JP4982578B2 - Resource allocation device, resource allocation method, and resource allocation control program - Google Patents

Description

  The present invention relates to a resource allocation device, a resource allocation method, and a resource allocation control program that allocate information processing apparatus resources to a plurality of virtual machines.

  Conventionally, a server virtualization technique that constructs a plurality of virtual machines on a physical server and operates each virtual machine independently has been used in a server system. This server virtualization technology adds the amount of resources required for each virtual machine to secure resources (resource request amount) by adding a buffer with a capacity obtained by multiplying the average resource usage in the virtual machine by a certain amount of deviation. In general, the obtained value is used, and resources are rearranged according to the required resource amount.

  Also known are a computer resource allocation method, a resource management server, and a computer system that periodically manage resource usage and dynamically allocate resources to a plurality of virtual machines (see, for example, Patent Document 1). The computer resource allocation method described in Patent Document 1 collects the resource usage status of a computer, calculates a correlation coefficient for the resource usage of each computer based on the collected data, and collects the collected data or resource. This is a method of calculating the resource allocation value of each computer based on the predicted value of the usage state and the calculated correlation coefficient, and allocating the resource of each computer according to the resource allocation value.

  Further, a method and a system for distributing the load by changing the resource allocation according to the resource consumption level, which is a resource load, are known (see, for example, Patent Document 2). The load distribution method described in Patent Document 2 defines a lower threshold value of resource consumption level for each computer, acquires the resource consumption level of each virtual dedicated server, acquires the resource consumption level of each computer, If each computer's consumption level is compared with its associated lower threshold and a computer with a resource consumption level higher than the lower threshold is identified, then each computer's virtual level corresponding to the current location of the virtual dedicated server is always For a dedicated server, a new arrangement is defined that defines where resources are provided from, and the virtual dedicated server is moved from the current position in the computer to a position corresponding to that computer or another computer according to the arrangement. is there.

Japanese Patent No. 4119239 JP-T-2005-501335

In the general server virtualization technology described above, the resource requirement amount of each virtual machine is calculated based on the average usage amount and its deviation without considering the usage status of the virtual machine. An appropriate resource amount cannot be secured according to the characteristics, and as a result, the aggregation rate of virtual machines cannot be increased.
That is, in order to improve the aggregation rate of virtual machines, it is necessary to calculate the resource request amount according to the usage status of each virtual machine. On the other hand, in the virtual machine, the resource contention rate between the virtual machines increases as the aggregation rate increases, and processing delays occur, which hinders calculation of an appropriate resource request amount. Furthermore, in order to appropriately evaluate the resource requirement amount, it is necessary to acquire corresponding measurement data, but the reliability of the measurement data and the load followability are in a trade-off relationship.

The inventions disclosed in Patent Document 1 and Patent Document 2 described above are not intended to improve the aggregation rate of virtual machines. In addition, these documents describe that the resource contention rate increases as the virtual machine aggregation rate increases, resulting in problems in calculating appropriate resource requirements, and measurement data to appropriately evaluate resource requirements. There is no disclosure of a problem about the point that the load followability deteriorates if an attempt is made to increase the reliability and a configuration for solving these problems.
Therefore, the present invention has been made to solve the above-described problem, and the resource requirement amount required for each of a plurality of virtual machines constructed on one or a plurality of information processing apparatuses is determined according to load followability and resource. To provide a resource allocation device, a resource allocation method, and a resource allocation control program that can be obtained in consideration of contention, and can increase the aggregation rate of virtual machines by securing appropriate resource capacity and allocating to virtual machines. With the goal.

[1] One aspect of the present invention for solving the above-described problem is that a measurement processing unit that measures a resource usage amount for a predetermined period for each of a plurality of virtual machines to which resources of an information processing device are allocated; Based on the resource usage measured by the measurement processing unit, the predicted value of the resource usage and the degree of change in the predicted load are calculated for each of the plurality of virtual machines, and the degree of change in the predicted load is equal to or greater than a predetermined threshold. If there is, the resource request amount is calculated based on the average resource usage amount of the virtual machine calculated based on the resource usage amount for the predetermined period and the predicted value of the resource usage amount, while the predicted load When the degree of change is less than the predetermined threshold, the average resource of the virtual machine calculated based on the resource usage for the predetermined period, respectively. The resource request amount is calculated based on a deviation of the usage amount and the resource usage amount, and a resource contention value indicating a resource contention value calculated based on an average resource usage rate and the resource usage amount deviation in the information processing apparatus. A distribution calculation unit for calculating, and a resource allocation unit for rearranging and allocating resources of the information processing apparatus to each virtual machine based on a resource request amount for each of the plurality of virtual machines calculated by the distribution calculation unit; A resource allocating device comprising:
Here, when the degree of change in the predicted load is a value equal to or greater than a predetermined threshold (for example, a switching parameter), the distribution calculation unit weights the average resource usage and the predicted value of the resource usage by the prediction evaluation parameter. The resource requirement amount Rreq _j may be calculated by adding the values. Further, when the degree of change in the predicted load is a value less than a predetermined threshold, the allocation calculation unit sets the resource contention value with the resource contention evaluation parameter so as to secure more buffers as resource contention easily occurs. You may make it weight.

  [2] The resource allocation device according to [1], further comprising: a classification processing unit that classifies each of the plurality of virtual machines into a predetermined category based on the average resource usage, For each of the plurality of virtual machines, the resource request amount is calculated in association with the classification classified by the classification processing unit.

  [3] One aspect of the present invention for solving the above-described problem is that a measurement step of measuring a resource usage amount for a predetermined period for each of a plurality of virtual machines to which resources of an information processing device are allocated, and the measurement When the predicted value of the resource usage and the degree of change in the predicted load are calculated for each of the plurality of virtual machines based on the resource usage measured in the step, and the degree of change in the predicted load is greater than or equal to a predetermined threshold Includes calculating the resource request amount based on the average resource usage amount of the virtual machine calculated based on the resource usage amount for the predetermined period and the predicted value of the resource usage amount, while changing the predicted load Is less than the predetermined threshold value, the virtual machine is calculated based on the resource usage for the predetermined period. The resource request based on an average resource usage and a deviation of the resource usage, and a resource contention value indicating a resource contention level calculated based on an average resource usage rate and the resource usage deviation in the information processing apparatus An allocation calculation step for calculating an amount, and a resource allocation step for rearranging and allocating resources of the information processing apparatus to each virtual machine based on a resource request amount for each of the plurality of virtual machines calculated in the allocation calculation step A resource allocation method characterized by comprising:

  [4] The resource allocation method according to [3], further including a classification step of classifying each of the plurality of virtual machines into a predetermined category based on the average resource usage, wherein the allocation calculation step includes: For each of a plurality of virtual machines, the resource requirement amount is calculated in correspondence with the classification classified in the classification step.

  [5] One aspect of the present invention for solving the above problem is a measurement step of measuring a resource usage amount for a predetermined period for each of a plurality of virtual machines to which information processing apparatus resources are assigned to a computer. Based on the resource usage measured in the measurement step, the predicted value of the resource usage and the degree of change in the predicted load are calculated for each of the plurality of virtual machines, and the degree of change in the predicted load is equal to or greater than a predetermined threshold. The resource request amount is calculated based on the average resource usage amount of the virtual machine calculated based on the resource usage amount for the predetermined period and the predicted value of the resource usage amount, while the predicted load When the degree of change is less than the predetermined threshold, it is calculated based on the resource usage for the predetermined period, respectively. Based on the average resource usage of the virtual machine and the deviation of the resource usage, and the resource contention value indicating the degree of resource contention calculated based on the average resource usage rate and the resource usage deviation in the information processing apparatus A resource calculating amount for calculating the resource request amount, and reallocating resources of the information processing apparatus to each virtual machine based on the resource request amount for each of the plurality of virtual machines calculated in the distribution calculating step. The resource allocation control program for performing the process of the resource allocation step to allocate.

  [6] The resource allocation control program according to [5] further includes a classification step of classifying each of the plurality of virtual machines into a predetermined category based on the average resource usage, and the allocation calculation step includes: For each of the plurality of virtual machines, the computer is caused to execute a process of calculating the resource request amount corresponding to the classification classified in the classification step.

  According to the present invention, a resource requirement amount required for each of a plurality of virtual machines constructed on one or a plurality of information processing apparatuses is obtained in consideration of load followability and resource competition, and an appropriate resource capacity is ensured. By assigning them to virtual machines, the aggregation rate of virtual machines can be increased.

1 is a schematic configuration diagram of a server system to which a resource allocation device according to an embodiment of the present invention is applied. It is a block diagram which shows the function structure of the resource control server and center server in this embodiment. In this embodiment, it is a figure which shows the classification | category of the management information which an allocation management part manages. It is a figure which shows the data structure of the resource information in this embodiment. It is a figure which shows the data structure of the physical server resource information in this embodiment. It is a figure which shows the data structure of the virtual server basic information in this embodiment. It is a figure which shows the data structure of virtual server resource information in this embodiment. It is a figure which shows the data structure of the basic information in this embodiment. It is a figure which shows the data structure of the classification definition information in this embodiment. It is a figure which shows the data structure of classification | category evaluation information in this embodiment. It is a figure which shows the data structure of measurement information in this embodiment. It is a figure which shows the data structure of arrangement | positioning information in this embodiment. It is a sequence diagram which shows the basic operation | movement procedure of a server system in this embodiment. It is a sequence diagram which shows the basic operation | movement procedure of a server system in this embodiment. In this embodiment, it is a flowchart which shows the procedure of the classification calculation process which a classification calculation part performs. In this embodiment, it is a flowchart which shows the procedure of the virtual server arrangement | positioning calculation process which an arrangement | positioning calculation part performs. In this embodiment, it is a flowchart which shows the procedure of the evaluation process of the evaluation arrangement | positioning data which an arrangement | positioning calculation part performs. It is a graph which shows the resource requirement amount with respect to the number of steps in the server system by this embodiment and a modification, and the server system by a prior art. It is a graph which shows the type of the resource requirement amount with respect to the number of steps in the server system according to the present embodiment and the modified example and the server system according to the conventional technology.

  Hereinafter, embodiments for carrying out the present invention will be described in detail with reference to the drawings. FIG. 1 is a schematic configuration diagram of a server system to which a resource allocation device according to an embodiment of the present invention is applied. As shown in the figure, the server system 100 includes six virtual servers in four physical servers 1 (in the figure, the four physical servers 1 are shown as physical servers 1-1 to 1-4). 2 (in the figure, six virtual servers 2 are shown as virtual servers 2-1 to 2-6), and these physical servers 1-1 to 1-4 and virtual servers 2-1 to 2- 6 is controlled by the center server 3 and the resource control server 4.

  In the server system 100, the number of physical servers need not be limited to four, and may be one or more. Also, the number of virtual servers is not limited to six, but can be a plurality of numbers.

The physical server 1 is a server device that is an information processing device, and includes resources such as a CPU (Central Processing Unit), a memory, a graphics interface, a network interface, a storage, and an I / O interface.
The virtual server 2 is a virtual machine built on the physical server 1 and uses the resources of the physical server 1 independently of the other virtual servers 2. That is, the server system 100 shown in FIG. 1 is virtualized as if there were six independent servers.

  The resource control server 4 manages the allocation of resources to the virtual server 2. In managing resource allocation, the resource control server 4 instructs to measure the resource usage of the virtual server 2, classify calculation instructions for classifying the virtual server 2 according to the load, and the physical server 1 of the virtual server 2. A distribution calculation instruction for determining allocation to the center server 3 is supplied to the center server 3. Further, the resource control server 4 supplies an initial placement instruction for the virtual server 2 to the center server 3 or sends a resource allocation instruction for the virtual server 2 based on the resource request amount of the virtual server 2 acquired from the center server 3 to the center. Or supply to the server 3.

  The center server 3 dynamically classifies the virtual server 2 constructed in the physical server 1 according to the load, based on the measurement instruction, the classification calculation instruction, and the distribution calculation instruction supplied from the resource control server 4. The resource request amount of the server 2 is calculated and supplied to the resource control server 4. Further, the center server 3 allocates and allocates the virtual server 2 to the physical server 1 based on the initial allocation instruction or the resource allocation instruction supplied from the resource control server 4.

  Although not shown in FIG. 1, the physical server 1, the center server 3, and the resource control server 4 are connected through an electric communication line such as a network.

Next, functional configurations of the resource control server 4 and the center server 3 will be described. FIG. 2 is a block diagram illustrating functional configurations of the resource control server 4 and the center server 3. As shown in the figure, the resource control server 4 and the center server 3 include an allocation management unit 10, a measurement processing unit 20, a classification calculation unit (classification processing unit) 30, a distribution calculation unit 40, and a resource allocation unit 50. With.
That is, the resource allocation device in the present embodiment is a device that includes the allocation management unit 10, the measurement processing unit 20, the classification calculation unit 30, the distribution calculation unit 40, and the resource allocation unit 50.

  The allocation management unit 10 has a control function related to resource allocation and a storage function for storing various information. That is, the allocation management unit 10 controls the measurement processing unit 20, the classification calculation unit 30, the distribution calculation unit 40, and the resource allocation unit 50 as control functions. The allocation management unit 10 includes a system information storage unit 11, a control information storage unit 12, a measurement information storage unit 13, and an arrangement information storage unit 14 as storage functions, and stores and manages management information. The system information storage unit 11, the control information storage unit 12, the measurement information storage unit 13, and the arrangement information storage unit 14 are configured by, for example, a hard disk device or a semiconductor storage device, or a combination thereof.

The system information storage unit 11 stores system information. The control information storage unit 12 stores control information. The measurement information storage unit 13 stores measurement information. The arrangement information storage unit 14 stores arrangement information. Details of the system information, control information, measurement information, and arrangement information included in the management information will be described later.
The allocation management unit 10 stores, for example, system information and control information input from the outside by an operation of an operation unit (not shown) by a system administrator of the server system 100 in the system information storage unit 11 and the control information storage unit 12, respectively. .

When the measurement processing unit 20 takes in the measurement instruction supplied from the allocation management unit 10, the measurement processing unit 20 acquires the resource usage of each resource for each virtual server 2 from the virtual server 2 and the physical server 1 which are execution environments. Then, the measurement processing unit 20 supplies measurement information including data indicating the acquired resource usage to the allocation management unit 10.
When the classification calculation unit 30 receives the classification calculation instruction supplied from the allocation management unit 10, the classification calculation unit 30 executes a classification calculation process for classifying the virtual server 2 based on the resource usage.
When the distribution calculation unit 40 receives the distribution calculation instruction supplied from the allocation management unit 10, the distribution calculation unit 40 executes a virtual server arrangement calculation process for determining the physical server 1 to which the virtual server 2 is allocated.
When the resource allocation unit 50 receives the initial allocation instruction or the resource allocation instruction supplied from the allocation management unit 10, the resource allocation unit 50 executes a process of allocating and allocating the virtual server 2 to the physical server 1.

  Next, various information stored in the system information storage unit 11, the control information storage unit 12, the measurement information storage unit 13, and the arrangement information storage unit 14 will be described. First, FIG. 3 is a diagram showing a classification of management information managed by the allocation management unit 10. As shown in the figure, the management information managed by the assignment management unit 10 includes system information, control information, measurement information, and arrangement information. The system information is classified into resource information, physical server resource information, virtual server basic information, and virtual server resource information. The control information is classified into basic information, classification definition information, and classification evaluation information.

  The resource information includes information that identifies a resource included in the physical server 1. FIG. 4 is a diagram illustrating a data configuration of resource information. As shown in the figure, the resource information includes resource identification information and a resource name in association with each other. The resource identification information is information for identifying a resource. The resource name is the name of the resource indicated by the resource identification information. The primary key in the resource information is resource identification information. The resource information is stored in advance in the system information storage unit 11 by external supply or the like.

  The physical server resource information includes information regarding resources provided in the physical server 1. FIG. 5 is a diagram illustrating a data configuration of physical server resource information. As shown in the figure, the physical server resource information includes physical server identification information, resource identification information, a resource upper limit value, and an average usage rate in association with each other. The physical server identification information is information for identifying the physical server 1. The resource identification information is identification information corresponding to a resource included in the physical server 1 indicated by the physical server identification information. The resource upper limit value is an upper limit value (maximum value) of the resource indicated by the resource identification information, and the average usage rate is an average value of the usage rate with respect to the resource upper limit value in a certain sampling period of the resource. The primary key in the physical server resource information is physical server identification information. The physical server resource information is stored in advance in the system information storage unit 11 by external supply or the like, and the average usage rate is updated by the allocation management unit 10 based on the subsequent measurement processing by the measurement processing unit 20.

  The virtual server basic information includes information on the initial placement destination and classification of the virtual server 2. FIG. 6 is a diagram illustrating a data configuration of the virtual server basic information. As shown in the figure, the virtual server basic information includes virtual server identification information, initial placement destination identification information, and classification identification information in association with each other. The virtual server identification information is information for identifying the virtual server 2. The initial placement destination identification information is identification information indicating the physical server 1 in which the virtual server 2 indicated by the virtual server identification information is first placed based on the initial placement instruction. The classification identification information is information for identifying the classification of the virtual server 2 indicated by the virtual server identification information. The primary key in the virtual server basic information is virtual server identification information. The virtual server basic information is stored in advance in the system information storage unit 11 by external supply or the like, and the classification identification information is updated by the allocation management unit 10 based on the classification calculation processing by the classification calculation unit 30 thereafter.

  The virtual server resource information includes information regarding resources reserved by the virtual server 2. FIG. 7 is a diagram illustrating a data configuration of virtual server resource information. As shown in the figure, the virtual server resource information includes virtual server identification information, resource identification information, average resource usage, and resource usage deviation in association with each other. The virtual server identification information is information for identifying the virtual server 2. The resource identification information is identification information corresponding to a resource secured by the virtual server 2 indicated by the virtual server identification information. The average resource usage is an average value of the resource usage of the resource indicated by the resource identification information in a certain sampling period, and the resource usage deviation is a deviation of the average resource usage of the resource. The primary key in the virtual server resource information is virtual server identification information. The physical server resource information is stored in advance in the system information storage unit 11 by external supply or the like, and is updated by the allocation management unit 10 based on the subsequent measurement processing by the measurement processing unit 20.

  The basic information includes control parameters in the resource allocation device. FIG. 8 is a diagram illustrating a data configuration of basic information. As shown in the figure, the basic information includes a virtual server movement upper limit parameter and a control interval parameter. The virtual server movement upper limit parameter is an allowable upper limit value of the number of times the virtual server 2 moves between physical servers when the virtual server 2 is rearranged. The control interval parameter is a time indicating a cycle in which the allocation management unit 10 controls resource allocation. The basic information is stored in advance in the control information storage unit 12 by external supply or the like.

  The classification definition information includes definition information for classifying the virtual server 2. FIG. 9 is a diagram illustrating a data configuration of the classification definition information. As shown in the figure, the classification definition information includes classification identification information, resource identification information, and a resource centroid value. The classification identification information is identification information indicating the classification of the virtual server 2. For example, the classification (heavy user) of the virtual server 2 that performs processing with a relatively large load is set as classification A, the classification (light user) of the virtual server 2 that performs processing with a relatively small load is set as classification C, and classification A and classification A classification (middle user) of the virtual server 2 that performs processing of a load that is in the middle of C is a classification C. The resource identification information is information corresponding to the resource identification information set in the resource information, and the resource centroid value is the centroid of the load when the classification load indicated by the classification identification information is applied to the resource indicated by the resource identification information. Value. The primary key in the classification definition information is classification identification information. The classification definition information is stored in advance in the control information storage unit 12 by external supply or the like.

  The classification evaluation information includes an evaluation parameter used when evaluating the resource requirement for each classification. FIG. 10 is a diagram illustrating a data configuration of the classification evaluation information. As shown in the figure, the classification evaluation information includes classification identification information, a deviation evaluation parameter, a prediction evaluation parameter, a switching parameter, and a resource competition evaluation parameter. The classification identification information is information corresponding to the classification identification information set in the classification definition information. The deviation evaluation parameter is a parameter for weighting the resource usage deviation included in the virtual server resource information, and determines how much the allocation calculation unit 40 secures the buffer. The prediction evaluation parameter is a weighting parameter for adjusting the balance between the predicted value of the resource usage calculated by the distribution calculation unit 40 and the average resource usage included in the virtual server resource information. The switching parameter is a threshold value for deciding whether or not to consider the predicted value of the resource usage when the distribution calculation unit 40 calculates the resource requirement. The resource competition evaluation parameter is a parameter for weighting the resource competition value calculated by the distribution calculation unit 40, and is a parameter for determining how much a buffer is to be secured when resource competition occurs. This resource contention value is a numerical value of the degree of resource contention.

  The measurement information is information including the resource usage acquired by the measurement processing unit 20 from the virtual server 2 and the physical server 1, which are execution environments, and related data. FIG. 11 is a diagram illustrating a data configuration of measurement information. As shown in the figure, the measurement information includes data acquisition time, virtual server identification information, resource identification information, and resource usage in association with each other. The data acquisition time is time information indicating the date and hour / minute / second when the measurement processing unit 20 acquires measurement data from the execution environment. The virtual server identification information is identification information for identifying the virtual server 2 that is the measurement target. Resource identification information is identification information of a resource to be measured. The resource usage is measurement data indicating the measured resource usage.

  The placement information is information indicating the placement relationship between the physical server 1 and the virtual server 2 that indicates the physical server 1 on which the virtual server 2 is placed. FIG. 12 is a diagram illustrating a data configuration of arrangement information. As shown in the figure, the arrangement information includes physical server identification information and virtual server identification information in association with each other. The physical server identification information is information for identifying the physical server 1. The virtual server identification information is identification information of the virtual server 2 arranged in the physical server 1 indicated by the physical server identification information.

  Next, the operation of the server system 100 in this embodiment will be described. 13 and 14 are sequence diagrams showing a basic operation procedure of the server system 100. FIG. Resource information, physical server resource information, virtual server basic information, and virtual server resource information (hereinafter referred to as system information) are stored in the system information storage unit 11 by the operation of the operation unit by the system administrator. After the definition information and the classification evaluation information (hereinafter referred to as control information) are stored in the control information storage unit 12, the process according to the sequence diagram shown in FIG. 13 is executed. After the process according to the sequence diagram of FIG. 13 is completed, the process according to the sequence diagram of FIG. 14 is executed.

  First, in step S <b> 1, the allocation management unit 10 supplies an initial allocation instruction command to the resource allocation unit 50. Then, the allocation management unit 10 updates the arrangement information stored in the arrangement information storage unit 14 based on the virtual server basic information read from the system information storage unit 11. Next, in step S <b> 2, when the resource allocation unit 50 takes in the initial arrangement instruction command supplied from the allocation management unit 10, the resource information, physical server resource information, virtual server basic information stored in the system information storage unit 11. Based on the virtual server resource information and the placement information stored in the placement information storage unit 14, the initial placement processing of the virtual server 2 is executed. In this initial arrangement process, the physical server 1, the virtual server 2 constructed in the physical server 1, and the resources secured by the virtual server 2 are associated, and virtualization is performed by a known server virtualization technique. .

  Next, in step S3, the allocation management unit 10 supplies a measurement instruction command to the measurement processing unit 20 and starts measuring time. Next, in step S4, when the measurement processing unit 20 fetches the measurement instruction command supplied from the allocation management unit 10, the measurement processing unit 20 reads the placement information stored in the placement information storage unit 14, and the virtual server 2 that is the execution environment and A resource usage measurement instruction is supplied to the physical server 1. Next, in step S5, when the virtual server 2 and the physical server 1 which are execution environments receive the measurement instruction supplied from the measurement processing unit 20, the virtual server 2 and the physical server 1 acquire the resource usage for each resource, and acquire the acquired resource usage. Measurement data including the value of is supplied to the measurement processing unit 20. Next, in step S6, the measurement processing unit 20 takes in the measurement data supplied from the virtual server 2 together with the current date and time information, generates measurement information including the taken-in data, and generates the allocation management unit 10. To supply.

  Next, in step S <b> 7, the assignment management unit 10 updates the measurement information stored in the measurement information storage unit 13 when the measurement information supplied from the measurement processing unit 20 is captured. Next, in step S8, the allocation management unit 10 determines whether or not the time indicated by the control interval parameter included in the basic information stored in the control information storage unit 12 has elapsed since the start of timing in the process of step S3. Determine. If the allocation management unit 10 determines that the elapsed time at the current time has not passed the time indicated by the control interval parameter (S8: YES), the allocation management unit 10 returns to the process of step S3, and the elapsed time at the current time. When it is determined that the time indicated by the control interval parameter has elapsed (S8: NO), the process proceeds to step S9.

  In step S <b> 9, the allocation management unit 10 calculates the average usage rate for each resource of the physical server 1 based on the measurement information stored in the measurement information storage unit 13, and stores the physical storage stored in the system information storage unit 11. Update server resource information. Next, in step S10, the allocation management unit 10 calculates an average resource usage and a resource usage deviation for each resource of the virtual server 2 based on the measurement information stored in the measurement information storage unit 13, The virtual server resource information stored in the system information storage unit 11 is updated.

  Next, in step S <b> 11 of the sequence diagram shown in FIG. 14, the allocation management unit 10 supplies a classification calculation instruction command to the classification calculation unit 30. Next, in step S12, when the classification calculation unit 30 takes in the classification calculation instruction command supplied from the allocation management unit 10, the classification calculation unit 30 executes a classification calculation process for classifying the virtual server 2 according to the load. Details of this classification calculation processing will be described later.

  Next, in step S <b> 13, the allocation management unit 10 supplies a distribution calculation instruction command to the distribution calculation unit 40. Next, in step S <b> 14, when the distribution calculation unit 40 fetches the distribution calculation instruction command supplied from the allocation management unit 10, the distribution calculation unit 40 executes a virtual server arrangement calculation process for relocating the virtual server 2. Details of the virtual server arrangement calculation processing will be described later. Next, in step S <b> 15, the distribution calculation unit 40 supplies the allocation information that is the return value of the virtual server allocation calculation process to the allocation management unit 10. Next, in step S <b> 16, the allocation management unit 10 updates the arrangement information stored in the arrangement information storage unit 14 when the arrangement information supplied from the distribution calculation unit 40 is fetched.

  Next, in step S <b> 17, the allocation management unit 10 supplies a resource allocation instruction command to the resource allocation unit 50. Next, in step S18, when the resource allocation unit 50 takes in the resource allocation instruction command supplied from the allocation management unit 10, the resource information, physical server resource information, virtual server basic information stored in the system information storage unit 11 Based on the virtual server resource information and the placement information stored in the placement information storage unit 14, the rearrangement process of the virtual server 2 is executed. Similar to the processing in step S2, this relocation processing is performed by associating the physical server 1, the virtual server 2 constructed in the physical server 1, and the resources secured by the virtual server 2 with known server virtualization. Resources are rearranged by technology. Next, in step S <b> 19, when the resource rearrangement process ends, the resource allocation unit 50 supplies an allocation response indicating the end of the process to the allocation management unit 10.

Next, the classification calculation process executed by the classification calculation unit 30 will be described. FIG. 15 is a flowchart showing the procedure of the classification calculation process executed by the classification calculation unit 30. When the classification calculation unit 30 fetches the classification calculation instruction command supplied from the allocation management unit 10, the classification calculation unit 30 starts the processing of this flowchart.
First, in step S101, the classification calculation unit 30 reads the average resource usage of each resource for each virtual server 2 from the virtual server resource information stored in the system information storage unit 11, and creates an average resource usage vector. .
Specifically, the classification calculation unit 30 uses, for example, the average resource usage for the record whose virtual server identification information is “1” in the virtual server resource information (see FIG. 7) stored in the system information storage unit 11. A quantity vector X (bold body) ₁ = (1 GHz, 100 MB,...) Is created. Here, the description “(bold)” indicates that the immediately preceding character or character string is represented in bold, and that the character or character string represents a vector.

Next, in step S102, the classification calculation unit 30 reads out the resource barycentric value for each classification identification information from the classification definition information stored in the control information storage unit 12, and creates a resource barycentric vector.
Specifically, the classification calculation unit 30 uses, for example, the resource centroid vector Y () for the record whose classification identification information is “A” in the classification definition information (see FIG. 9) stored in the control information storage unit 12. Bold body) _{C = A} = (aaa1, bbb1,...) Is created.

Next, in step S103, the classification calculation unit 30 selects one virtual server 2, and the distance between the average resource usage vector X (bold body) and the resource centroid vector Y (bold body) _{C = n} is The smallest resource centroid vector Y (bold body) _{C = k} is selected. This distance calculation is, for example, the square of the difference value between the average resource usage vector X (bold body) and the resource centroid vector Y (bold body) _{C = n} . Next, in step S104, the classification calculation unit 30 stores the virtual server basic stored in the system information storage unit 11 with the classification identification information k in the resource centroid vector Y (bold body) _{C = k} selected in the process of step S103. The classification identification information corresponding to the currently selected virtual server 2 is updated.

  Next, in step S105, the classification calculation unit 30 determines whether or not calculation has been performed for all virtual servers 2 based on the virtual server identification information included in the virtual server resource information stored in the system information storage unit 11. If it is determined that there are still uncalculated virtual servers 2 (S105: NO), the process returns to step S103, and if it is determined that the calculation for all the virtual servers 2 is completed (S105: YES). Then, the process of this flowchart is terminated.

Next, the virtual server arrangement calculation process executed by the distribution calculation unit 40 will be described. FIG. 16 is a flowchart showing the procedure of the virtual server arrangement calculation process executed by the arrangement calculation unit 40. When the distribution calculation unit 40 takes in the distribution calculation instruction command supplied from the allocation management unit 10, the distribution calculation unit 40 starts the processing of this flowchart.
First, in step S201, the distribution calculation unit 40 calculates the arrangement combination of the virtual server 2 and the physical server 1 based on the arrangement information stored in the arrangement information storage unit 14 and creates evaluation arrangement information. .
Specifically, the distribution calculation unit 40 applies the all search method algorithm which is a known search algorithm to obtain the arrangement combination of the virtual server 2 and the physical server 1 by calculation and creates evaluation arrangement information. For example, in the case of an arrangement combination of two physical servers 1-1 and 1-2 and three virtual servers 2-1 to 2-3, the distribution calculation unit 40 evaluates including eight evaluation arrangement data. Create placement information.

  In addition to the full search algorithm, the allocation calculator 40 can apply, for example, a dynamic programming (DP) method, a genetic algorithm (GA) GA method, and the like as a search algorithm. is there.

  Next, in step S202, the distribution calculation unit 40 determines whether or not evaluation of all evaluation arrangement data included in the evaluation arrangement information created in the process of step S201 has been completed. When it is determined that the evaluation of all the evaluation arrangement data has been completed (S202: YES), the distribution calculation unit 40 sets the current arrangement information as a return value and ends the processing of this flowchart. On the other hand, when the distribution calculation unit 40 determines that unevaluated evaluation arrangement data remains (S202: NO), the process proceeds to step S203.

  In step S203, the distribution calculation unit 40 selects one evaluation arrangement data from the evaluation arrangement information. Next, in step S204, the distribution calculation unit 40 executes an evaluation process on the selected evaluation arrangement data, and acquires an evaluation value and the number of movements. Details of the evaluation arrangement data evaluation process will be described later.

Next, in step S205, the distribution calculation unit 40 compares the evaluation value that is the return value of the evaluation arrangement data evaluation process in step S204 with the current minimum evaluation value that is a variable in the virtual server arrangement calculation process. If it is determined that the evaluation value is equal to or less than the minimum evaluation value (S205: YES), the process proceeds to step S206. If it is determined that the evaluation value is greater than the minimum evaluation value (S205: NO), the process proceeds to step S202. Return to processing.
In step S206, the distribution calculation unit 40 compares the movement number that is the return value of the evaluation processing of the evaluation arrangement data in step S204 with the current minimum movement number that is a variable in the virtual server arrangement calculation processing, and moves If it is determined that the number is smaller than the minimum movement number (S206: YES), the process proceeds to step S207. If it is determined that the movement number is equal to or greater than the minimum movement number (S206: NO), the process returns to step S202. .

  In step S207, the distribution calculation unit 40 substitutes the evaluation value for the minimum evaluation value, substitutes the number of movements for the minimum movement number, and substitutes the currently selected evaluation arrangement data for the arrangement information. Move.

Next, the evaluation processing of evaluation arrangement data executed by the distribution calculation unit 40 will be described. FIG. 17 is a flowchart illustrating a procedure of evaluation processing of evaluation placement data executed by the placement calculation unit 40.
First, in step S <b> 301, the distribution calculation unit 40 performs physical server resource information, virtual server basic information, and virtual server resource information stored in the system information storage unit 11, and classification evaluation information stored in the control information storage unit 12. The basic information and the arrangement information stored in the arrangement information storage unit 14 are read.

  Next, in step S <b> 302, the distribution calculation unit 40 calculates a predicted value of the resource usage for each resource for each virtual server 2. The calculation of the predicted value of the resource usage is calculated by, for example, an exponential smoothing method known as an analysis method for predicting a future value based on time series data of the resource usage. The exponential smoothing method is a method of calculating and smoothing a weighted average with respect to time-series data, and performing exponential weighting that decreases as it goes back in the past. In the present embodiment, the allocation calculation unit 40 calculates a predicted value of resource usage by applying a secondary exponential smoothing method in the exponential smoothing method. Next, the distribution calculation unit 40 calculates the degree of change in the predicted load, which is the magnitude of the gradient of the predicted value of the resource usage, in association with each predicted value of the resource usage. The degree of change in the predicted load is, for example, an absolute value of a value representing the slope of the predicted value of resource usage.

Next, in step S <b> 303, the distribution calculation unit 40 calculates a resource request amount for each virtual server 2 according to the classification.
Specifically, the distribution calculation unit 40 calculates the resource request amount according to the degree of change in the predicted load calculated in the process of step S302. That is, the distribution calculation unit 40 first extracts the classification identification information corresponding to the virtual server identification information of the virtual server 2 that is the calculation target from the virtual server basic information, and corresponds to the extracted classification identification information from the classification evaluation information. A deviation evaluation parameter, a prediction evaluation parameter, a switching parameter, and a resource competition evaluation parameter are extracted. Next, when the degree of change in the predicted load is a value greater than or equal to the switching parameter, the distribution calculation unit 40 calculates the resource request amount Rreq _j using the following equation (1). In equation (1), j is virtual server identification information, β _j is a prediction evaluation parameter (0 <β _j <1), m is an average resource usage, pv is a predicted value of resource usage, and a is a prediction load. , Γ _j is a switching parameter.

That is, when the absolute value | a | (the degree of change in the predicted load) of the predicted load is a value greater than or equal to the switching parameter γ _j , the predicted load fluctuates relatively greatly, and thus the deviation from the average resource usage amount. Has a large percentage of errors. Therefore, when the absolute value | a | of the gradient of the predicted load is a value greater than or equal to the switching parameter γ _j , the distribution calculation unit 40 uses the average resource usage m and the predicted value pv of the resource usage as the prediction evaluation parameter. The resource requirement amount Rreq _j is calculated by weighting and adding by β _j .
Thus, by calculating the resource request amount using the average resource usage amount m and the predicted value pv of the resource usage amount, the load followability can be improved. For example, if the predicted value pv of the resource usage is interpolated by changing the weighting ratio of the prediction evaluation parameter β _j according to the length of the measurement period, the load followability can be finely adjusted. The load can be followed with high accuracy.

  In equation (1), the weighting of the average resource usage m and the predicted value pv of the resource usage may be set independently.

On the other hand, when the absolute value | a | of the predicted load gradient is less than the switching parameter γ _j , the distribution calculation unit 40 calculates the resource request amount Rreq _j using the following equation (2). In Expression (2), α _j is a deviation evaluation parameter, d is a resource usage deviation, ζ _j is a resource competition evaluation parameter, rc is a resource competition value, and mr is an average usage rate. Sig () is a sigmoid function.

That is, when the absolute value | a | of the predicted load slope is less than the switching parameter γ _j , the load fluctuation is in a steady state or a state close thereto, and the resource request amount is evaluated in consideration of the resource usage deviation d. It can be an expression. However, as the average resource utilization rate mr of the physical server 1 is higher and the resource usage deviation d is lower, resource competition is more likely to occur. Therefore, the distribution calculation unit 40 increases the buffer as the resource competition is more likely to occur. the to ensure lot, giving weight resource contention value rc at the resource contention evaluation parameter zeta _j.

  Next, in step S <b> 304, the distribution calculation unit 40 calculates a total resource request amount that is the total of the resource request amounts for each physical server 1. Next, in step S305, the distribution calculation unit 40 calculates the number of movements of the virtual server 2 based on the evaluation placement data and placement information that are arguments. That is, the evaluation arrangement data is compared with the arrangement information, and the number of movements of the virtual server 2 necessary for changing from the arrangement state of the virtual server 2 in the arrangement information to the arrangement state of the virtual server 2 in the evaluation arrangement data is calculated. .

Next, in step S306, the distribution calculation unit 40 determines, for each physical server 1, whether the total resource request amount is equal to or less than the resource upper limit and the number of movements of the virtual server 2 is equal to or less than the movement upper limit.
Specifically, for each physical server 1, the distribution calculation unit 40 determines that the total resource request amount calculated in the process of step S304 is equal to or less than the resource upper limit value included in the physical server resource information, and in the process of step S305. It is determined whether or not the calculated movement number of the virtual server 2 is equal to or less than the virtual server movement upper limit parameter included in the basic information.
If the result of the determination is a positive result (step S306: YES), the process proceeds to step S307. If the result is negative (step S306: NO), the process proceeds to step S308. .

  In step S307, the distribution calculation unit 40 sets the value of the number of physical servers 1 for which the placement of the virtual server 2 is requested as an evaluation value as a return value. On the other hand, in step S308, the distribution calculation unit 40 sets, for example, an abnormal value such as a value representing infinity or a negative value as an evaluation value that is a return value.

  As described above, according to the embodiment of the present invention, the resource request amount required for each of the virtual servers 2-1 to 2-6 constructed in the physical servers 1-1 to 1-4 is determined according to the load followability. In addition, the aggregation rate of the virtual servers 2-1 to 2-6 can be increased by obtaining in consideration of resource competition and securing an appropriate resource capacity and allocating them to the virtual servers 2-1 to 2-6.

  In the above-described embodiment, an example is shown in which the resource allocation device calculates and evaluates the resource request amount after dynamically classifying the virtual server 2 according to the load. As a modified example of this, the resource allocation device may calculate and evaluate the resource request amount by omitting the dynamic classification process according to the load.

Here, a result of an operation experiment between the server system according to the above-described embodiment and the above-described modification and the server system according to the related art that does not consider resource usage prediction and resource competition will be described. FIG. 18 is a graph showing the resource request amount with respect to the number of steps in the server system according to the above-described embodiment and the server system according to the related art. As is apparent from the figure, the embodiment and the modification of the present invention are much more excellent in fitting to the load than the prior art. In particular, the followability at a location where the load changes from a large load to a small load is extremely superior to that of the prior art.
FIG. 19 is a graph in which the experiment shown in FIG. As is apparent from FIG. 19, according to the embodiment of the present invention, it can be seen that the distance of the resource request amount with respect to the load is smaller than that of the prior art, and an appropriate resource can be requested according to the load.

  That is, according to the server system that performs both load prediction and dynamic classification in consideration of resource competition according to the present embodiment, the aggregation rate of the virtual server 2 can be improved by about 25% with respect to the prior art. . Moreover, according to the modification of this embodiment, according to only the load prediction in consideration of resource competition, the aggregation rate of the virtual server 2 can be improved by about 22% with respect to the prior art. Thereby, according to the server system by this embodiment and the modification of this embodiment, the number of physical servers 1 can be reduced compared with the server system by a prior art.

  In the server system 100 according to the above-described embodiment, a server in which the function of the resource control server 4 and the function of the center server 3 are integrated may be used. In this case, the integrated server resource allocation control function may be realized by a computer. In this case, a resource allocation control program for realizing the resource allocation control function is recorded on a computer-readable recording medium, and the resource allocation control program recorded on the recording medium is read by the computer system and executed. It may be realized by. Here, the “computer system” includes an OS (Operating System) and hardware of peripheral devices. The “computer-readable recording medium” refers to a portable recording medium such as a flexible disk, a magneto-optical disk, an optical disk, and a memory card, and a storage device such as a hard disk built in the computer system. Furthermore, the “computer-readable recording medium” dynamically holds a program for a short time like a communication line when transmitting a program via a network such as the Internet or a communication line such as a telephone line. In this case, a volatile memory in a computer system serving as a server or a client in that case may be included and a program that holds a program for a certain period of time may be included. Further, the above program may be for realizing a part of the functions described above, or may be realized by a combination with the program already recorded in the computer system. .

  As mentioned above, although embodiment of this invention was explained in full detail with reference to drawings, the concrete structure is not restricted to this embodiment, The design etc. of the range which does not deviate from the summary of this invention are included.

DESCRIPTION OF SYMBOLS 1,1-1 to 1-4 Physical server 2, 2-1 to 2-6 Virtual server 3 Center server 4 Resource control server 10 Assignment management part 11 System information storage part 12 Control information storage part 13 Measurement information storage part 14 Arrangement Information storage unit 20 Measurement processing unit 30 Classification calculation unit 40 Allocation calculation unit 50 Resource allocation unit 100 Server system

Claims (6)

For each of a plurality of virtual machines to which resources of the information processing device are allocated, a measurement processing unit that measures resource usage for a predetermined period;
Based on the resource usage measured by the measurement processing unit, the predicted value of the resource usage and the degree of change in the predicted load are calculated for each of the plurality of virtual machines, and the degree of change in the predicted load is equal to or greater than a predetermined threshold. The resource request amount is calculated based on the average resource usage amount of the virtual machine calculated based on the resource usage amount for the predetermined period and the predicted value of the resource usage amount, while the predicted load And the deviation of the average resource usage of the virtual machine and the resource usage calculated respectively based on the resource usage for the predetermined period, and the information processing device A resource contention value indicating the degree of resource contention calculated based on the average resource usage rate and the deviation of the resource usage; A distribution calculator for calculating the resource demand based,
A resource allocation unit that rearranges and allocates resources of the information processing apparatus to each virtual machine based on a resource request amount for each of the plurality of virtual machines calculated by the distribution calculation unit;
A resource allocation device comprising: A classification processing unit that classifies each of the plurality of virtual machines into a predetermined classification based on the average resource usage;
The resource allocation apparatus according to claim 1, wherein the distribution calculation unit calculates the resource request amount for each of the plurality of virtual machines in association with the classification classified by the classification processing unit. For each of a plurality of virtual machines to which resources of the information processing device are allocated, a measurement step for measuring a resource usage amount for a predetermined period;
Based on the resource usage measured in the measurement step, for each of the plurality of virtual machines, a predicted value of the resource usage and a degree of change in the predicted load are calculated, and the degree of change in the predicted load is equal to or greater than a predetermined threshold. If there is, the resource request amount is calculated based on the average resource usage amount of the virtual machine calculated based on the resource usage amount for the predetermined period and the predicted value of the resource usage amount, while the predicted load When the degree of change is less than the predetermined threshold, the deviation of the average resource usage of the virtual machine and the resource usage calculated based on the resource usage for the predetermined period, and the resource in the information processing apparatus A resource indicating the degree of resource contention calculated based on the average usage rate and deviation of the resource usage A distribution calculating step of calculating the resource request amount based on the focus value,
A resource allocation step of rearranging and allocating resources of the information processing apparatus to each virtual machine based on a resource request amount for each of the plurality of virtual machines calculated in the distribution calculation step;
A resource allocation method characterized by comprising: A classifying step of classifying each of the plurality of virtual machines into a predetermined category based on the average resource usage;
The resource allocation method according to claim 3, wherein the allocation calculation step calculates the resource requirement amount corresponding to the classification classified in the classification step for each of the plurality of virtual machines. On the computer,
For each of a plurality of virtual machines to which resources of the information processing device are allocated, a measurement step for measuring a resource usage amount for a predetermined period;
Based on the resource usage measured in the measurement step, for each of the plurality of virtual machines, a predicted value of the resource usage and a degree of change in the predicted load are calculated, and the degree of change in the predicted load is equal to or greater than a predetermined threshold. If there is, the resource request amount is calculated based on the average resource usage amount of the virtual machine calculated based on the resource usage amount for the predetermined period and the predicted value of the resource usage amount, while the predicted load When the degree of change is less than the predetermined threshold, the deviation of the average resource usage of the virtual machine and the resource usage calculated based on the resource usage for the predetermined period, and the resource in the information processing apparatus A resource indicating the degree of resource contention calculated based on the average usage rate and deviation of the resource usage A distribution calculating step of calculating the resource request amount based on the focus value,
A resource allocation step of rearranging and allocating resources of the information processing apparatus to each virtual machine based on a resource request amount for each of the plurality of virtual machines calculated in the distribution calculation step;
A resource allocation control program for executing the process. A classifying step of classifying each of the plurality of virtual machines into a predetermined category based on the average resource usage;
The resource allocation control program according to claim 5, wherein the allocation calculation step causes the computer to execute a process of calculating the resource request amount corresponding to the classification classified in the classification step for each of the plurality of virtual machines. .

Images