by 1000 (that is, converting to a usage rate per second). The usage rate of the switch or router can be calculated by dividing the message size by the throughput.

  Finally, the performance evaluation means 50 performs performance evaluation based on the usage rate per request of each element in the physical server model 15 and the arrival rate (frequency) of requests input to the system. The performance evaluation may be calculated analytically using a known queuing model or may be calculated by simulation. For example, the performance evaluation unit 50 performs performance evaluation on the maximum throughput of the entire system, response time, usage rate of each element, and the like, and presents the performance evaluation result 60 to the SE.

  As described above, according to the present embodiment, the performance evaluation model generation unit 20 (more specifically, the virtual data flow calculation unit 21, the physical data flow calculation unit 22, and the load calculation unit 23) From the attributes in the process flow data model specified by the virtual realization relation information 12 and the physical realization relation information 14 based on the rules for calculating A performance evaluation model to be evaluated is generated, and the performance evaluation unit 50 performs system performance evaluation based on the performance evaluation model. At that time, the virtual data flow calculation means 21 calculates a virtual data flow on the virtual server model 13 from the data flow of the process data flow model 11 based on the virtual realization relation information 12. Further, the physical data flow calculation means 22 calculates a physical data flow on the physical server model 15 from the virtual data flow based on the physical realization relationship information 14. Then, the load calculation unit 23 calculates the load of the object in the physical server model from the physical data flow based on the attribute set in the element in the process data flow model 11. With such a configuration, it is possible to easily estimate the performance of the virtual system.

  Specifically, the SE stores the process data flow model 11, the virtual server model 13, the physical server model 15, the virtual realization relation information 12, and the physical realization relation information 14 in the system model storage means 10, and the performance evaluation means 50 By using such information, it is possible to evaluate the performance when the system is operated on the virtual system.

  When performing the performance evaluation, the SE does not need to know the details of the overhead due to virtualization, and only has to input design information created using UML or a similar modeling language. As a result, the performance evaluation process is greatly facilitated. This is because, based on the virtualization DB created in advance as information considering the virtualization overhead, the load that is finally applied to the physical server is changed by the performance evaluation model generation means 20 by the process data flow model 11, the virtual server model. 13 because it is automatically calculated from the physical server model 15, the virtual realization relation information 12, and the physical realization relation information 14.

Embodiment 2. FIG.
FIG. 16 is a block diagram illustrating an example of a performance evaluation system according to the second embodiment of the present invention. In addition, about the structure similar to 1st Embodiment, the code | symbol same as FIG. 1 is attached | subjected and description is abbreviate | omitted. The performance evaluation system in the present embodiment includes a system model storage unit 10a, a dynamic evaluation model generation unit 70, a virtualization performance storage unit 30, a physical performance storage unit 40, and a performance evaluation unit 50.

  The system model storage unit 10 a stores a process data flow model 11, virtual realization relationship information 12, a virtual server model 13, a physical server model 15, and dynamic physical realization relationship information 16. That is, the second embodiment is different from the first embodiment in that the dynamic physical realization relationship information 16 is stored instead of the physical realization relationship information 14. The process data flow model 11, the virtual realization relation information 12, the virtual server model 13, the physical server model 15, and the dynamic physical realization relation information 16 are stored in advance in the system model storage unit 10a.

  The dynamic physical realization relationship information 16 is information describing a plurality of correspondence relationships between each object of the virtual server model 13 and each object of the physical server model 15 in association with a physical system state transition. . That is, the physical realization relationship information 14 in the first embodiment is information indicating a fixed correspondence relationship, but the dynamic physical realization relationship information 16 in the second embodiment changes depending on the situation. It differs from the physical realization relationship information 14 in that it is information indicating a correspondence relationship. As a situation where the correspondence between objects is changed, for example, the usage rate of the server device can be cited. However, the situation in which the correspondence between objects is changed is not limited to the usage rate of the device. In addition, the correspondence may be changed according to the number of accesses from other devices.

  FIG. 17 is an explanatory diagram showing an example of the dynamic physical realization relationship information 16. FIG. 17A is an example of a correspondence table showing the correspondence between the elements (objects) of the virtual server model 13 and the elements (objects) of the physical server model 15 for each of the two types of states σ1 and σ2. FIG. 17B is an example of state transition in which the physical system changes into two types of states σ1 and σ2 according to the usage rate of the object. In the example shown in FIG. 17B, an operation of transitioning to the state σ2 when the usage rate of “: pSV” exceeds 80% and transitioning to the state σ1 when the usage rate of “: pSV” becomes 10% or less. Represents. The state transition in the dynamic physical realization relationship information 16 can be described using, for example, a UML state transition diagram.

  The dynamic evaluation model generation unit 70 includes a virtual data flow calculation unit 21, a physical data flow calculation unit 22, a load calculation unit 23, and a usage rate calculation unit 24. The virtual data flow calculation unit 21 is the same as that in the first embodiment.

  The physical data flow calculation means 22 in this embodiment calculates the physical data flow in each state. That is, the physical data flow calculation unit 22 calculates a physical data flow for each correspondence according to the physical system state. The physical data flow calculation method is the same as the method in the first embodiment.

  Similarly, the load calculation unit 23 calculates the load applied to each element in the physical server model 15 in each state. Further, the usage rate calculation means 24 calculates the usage rate of each element in the physical server model 15 in each state, and evaluates the usage rate of each element and the state transition condition in each state in the dynamic performance evaluation. Output as a model.

  Further, the performance evaluation unit 50 evaluates the dynamic performance evaluation model by simulation or the like, and presents the maximum throughput of the entire system, the response time, the usage rate of each element, and the like as the performance evaluation result 60 to the SE.

  As in the first embodiment, the virtual data flow calculation unit 21, the physical data flow calculation unit 22, the load calculation unit 23, the usage rate calculation unit 24, and the performance evaluation unit 50 are a program (performance evaluation unit). This is realized by a CPU of a computer that operates in accordance with a computer program.

  As described above, according to the present embodiment, the physical data flow calculation unit 22 calculates a physical data flow for each correspondence according to the state of the physical system. Therefore, the performance of the virtual system can be easily estimated even when the method of assigning the virtual server to the physical server changes dynamically.

  In this case, for example, the SE describes the method of allocating the virtual server to the physical server as the dynamic physical realization relation information 16 and stores the dynamic physical realization relation information 16 in the system model storage unit 10a, thereby improving the system performance. Can be easily evaluated. This is because the dynamic physics realization relationship information 16 can be described by a correspondence table and a state transition diagram as illustrated in FIG. 17, so that the input is relatively easy and the performance evaluation process is greatly facilitated. Because.

  Next, an example of the minimum configuration of the present invention will be described. FIG. 18 is a block diagram showing an example of the minimum configuration of the performance evaluation system according to the present invention. The performance evaluation system according to the present invention includes an object on the path and a connection relationship between the objects as elements of a data flow that is a path of data transmitted and received between processes defined on the IT system. Between the process data flow model 81 (for example, the process data flow model 11), which is a model in which attributes indicating data flow characteristics (for example, message size, CPU time, etc.) are set, and objects in the virtualized system A virtual server model 82 (for example, virtual server model 13) describing connection relationships, a physical server model 83 (for example, physical server model 15) describing connection relationships between objects in a physical system, and a process data flow model 81 elements and virtual server model 82 objects Virtual realization relationship information 84 (for example, virtual realization relationship information 12) that is information indicating the correspondence relationship between the virtual server model 82 and physical realization information that is information indicating the correspondence relationship between the object of the virtual server model 82 and the object of the physical server model 83. System model storage unit 80 (for example, system model storage unit 10) that stores relationship information 85 (for example, physical realization relationship information 14), and rules for calculating physical system performance (for example, virtualization overhead function) Based on the above, the performance of the virtualized system is evaluated from the attributes (for example, information indicating the load) set in the elements of the process flow data model 81 specified by the virtual realization relation information 84 and the physical realization relation information 85 Performance evaluation to generate a performance evaluation model (eg, performance evaluation model, dynamic performance evaluation model) Del generating means 90 (e.g., performance evaluation model generation unit 20) and performance evaluation means 91 evaluate the performance of the virtualized system based on the performance evaluation model (e.g., performance evaluation means 50) and a.

  With such a configuration, it is possible to easily estimate the performance of the virtual system.

  Further, the performance evaluation model generation unit 90 calculates a virtual data flow calculation unit that calculates a virtual data flow that is a data flow on the virtual server model 82 from the data flow of the process data flow model 81 based on the virtual realization relationship information 84 ( For example, based on the virtual data flow calculation means 21) and the physical realization relationship information 85, a physical data flow calculation means (for example, physical data) that calculates a physical data flow that is a data flow on the physical server model 83 from the virtual data flow. A flow calculation unit 22), a load calculation unit (for example, a load calculation unit 23) that calculates the load of the object in the physical server model 83 from the physical data flow based on attributes set in the elements in the process data flow model 11. The load calculated by the load calculation means and the physical support Usage rate calculation means (for example, a usage rate calculation means for calculating the usage rate of each element in the physical server model as a performance evaluation model from physical performance specifications of each object in the bus model (for example, throughput in the physical performance DB, CPU benchmark, etc.) Usage rate calculation means 24).

  The system model storage unit 80 is information that describes a plurality of correspondence relationships between each object of the virtual server model 82 and each object of the physical server model 83 in association with a physical system state transition. Dynamic physical realization relationship information (for example, dynamic physical realization relationship information 16) may be stored. The physical data flow calculation unit may calculate the physical data flow for each correspondence according to the physical system state.

  Further, the system model storage unit 80 may store a process data flow model 81 in which attributes (for example, average message size, average CPU time) indicating a physical system load are set in each element of the data flow. . The usage rate calculating means may generate a performance evaluation model from an attribute indicating a physical system load.

  The present invention can be applied to applications such as IT system construction support tools and operation management tools.

10, 10a System model storage means 11 Process data flow model 12 Virtual realization relation information 13 Virtual server model 14 Physical realization relation information 15 Physical server model 16 Dynamic physical realization relation information 20 Performance evaluation model generation means 21 Virtual data flow calculation means 22 Physical data flow calculation means 23 Load calculation means 24 Usage rate calculation means 30 Virtualization performance storage means 40 Physical performance storage means 50 Performance evaluation means 60 Performance evaluation result 70 Dynamic evaluation model generation means

Claims (8)

As an element of a data flow that is a path of data transmitted / received between processes defined on the IT system, an object on the path and a connection relationship between the objects are included, and each element indicates the nature of the data flow. A process data flow model that is a model in which attributes are set, a virtual server model that describes connection relationships between objects in a virtualized system, and a physical server model that describes connection relationships between objects in a physical system , Virtual realization relationship information which is information indicating the correspondence between the elements of the process data flow model and the object of the virtual server model, and information indicating the correspondence between the object of the virtual server model and the object of the physical server model That stores physical realization-related information And Dell storage means,
Based on the rules for calculating the performance of the physical system, the performance of the virtualized system is evaluated from the attributes set in the elements of the process data flow model specified by the virtual realization relation information and the physical realization relation information. A performance evaluation model generating means for generating a performance evaluation model to be
A performance evaluation system comprising: performance evaluation means for performing performance evaluation of a virtualized system based on the performance evaluation model. The performance evaluation model generation means
Virtual data flow calculation means for calculating a virtual data flow that is a data flow on the virtual server model from the data flow of the process data flow model based on the virtual realization relationship information;
Physical data flow calculation means for calculating a physical data flow that is a data flow on a physical server model from the virtual data flow based on physical realization relationship information;
Load calculating means for calculating the load of the object in the physical server model from the physical data flow based on the attribute set in the element in the process data flow model;
A usage rate calculating unit that calculates a usage rate of each element in the physical server model as a performance evaluation model from the load calculated by the load calculating unit and physical performance specifications of each object in the physical server model. Item 1. The performance evaluation system according to Item 1. The system model storage means is a dynamic physical realization relationship, which is information describing a plurality of correspondence relationships between each object of the virtual server model and each object of the physical server model in association with a physical system state transition. Remember information,
The performance evaluation system according to claim 2, wherein the physical data flow calculation unit calculates a physical data flow for each correspondence according to a physical system state. The system model storage means stores a process data flow model in which an attribute indicating a physical system load is set in each element of the data flow,
The performance evaluation system according to claim 2 or 3, wherein the usage rate calculation means generates a performance evaluation model from an attribute indicating a physical system load. The performance evaluation model generation means includes an object on the path and a connection relationship between the objects as elements of a data flow that is a path of data transmitted and received between processes defined on the IT system. Information indicating the correspondence between the element in the process data flow model, which is a model in which an attribute indicating the nature of the data flow is set, and the object in the virtual server model describing the connection relationship between the objects in the virtualized system Identified by certain virtual realization relationship information and physical realization relationship information, which is information indicating the correspondence relationship between the virtual server model object and the physical server model object describing the connection relationship between the objects in the physical system. Of the process data flow model Generate a performance evaluation model that evaluates the performance of the virtualized system based on the rules for calculating the performance of the physical system from the attributes set in the element,
A performance evaluation method, wherein the performance evaluation means performs a performance evaluation of a virtualized system based on the performance evaluation model. When the virtual data flow calculation means generates the performance evaluation model, based on the virtual realization relationship information, calculates a virtual data flow that is a data flow on the virtual server model from the data flow of the process data flow model,
The physical data flow calculation means calculates a physical data flow that is a data flow on the physical server model from the virtual data flow based on the physical realization relationship information,
The load calculation means calculates the load of the object in the physical server model from the physical data flow based on the attribute set in the element in the process data flow model,
The performance according to claim 5 , wherein the usage rate calculating means calculates the usage rate of each element in the physical server model as a performance evaluation model from the calculated load and physical performance specifications of each object in the physical server model. Evaluation method. As an element of a data flow that is a path of data transmitted / received between processes defined on the IT system, an object on the path and a connection relationship between the objects are included, and each element indicates the nature of the data flow. A process data flow model that is a model in which attributes are set, a virtual server model that describes connection relationships between objects in a virtualized system, and a physical server model that describes connection relationships between objects in a physical system , Virtual realization relationship information which is information indicating the correspondence between the elements of the process data flow model and the object of the virtual server model, and information indicating the correspondence between the object of the virtual server model and the object of the physical server model That stores physical realization-related information A performance evaluation program applied to a computer equipped with a Dell storage means,
In the computer,
Based on the rules for calculating the performance of the physical system, the performance of the virtualized system is evaluated from the attributes set in the elements of the process data flow model specified by the virtual realization relation information and the physical realization relation information. A performance evaluation model generation process for generating a performance evaluation model to be performed, and
A performance evaluation program for executing performance evaluation processing for performing performance evaluation of a virtualized system based on the performance evaluation model. On the computer,
In the performance evaluation model generation process,
A virtual data flow calculation process for calculating a virtual data flow that is a data flow on the virtual server model from the data flow of the process data flow model based on the virtual realization relationship information;
A physical data flow calculation process for calculating a physical data flow that is a data flow on a physical server model from the virtual data flow based on physical realization relationship information;
A load calculation process for calculating an object load in the physical server model from the physical data flow based on attributes set in the element in the process data flow model; and
Based on the load calculated in the load calculation process and the physical performance specifications of each object in the physical server model, a usage rate calculation process is performed to calculate the usage rate of each element in the physical server model as a performance evaluation model. The performance evaluation program according to claim 7.

Images