WO2017056238A1 - Vm assignment management system and vm assignment management method - Google Patents

Abstract

This VM assignment management system is provided with a deployment management unit and a virtual machine assignment management unit. The deployment management unit: generates, for a system to be created, inter-VM communication relationship graph information which indicates that virtual machines are connected to each other via a logical network, on the basis of logical configuration templates, a logical configuration template management table, and a definition table for calculating an amount of communication on a per-middleware basis, said templates and said tables being retained by the deployment management unit; and requests the virtual machine assignment management unit to provision a virtual machine system over physical machines. The virtual machine assignment management unit then provisions the virtual machine system over physical machines, as requested, on the basis of physical configuration graph information, a virtual-to-physical machine mapping table, and a logical-to-physical network mapping table, which are retained by the virtual machine assignment management unit, and on the basis of the inter-VM communication relationship graph information. Further, the manner in which the physical machines and physical switches are allocated is changed according to the manner in which the logical network indicated by the inter-VM communication relationship graph information is connected.

Description

VM placement management system and VM placement management method

The present invention relates to a VM placement management system and a VM placement management method, and in particular, a virtual machine (hereinafter referred to as a VM (Virtual Machine)) on a cloud infrastructure (IaaS (Infrastructure as a Service)) on which a plurality of business systems operate. In addition, the present invention relates to a VM placement management system and a VM placement management method suitable for performing optimal VM placement that conforms to the specifications of a physical network.

In recent years, due to the rapid spread of the Internet, software and data that were previously installed and used on local computers, or the technical infrastructure (servers, etc.) for providing them have become necessary through networks such as the Internet. In response, cloud services, which are services provided to users, are attracting attention.

As one form of this cloud service, there is a use form called IaaS (Infrastructure as a Service). In IaaS, a platform (infrastructure such as a virtual machine and a network) for constructing and operating a computer system itself is provided as a service via the Internet.

Also, for example, Patent Literature 1 is a technology aiming at ensuring performance stability in a virtualized environment and improving the aggregation rate. In Patent Document 1, in a system composed of a plurality of physical servers, a virtual server operating on a certain physical server is transferred to another physical server using live migration technology when a load is unbalanced among the physical servers. Move to. As a result, the load is leveled, and the risk of performance degradation is reduced.

US Patent No. 8095929

The technology described in Patent Document 1 can be used when a VM is configured with an optimal configuration on a physical machine in a system composed of a plurality of physical machines on an IaaS platform. However, in the method described in Patent Document 1, the VM placement destination physical server is determined based on the CPU and memory utilization in determining the VM placement destination, and no consideration is given to the network. Regarding the network, it is necessary to consider not only the physical machine that is the end point of communication but also the bandwidth of all network devices that are routed in the middle of communication, and the topology of the physical network must be considered. Further, the optimal communication path may differ depending on the number of communication destinations of applications on the tenant (cloud service service recipient) and the communication mode (unidirectional, bidirectional, synchronous, asynchronous). For example, since the load balancer communicates with a large number of other VMs, operation on a physical machine with a large amount of communication of other VMs is not desirable. On the other hand, a Web server and an AP server that are determined to communicate only with a specific server can be localized in the same physical machine by placing them on the same physical machine. This is desirable from the cloud administrator's point of view to reduce the possibility of communication affecting other tenants. In the prior art, such a viewpoint has not been considered. In addition, since big data analysis software represented by open source Hadoop performs asynchronous communication among a large number of VMs, it is desirable that the software be separated into clusters that have as little influence as possible from other tenants. If there is no limit on job completion time, it is desirable for cloud administrators to share resources between the same type of applications. However, in applications that place importance on synchronous response time, there are many switches for data transfer that are critical paths. It is not desirable to relay

The present invention has been made to solve the above-mentioned problems, and its purpose is to meet the specifications of the physical network and to create an optimal virtual machine on a cloud platform on which a plurality of business systems operate. An object of the present invention is to provide a VM placement management system that dynamically performs VM placement so that a service can be provided.

The configuration of the VM placement management system of the present invention is a VM placement management system that provisions a virtual machine (VM) system to one or more physical machines connected by a network by a computer, and sets up logical information of the system to be constructed. A virtual machine placement management unit that has a deployment management unit to hold, physical information of a system to be built, and relationship information between virtual machines and physical machines, and provisions virtual machines to physical machines in response to requests from the deployment management unit Are provided. The deployment management unit includes a logical configuration template that associates information on a virtual machine, information on software running on the virtual machine, information on a system to be constructed, a logical configuration template management table that holds the logical configuration template, and a logical configuration template And a communication amount calculation definition table for each middleware that defines bandwidth information at the time of communication, and the virtual machine arrangement management unit includes a physical switch that constructs a network, Connection structure of physical machines placed on the network, physical configuration graph information that defines bandwidth information in the network, a virtual machine-physical machine correspondence table that associates virtual machines with physical machines, and a virtual that associates logical networks with physical networks A machine-physical machine correspondence table. Then, the deployment management unit, based on the logical configuration template, the logical configuration template management table, and the communication amount calculation definition table for each middleware, establishes a communication relationship between VMs in which virtual machines are connected by a logical network. Generates graph information and requests the deployment management unit to provision the physical machine of the virtual machine system. The virtual machine arrangement management unit receives the physical configuration graph information, the virtual machine-physical machine correspondence table, and the virtual machine- Based on the physical machine correspondence table and the inter-VM communication relationship graph information, provision of the requested virtual machine system to the physical machine is performed.

According to the present invention, when creating a virtual machine on a cloud platform on which a plurality of business systems operate, a VM that dynamically arranges VMs so as to provide an optimal service that conforms to the specifications of the physical network. An arrangement management system can be provided.

It is a system configuration figure of VM arrangement management system. It is a hardware block diagram of the computer which mounts VM arrangement management system. It is a figure which shows the structural example of the structure and topology information 151 (the 1) It is a figure which shows the structural example of the structure and topology information 151 (the 2). 6 is a diagram illustrating an example of a logical configuration template management table 121. FIG. It is a figure which shows the structural example of the communication amount calculation formula definition table classified by middleware (the 1). It is a figure which shows the structural example of the communication amount calculation formula definition table classified by middleware (the 2). It is a figure which shows the structural example of the communication amount calculation formula definition table classified by middleware (the 3). It is a figure explaining the definition and kind of a logical network. It is a figure which shows the structural example of the communication relation graph information 141 between VMs (the 1). It is a figure which shows the structural example of the communication relation graph information 141 between VMs (the 2). It is a figure which shows the example of the cloud infrastructure physical structure graph information 131 (the 1). It is a figure which shows the example of the cloud infrastructure physical structure graph information 131 (the 2). 6 is a diagram illustrating an example of a virtual machine-physical machine association table 132. FIG. 6 is a diagram illustrating an example of a logical network-physical network correspondence table 133. FIG. It is the figure which showed the image by the graph of the logical network-physical network matching table 133 of FIG. 10A. It is a general flowchart which shows a VM provisioning process from the initial construction request | requirement of the cloud user 16. FIG. It is a flowchart which shows the production | generation process of the communication relation graph information 141 between VMs. It is a flowchart which shows the server addition request process by the cloud user 16 (the 1). It is a flowchart which shows the server addition request | requirement process by the cloud user 16 (the 2). It is a flowchart which shows the procedure which selects a physical machine. It is a flowchart which shows the allocation process of the logical network to a new physical switch. It is a flowchart which shows the selection process of the physical machine which minimizes the physical switch of the communication used from the allocated logical network group. It is a flowchart which shows the process which selects the physical machine with a low competition probability of network resource utilization.

Hereinafter, an embodiment according to the present invention will be described with reference to FIGS.
(I) System Configuration of VM Placement Management System Hereinafter, a system configuration of a VM placement management system according to an embodiment of the present invention will be described with reference to FIGS. 1 and 2. FIG. 1 is a system configuration diagram of a VM arrangement management system. FIG. 2 is a hardware configuration diagram of a computer on which the VM arrangement management system is mounted.

The VM placement management system of this embodiment performs management of a tenant's system configuration on the IaaS platform, and generation and placement (provisioning) of VMs based on the system configuration.

The VM placement management system includes a deployment management unit 10 and a virtual machine placement management unit 11 as shown in FIG.

The deployment management unit 10 and the virtual machine arrangement management unit 11 can be implemented as programs that run on the computer 200 shown in FIG. As shown in FIG. 2, the computer 200 has a form in which a CPU (Central Processing Unit) 201, a memory 202, a storage device 203, an input device 204, an output device 205, and a communication device 206 are coupled by a bus. The CPU 201 controls each unit of the computer 200, refers to data on the memory 202, and executes the loaded program. The memory 202 is also called a main storage device, and is a volatile semiconductor storage device. The storage device 203 is a device that stores large-capacity data such as an HDD (Hard Disk Drive) or an SSD (Solid State Drive), and stores programs, tables described later, and other data. The input device 204 is a device for inputting information such as a mouse and a keyboard. The output device 205 is a device that displays and prints information such as a liquid crystal display and a printing device. The communication device 206 is a device that is connected to the network 207 and can exchange information, and is, for example, an Ethernet (registered trademark) network card. The shared data store 14 is a mechanism in which the deployment management unit 10 and the virtual machine arrangement management unit 11 share information.

The deployment management unit 10 and the virtual machine arrangement management unit 11 each store internal management information in the storage device 208. However, the data shared between both systems is designed so that both can access by some sharing means. For example, there is a method in which the shared data store 14 is used to access management information common to both. To realize the shared data store 14, existing means such as NFS (Network File System) and RDB (Relational Data Base) can be used.

The deployment management unit 10 receives tenant creation / deletion and other configuration change (server addition / deletion) requests on the IaaS from the cloud user 16 managing each tenant, and manages the logical system configuration information of the tenant. After editing, the virtual machine placement management unit 11 is requested to provision an actual VM.

The virtual machine arrangement management unit 11 receives the VM provisioning request from the deployment management unit 10, determines the VM arrangement destination while considering the actual physical system usage status, and provisions such as creation and activation of the VM Perform the process.

The deployment management unit 10 receives a business system construction request from the cloud user 16. The cloud user 16 inputs the logical configuration template 15 together with the user ID serving as the identifier and the system ID for identifying the system that requests construction, and requests the deployment management unit to construct a business system. The construction of the business system is performed based on the configuration / topology information 151 described in the logical configuration template 15. This configuration / topology information 151 includes, for example, OASIS TOSCA (https://www.oasis-open.org/committees/tc_home.php?wg_abbrev=tosca), Amazon Web Services CloudFormation service template configuration information, and open source It conforms to the OpenStack HOT format, which is the IaaS infrastructure management software, and includes virtual machine information including the type of virtual machine operating in the tenant and software information operating on the server.

The deployment management unit 10 includes a logical configuration / template reception unit 101, an inter-VM communication relationship graph creation unit 102, and a resource request transmission unit 103.

The logical configuration / template reception unit 101 receives a business system construction request from the cloud user 16, receives input information, and records the user ID, system ID, and logical configuration information template in association with each other in the logical configuration template management table 121. It has a function. Further, based on the recorded information, logical configuration template information corresponding to the business system and a list of VMs provisioned based on the logical configuration template information are returned.

The inter-VM communication relationship graph creation unit 102 has a graph structure of a list of VMs to be provisioned and logical network configuration information indicating communication relationships between VMs based on the VM type and software dependency information included in the logical configuration template information. And is recorded as the inter-VM communication relationship graph information 141.

The resource request transmission unit 103 extracts a list of VMs and storage volumes included in the graph structure created by the inter-VM communication relationship graph creation unit, and sends a request to provision each virtual resource to the virtual machine arrangement management unit 11 Send. This is provided as a general IaaS-based function and will not be described in detail.

The virtual machine arrangement management unit 11 includes a resource request receiving unit 111, an arrangement destination physical machine selection unit 114, an arrangement destination physical network selection unit 113, and a virtual machine-physical machine association management unit 115.

The resource request reception unit 111 receives a VM creation request from the resource request unit 103 of the deployment management unit 10 and transfers the request to the placement destination physical machine selection unit 114.

The placement destination physical machine selection unit 114 is abbreviated as “cloud base physical configuration graph information 131”, “virtual machine-physical machine association table 132”, “virtual network-physical network association table” (hereinafter, “virtual NW—physical NW association table”). 133) and the placement relationship of the VM to be newly provisioned are determined using 133 and the inter-VM communication relation graph information 141. If necessary, the placement destination physical network selection unit 113 is called to assign a logical network to a set of physical switches. The called placement destination physical network selection unit 113 places a logical network in a physical network composed of a group of physical switches based on a network allocation status and an application communication form (topology) according to a procedure described later.

The virtual machine-physical machine correspondence management unit 115 records the placement destination of the VM and the logical network determined by the placement destination physical machine selection unit 114 and the placement destination physical network selection unit 113, and associates them with each other as an identifier. Has a function to return resources. Since the virtual machine-physical machine association management unit is a general management function based on IaaS, detailed description thereof is omitted.
(II) Data Structure of Deployment Management Unit Next, the data structure of the deployment management unit will be described with reference to FIGS. 3A to 5C. 3A and 3B are diagrams illustrating a configuration example of the configuration / topology information 151. FIG. FIG. 4 is a diagram illustrating an example of the logical configuration template management table 121.

The configuration / topology information 151 is data for defining a logical system configuration and software configuration on a virtual machine, which is passed as an input from the cloud user 16 and held in the logical configuration template 15.

The configuration example shown in FIG. 3A is a configuration example of a Web three-tier system, and the configuration example shown in FIG. 3B is a configuration example of a big data analysis system. The three-tier system here is a system constructed by dividing the client-server system into three layers of “presentation layer”, “application layer”, and “data layer”, and the processing of the client and server is divided into a plurality of layers. It aims to be able to flexibly cope with the need to make changes to a certain hierarchy by arranging them separately.

The configuration / topology information 151 includes virtual machine information 31 and software information 32. The virtual machine information 31 indicates the type and number of virtual resources such as VMs and other storage volumes required in the business system. The software information 32 is information relating to software for running on the virtual machine whose specification is indicated by the virtual machine information 31.

The virtual machine information 31 is expressed as a list composed of a set of virtual resource type, virtual resource class name, required amount, and initial number (not shown). The virtual resource type indicates the type of resource, and includes, for example, a VM, a storage volume, and a group. A group is a grouping of VM functions, and is expressed as a pointer to other virtual machine information 31. By using this group, other subsystems can be recursively subordinated to their definition, and can be used to indicate additional units when configuring a cluster to be scaled out. The virtual resource class name is an identifier in the list. The required amount indicates the amount of resource required for the virtual resource, and is determined for each virtual resource type. For example, the required amount in the VM is CPU capacity (clock, number of cores, etc.) and memory capacity. A necessary amount in the storage volume is a storage capacity. The initial number is the number provisioned at the time of initial construction. The initial value is one for each resource type. However, it is possible to request the deployment platform to change the number of units (scale out / in) by a designation from a cloud user after construction or a program that performs external system sizing.

In the example of the virtual machine information 31A shown in FIG. 3A, four types of VMs, that is, a load balancer 311, a Web server 314, an AP server 315, and a DB server 316, and one storage volume are defined. Web servers and AP servers are grouped as a group 312 and are shown to be added in this unit. The initial number of groups 312 is three.

Further, in the example of the virtual machine information 31B shown in FIG. 3B, three types of VMs, HadoopMaster, HadoopTask, and HadoopData, are defined, and storage volumes are also defined. Group 1 is composed of HadoopTask, the initial number is three, and Group 2 is composed of HadoopData, and the initial number is four.

The software information 32 indicates the type of software running on each VM defined by the virtual machine type 31 and the communication status between the software. The software information 32 is expressed as a list having a combination of a software name, an installed virtual resource class name, and one or more communication destination software names (not shown). The software name is an identifier of the corresponding software and needs to be unique in the logical configuration template 15 at a minimum. When the middleware-specific communication amount calculation formula definition table 122 described later is prepared in a batch by the deployment management platform, it is necessary to have a unique identifier for each software in the entire deployment management platform. The system may be configured such that the table 122 is described in the logical configuration template 15 and is defined by each cloud user. The mounting destination virtual resource is for designating the VM on which the software runs with one or more virtual resource class names. The communication destination software name indicates software that performs communication in the business system. Although the name software is used in the present invention, it is inherently necessary to distinguish between a software package to be installed and a process / service of a program executed by the software. In the present invention, both are described as being the same for simplicity. Further, communication with the outside of the system cannot be expressed by the communication destination software name. Therefore, a message from outside the system is expressed by a special element called “sink” while a message outside the “source” system is expressed so that communication with the outside of the tenant can be expressed.

In the example of the software information 32A shown in FIG. 3A, HaProxy 323 operates on the load balancer 311 (specified by the installed virtual resource class name 327), Apache HTTPD 324 operates on the Web server 314, Django 325 operates on the AP server 315, and PostgreSQL 326 operates on the DB server 316. It shows that. In addition, it indicates that HaProxy and Apache HTTPD (notation by arrow 328), Apache HTTPD and Django, Django and PostgreSQL are communicating, respectively. The HaProxy 323 receives a message from the source 321 and transmits the message to the sink 322.

Next, the logical configuration template management table will be described with reference to FIG.

The logical configuration template management table 121 is a table for associating business systems, logical configuration templates, and provisioned resources. As shown in FIG. 4, a user ID 401, system ID 402, system name 403, logical A configuration template 404 and a list including a resource list 405 are configured.

User ID 401 is an identifier for identifying a user. A system ID 402 and a system name 403 are an identifier for identifying a business system and a system name, respectively. The logical configuration template 404 stores a pointer to the logical configuration template that is passed when the business system configuration request illustrated in FIGS. 3A and 3B is requested. The resource list 405 stores a pointer to a resource list 406 for identifying a resource such as a VM created according to the logical configuration template 404.

The resource list 406 associates and manages a virtual resource class name 411 in the logical configuration template, a virtual machine ID 412 that is a resource identifier, and a state 413 that indicates the current virtual machine state, such as the progress of the creation task. Yes. Here, it should be noted that the virtual machine ID 412 is an instance of the class of the virtual resource class name 411 in the object-oriented manner.

Next, the communication amount calculation formula definition table for each middleware will be described with reference to FIGS. 5A to 5C. 5A to 5C are diagrams illustrating a configuration example of a communication amount calculation formula definition table for each middleware.

The communication amount calculation formula definition table for each middleware is a table describing the network bandwidth necessary for network communication for each business system. As shown in FIGS. 5A to 5C, the system ID 511 for identifying the system and the system It is expressed by a set of associated bandwidth calculation definition table 512. The bandwidth calculation definition table 512 stores a pointer to the bandwidth calculation definition table. The bandwidth calculation definition table is a combination table of communication source software (communication source 521) and communication destination software (communication destination 522). The presence / absence of communication is indicated by ○ and × in the figure. For the combination of ◯, the definition table 53 for calculating the communication characteristics is provided. This figure is a representation of the graph structure of the software information 32A shown in FIG. 3A in another format (in the case of the Web three-tier system), and substantially to the communication destination software having the graph structure shown in the software information 32 of FIG. 3A. Each dependent link can be interpreted as having a bandwidth definition table 53. The bandwidth definition table 53 includes a set of a processing type 531, a response time 532, a data transfer frequency 533, and a throughput 534. The process type 531 indicates that the communication is synchronous or asynchronous. The response time 532 is an average response speed to be obtained.

Each value is an immediate value or is expressed as a calculation formula calculated with reference to other variables. In the case of including a variable, it is assumed that a value is determined by a method such as providing a variable column in the logical configuration template and allowing the cloud user 16 to input the value, such as confirming at the time of deployment request.

Also, the example shown in FIG. 5B is an expression in another form of the graph structure of the software information 32A shown in FIG. 3B (in the case of a big data analysis system).

The middleware-specific communication fee calculation formula definition table 122 may be defined by the cloud user 16 in combination with the logical configuration template 15.

Further, as in the example of FIG. 5C, the cloud administrator 17 separately defines a plurality of types in advance (the pre-defined table 50 in the figure), and the pre-defined name of the pre-defined table is given to the cloud user 16 at the time of deployment. 52 methods such as displaying a list and its contents and allowing the cloud user 16 to select an appropriate one, and combining and analyzing data at the time of execution in combination with performance monitoring software to dynamically update the data. In the present embodiment, it is assumed that a value can be acquired at the time of deployment execution.
(III) Shared Data Structure of Deployment Management Unit and Virtual Machine Placement Management Unit Next, a data structure shared between the deployment management unit and the virtual machine placement management unit will be described with reference to FIGS. 6, 7A, and 7B. FIG. 6 is a diagram for explaining the definition and types of logical networks. 7A and 7B are diagrams illustrating a configuration example of the inter-VM communication relationship graph information 141. FIG.

The inter-VM communication relationship graph information 141 includes, for each of all VMs constituting the business system, the communication partner of each VM derived from the logical configuration template 15 by the procedure described later, and the connection form (1: 1 communication, N 1: communication, N: M unidirectional communication, N: N bi-directional communication), and a graph structure for converting a business system communication amount into a resource usage amount.

A logical network is a logical object that represents the communication bandwidth and characteristics between VMs having the type shown in FIG. Using the logical network, it is possible to grasp the locality of communication in the system and the size of the communication band. The logical network has a link with one or more VM nodes. A VM node is an object corresponding to a VM described in the resource list, and each is uniquely identified by a virtual resource ID.

The 1: 1 logical network 61 has a network structure indicating that 1: 1 communication is performed between specific VMs as illustrated in FIG. The VMs that communicate with each other are connected to the one-side terminal 611 of the logical network 61. Each terminal has band information 612 indicating a band necessary for each communication. Here, “1 side terminal” implies that the number of VMs connected to the terminal is one, and hereinafter, “N side terminal” and “M side terminal” are also respectively the same. It is implied that the number of VM connections is N and M (N, an integer of M ≧ 2).

The N: 1 logical network 62 has a network structure in which a specific VM as shown in FIG. 6B communicates with many other VMs. This is the case when the load balancer communicates with many worker nodes. The VM connected to the 1-side terminal 621 communicates with other VMs connected to the N-side terminal, and each terminal has band information 623 and 624. The band information 624 of the 1 side terminal is expressed as a sum of bands necessary for communication with N VMs.

In the N: M unidirectional logical network 63, VMs as shown in FIG. 6C are classified into two clusters, and VMs in one cluster are 1: N respectively with VMs in the other cluster. It is a network structure which shows performing communication. The feature is that there is no communication between VMs in the same cluster. This is a communication pattern in which the AP server cluster of the Web 3-tier system accesses the Memcache cluster. The VM in one cluster is connected to the M-side terminal 631, the VM in the other cluster is connected to the N-side terminal 632, and each terminal has band information 633 and 634. The band information of the M side terminal is the total amount of communication with the VM connected to the N side terminal. Similarly, the band information of the N-side terminal is the total amount of communication with the VM of the M-side terminal.

The N: N bidirectional logical network 64 has a network structure in which VMs in a cluster communicate with each other as shown in FIG. The VMs in the cluster are connected to the N-side terminal 641 and communicate with other VMs, and the terminal has band information 642. This corresponds to a Hadoop Shuffle operation or an MPI node in HPC.

FIG. 7A is a configuration example of the communication relationship graph information 141 between the Web 3-tier systems VM in the configuration / topology information 151 illustrated in FIG. 3A, and FIG. 7B is a VM of the analysis system in the configuration / topology information 151 illustrated in FIG. 3B. 5 is a configuration example of inter-communication relationship graph information 141. The inter-VM communication relationship graph information 141 is automatically generated by a method described later.
(IV) Data Structure of Virtual Machine Placement Management Unit Next, a data structure of the virtual machine placement management unit will be described with reference to FIGS. 8A to 10B. 8A and 8B are diagrams illustrating examples of the cloud-based physical configuration graph information 131. FIG. FIG. 9 is a diagram illustrating an example of the virtual machine-physical machine association table 132. FIG. 10A is a diagram showing an example of the logical network-physical network correspondence table 133. FIG. 10B is a diagram showing a graph of the logical network-physical network correspondence table 133 in FIG. 10A.

The cloud-based physical configuration graph information 131 used in the virtual machine arrangement management unit 11 is information that records the physical machine group in the data center, the connection relationship between the switches that connect the physical machines, and individual identification numbers and specifications. It is expressed as a graph structure that retains the topology (structure) of the connection relationship. In order to represent physical configuration information in the data center, each L2 / L3 switch and physical machine in the data center is expressed as a node 811, and a network connection between the nodes is used as an edge. Each node has a physical resource class representing a type such as a switch or a physical machine (synonymous with a physical server in this embodiment), and a node ID for uniquely identifying each node as an attribute value of the node.

In addition, existing techniques such as conversion to an RDB structure using RDF (http://www.w3.org/RDF/), GraphDB, and ActiveRecord have been established as a general expression / management method of the graph structure. .

Specific examples of the cloud-based physical configuration graph information 131 are as shown in FIGS. 8A and 8B. The cloud-based physical configuration graph information 131 illustrated in FIG. 8A is a configuration example simulating a general data center, and the cloud-based physical configuration graph information 131 illustrated in FIG. 8B is used in a large-scale IaaS in recent years. This is a configuration example that simulates a data center that uses a plurality of SDN (Software Defined Network) technologies and has a network having a plurality of paths. It should be noted that the switches 4 to 7 in FIG. 8A and the switches 4 to 8 in FIG. 8B are respectively connected to the physical machine group surrounded by a square in the lower part of the figure in a 1: 1 ratio. Further, it is indicated that both the switch 1 and the switch 2 and the switch 1 and the switch 3 have the band 813 (40 Gps).

As shown in FIG. 9, the virtual machine-physical machine correspondence table 132 includes a physical machine ID 901, a physical machine IP address 902, a physical machine CPU core number 903, a physical memory size 904, a user ID 905, a virtual machine ID 906, and a virtual machine. Each field includes an IP address 907, a virtual machine CPU core count 908, and a virtual memory size 909. This is used to associate and manage physical machines on which virtual machines are operating, and extract the associated physical machines or virtual machines from each other using the identifiers of both as search keys. In addition, when newly allocating a VM, it is also used to extract a physical machine having a corresponding resource amount available. If there is no VM running on the physical machine, it is assumed that an entry for the physical machine in which the VM information is not described is included as shown in the figure. As a result, it is possible to easily search for physical machines that have available resources. For the association, the same technology as that based on a general IaaS basis is used.

As shown in FIG. 10A, the logical network-physical network correspondence table 133 includes fields of switch ID 1001, terminal-switch port mapping 1002, user ID 1003, system ID 1004, and logical NWID 1005.

The logical network-physical network association table 133 is a table showing the association of logical networks operating on the physical switch. Here, that the logical network is operating means that communication between VMs belonging to the logical network is performed on the switch. In the example of FIG. 10A, a logical network is uniquely identified by a combination of a user ID 1003, a system ID 1004, and a logical network ID 1005. In addition, since a VM belonging to a specific logical network may relay a plurality of devices, one logical network may be associated with a plurality of switch IDs 1001 at the same time. FIG. 10B shows an image based on a graph of correspondence of the configuration described in the configuration example of FIG. 10A to an actual physical device, and the physical configuration assumes the configuration described in FIG. 8A.
(V) Processing of VM Arrangement Management System (V-1) VM Provisioning Processing from Initial Construction Request of Cloud User 16 First, an outline of VM provisioning processing from the initial construction request of the cloud user 16 will be described with reference to FIG. FIG. 11 is a general flowchart showing the VM provisioning process from the initial construction request of the cloud user 16.

Hereinafter, a procedure when the deployment management unit 10 receives a tenant creation request on the IaaS from the cloud user 16 managing each tenant and provisions VM resources based on the request will be described.
(Step 1101)
First, the cloud user 16 registers the logical configuration template 15 as shown in FIGS. 3A and 3B. The cloud user 16 uses the means such as GUI, CLI, or REST API (http://en.wikipedia.org/wiki/REST), and the user ID 401 and the original data of the system ID 402 shown in FIG. The defined logical configuration template 15 is input to the deployment management unit 10. The logical configuration template 15 is transmitted to the deployment management unit 10 in a text format or a graph structure displayed and managed using a GUI.
(Step 1102)
Next, the logical configuration template receiving unit 101 associates the user ID, the system ID, and the logical configuration template 15 input from the cloud user 16 in step 1101 with the logical configuration template management table 121 shown in FIG. save
(Step 1103)
Next, the inter-VM communication relationship graph creation unit 102 uses the logical configuration template registered in step 1101 and the communication fee calculation formula definition table 122 for each middleware as shown in FIGS. The graph information 141 is generated (the procedure for generating the inter-VM communication relationship graph information 141 will be described in detail later with reference to FIG. 12) and stored in the shared data store 14. Note that the inter-VM communication relationship graph information 141 may be pre-defined in the deployment management unit, or may be information explicitly specified by the cloud user 16 in step 1101.
(Step 1104)
Next, the logical configuration template receiving unit 101 includes a VM list (hereinafter referred to as “newly created VM list”) and a storage volume list (hereinafter referred to as “newly created”) included in the inter-VM communication relationship graph information 141 generated in step 1103. "Volume List") is extracted, the row identified by the system ID in the logical configuration template management table 121 is searched, and the newly created VM list 1153 and the newly created volume 1154 are displayed in the resource list 405 in the row. 413 is registered as “waiting for execution”.
(Step 1105)
Next, the resource request transmission unit 103 extracts each row in the resource list 405 registered in step 1104 that has a status 413 of “waiting for execution” and issues a provisioning request to the virtual machine arrangement management unit 11 in order. .

(V-2) Generation Process of Inter-VM Communication Relationship Graph Information 141 Next, the generation process of the inter-VM communication relationship graph information 141 will be described with reference to FIG. FIG. 12 is a flowchart showing a process for generating the inter-VM communication relationship graph information 141.

This generation process is executed by a program that generates the inter-VM communication relationship graph information 141. In this program, template information registered in the logical configuration template 404 (illustrated in FIG. 3A, FIG. 3B, etc.) is used as an input. A communication relationship graph 141 is created.

In the flowchart shown in FIG. 12, one group (which is an expansion unit of the VM function (logical server) shown in FIG. 3A or the like) is received as an argument, and a communication relationship graph between VMs belonging to the group is calculated. And return the graph structure. By applying this algorithm recursively to each group in the system, an inter-VM communication relationship graph between servers organized in groups can be acquired. The initial value of the argument (group) is a reference link pointing to the logical configuration template 404 stored in the logical configuration template management table 121 in association with the business system uniquely specified by the user ID and the system ID.
(Step 1201)
First, a graph structure storage area (hereinafter referred to as “communication graph structure”, not shown) is generated. In this data area, a data structure representing the VM in the logical network shown in FIG. 6, the logical network, and the link therebetween is stored. For the data representation of the graph structure, a structure such as a structure or a pointer that is generally used for computer programming can be used. In the initial state, the communication graph structure is generated without any data structure.
(Step 1202)
Next, the virtual resource types included in the group given by the argument (hereinafter referred to as “communication relationship extraction target group”) are acquired with reference to the logical configuration template 404 as shown in FIGS. 3A and 3B.
(Steps 1203a and 1203b)
(Recursion processing with depth priority) Next, referring to the logical configuration template 404, if a group is included in the communication relationship extraction target group (step 1203a: YES), the child group is used as an argument. Perform this process recursively and add the result to the communication graph structure.
(Step 1204)
The number of virtual machine node information (hereinafter referred to as “newly generated VM”) described in the initial number attribute of the communication relationship extraction target group is generated based on the virtual resource type and added to the communication graph structure. The virtual resource class name described in the resource type 310 is used for the resource class of the generated resource.
(Steps 1205a and 1205b)
Steps 1206, 1207a-1207b (steps 12071 to 12076) are performed on each newly created VM created in step 1204 (hereinafter simply referred to as VM in the loop).

(Step 1206)
All the software information 422 included in the software information (illustrated by 32A and 32B) of the logical configuration template 404 is acquired software information (hereinafter referred to as “installed software group”) with a link to the VM.

(Steps 1207a and 1207b)
The following steps (12071) to (including 12076) are performed on each of the connection links 424 (hereinafter simply referred to as “links”) in FIG. 4 to which each of the installed software groups is connected.

(Step 12071)
Obtain linked software information (hereinafter referred to as “connected software group”) (step 12072)
In the communication graph structure, with reference to the logical configuration template 404, all the virtual machines of the resource class 410 to which the mounted links are established are extracted from the connection destination software group (hereinafter, the extracted virtual machines are referred to as “connection destination VM group”). ).

(Step 12073)
A 1: 1 logical network node is created for each pair of newly created VMs and VMs in the connection destination VM group (VMs in the newly created VM and VMs in the connection destination VM group) and added to the communication graph structure To do.

(Step 12074)
In the 1: 1 logical network node in the communication graph structure generated in step 1208, logical network nodes having the same connection destination are aggregated to create and replace an N: 1 logical network node. When replacing, the bandwidth information of each terminal is updated according to the communication amount between VMs described in the middleware communication amount calculation formula definition table 122 and the bandwidth definition of each terminal.

(Step 12075)
In the N: 1 logical network node in the communication graph structure, logical network nodes having the same connection destination set at the N-side terminal are aggregated to create and replace an N: M unidirectional logical network node. When replacing, the bandwidth information of each terminal is updated according to the communication amount between VMs described in the middleware communication amount calculation formula definition table 122 and the bandwidth definition of each terminal.

(Step 12076)
In the N: M unidirectional logical network node in the communication graph structure, when the VM group of the N side terminal and the M side terminal is the same, the N: M bidirectional logical network node is replaced. When replacing, the bandwidth information of each terminal is updated according to the communication amount between VMs described in the middleware communication amount calculation formula definition table 122 and the bandwidth definition of each terminal.
(Step 1208) The communication graph structure is returned as a result of this processing.
(V-3) Server Addition Request Processing by Cloud User 16 Next, server addition request processing by the cloud user 16 will be described with reference to FIGS. 13A and 13B.

FIG. 13A and FIG. 13B are flowcharts showing server addition request processing by the cloud user 16.

This is a processing procedure for determining a VM placement destination when the deployment management unit 10 receives a VM addition request from the cloud user 16 after the construction of the business system.
(Step 1301)
First, the cloud user 16 requests a scale-out by specifying the virtual resource class name and the additional number included in the logical configuration template specified when building the business system identified by the user ID, the system ID, and the system ID. To do.
(Step 1302)
Next, the deployment management unit 10 adds the virtual resource class name and the additional number included in the request of the cloud user 16 to the value of the initial number attribute 313 in the logical configuration template management table 121, and the above-described step 1102. The logical configuration template management table 121 is updated according to the procedure. After that, according to the procedure shown in FIG. 12, the inter-VM communication relationship graph generation unit 102 updates the inter-VM communication relationship graph information 141, specifies the virtual resource name and the added virtual machine ID 412, and sends a VM addition command to the virtual machine. The data is transmitted to the arrangement management unit 11.
(Step 1303)
Next, the resource request reception unit 111 in the virtual machine arrangement management unit 11 receives a VM addition request from the deployment management unit 10. Here, the VM addition request includes a virtual machine ID, a necessary resource amount (CPU, memory), and the like.

The resource request accepting unit 111 activates the placement destination physical machine selection task in the placement destination physical machine selection unit 114 using the specified list of VMs as an argument.
(Steps 1304a and 1304b)
The placement destination physical machine selection unit 114 executes Steps 1305 to 1310 for each VM in the VM list specified in Step 1302 (hereinafter referred to as “additional VM”).

(Step 1305)
The placement destination physical machine selection unit 114 executes the following steps 13051 to 13055, and extracts a physical machine group to which a VM and a logical network in the business system have been allocated and a required resource amount is available (business system) Allocated physical machine candidate acquisition processing (FIG. 13B)).

(Step 13051)
The placement destination physical machine selection unit 114 refers to the virtual machine-physical machine association table 132 shown in FIG. 9 and calculates the actual resource amount and the allocated resource amount of each physical machine in the cloud infrastructure. Then, these two differences are taken to calculate the unallocated resource amount. A list of physical machines having a free space to be allocated to the newly requested virtual machine resource amount (hereinafter referred to as “free physical machine group”) is extracted.
(Step 13052)
If one or more physical machines corresponding to the free physical machine group 1352 are found in step 13051, steps 13053 to 1355 are executed. If there is no corresponding physical machine, an allocation failure in step 1310b described later is executed.

(Step 13053)
A VM node corresponding to the virtual resource ID to be added (hereinafter referred to as “new VM”) is extracted from the inter-VM communication relationship graph information 141. Next, all logical networks connected to the new VM are extracted (hereinafter referred to as connected logical network group).

(Step 13054)
With reference to the logical network-physical network correspondence table 133 shown in FIG. 10A, all the physical switches used by the respective logical networks in the connected logical network group are acquired (hereinafter referred to as “logical NW assigned switches”). Group)). For example, in the example of FIG. 10A, the logical NW assigned switch group used by the logical network “Web server AP server” is “switches 2, 4 and 5”.

(Step 13055)
With reference to the cloud-based physical configuration graph information 131, a physical machine connected to a switch in the logical NW assigned switch group (hereinafter referred to as “physical machine candidate group”) is extracted from the free physical machine group. When there are a plurality of connection destination logical network groups connected to the new VM 1353, those having connections with the logical NW assigned switch group 1355 corresponding to each connection destination logical network group are extracted as physical machine candidate groups.
(Step 1306)
If there is one or more VMs corresponding to the physical machine candidate group 1356 in step 1305, steps 1307 and 1308 are executed, and one physical machine is selected from the corresponding physical machine list based on the network requirements. If there is no corresponding physical machine, a logical network is allocated on the physical switch and a physical machine is selected by the procedures of steps 1309 and 1310.
(Step 1307)
A physical machine is selected based on the network requirements (search / selection of a physical machine to which a business system has been assigned) according to the procedure described later in FIG.
(Step 1308)
A virtual machine is generated and arranged (provisioned) for the physical machine selected in Step 1307. Since this procedure is the same as a general IaaS VM provisioning procedure, a detailed procedure is omitted.
(Step 1309)
According to the procedure described later in FIG. 15, logical network assignment processing is performed (logical network assignment processing to a new physical switch), and one physical machine is selected from the free physical machine group 1352.
(Steps 1310a and 1310b)
In step 1309 (or step 13052), if logical network allocation fails, error processing is performed and the process ends.

(V-4) Search / Selection of Physical Machine Already Assigned to Business System Next, a procedure for selecting a physical machine will be described with reference to FIG.

FIG. 14 is a flowchart showing a procedure for selecting a physical machine.

The procedure shown in FIG. 14 has free resources in the CPU and memory, and all the logical networks of the business system are connected. That is, the bandwidth of the physical machine on which the VM of the business system is already operating is limited. It is a process of selecting physical machines that do not overflow. According to this procedure, if there is a business system that has already been deployed, the VM is preferentially assigned to the physical machine on which the VM operates, and communication is localized. Thereby, when tenants are mixed as an entire IaaS platform, it is possible to avoid that many tenants are affected by performance due to a communication bottleneck.
(Steps 1401a and 1401b)
Steps 1402 to 1404 are performed on all physical machines in the physical machine candidate list (hereinafter referred to as “candidate physical machines”).
(Step 1402)
By arranging VMs, network packets for inter-VM communication increase, and all physical switches that may consume bandwidth are extracted (hereinafter referred to as “switch group with VM placement effect”).

The switch group with VM placement influence is calculated by the following steps 14021 and 14022.

(Step 14021)
In accordance with the same procedure as in step 13053, all the logical networks connected to the additional VM (hereinafter referred to as “VM connection logical network group A”) are extracted from the inter-VM communication relationship graph 141. For each VM connection logical network group A, all the physical switches to which the VM connection logical network is assigned (hereinafter referred to as a “logical network assigned switch group”) are extracted by the same procedure as in step 13054. .

(Step 14022)
For each of the extracted logical network assigned switch groups, the respective terminals on the 1 side, N side, and M side of the logical network are mapped with reference to the logical NW-physical NW correspondence table 133 shown in FIG. 10A. Get the switch port. If there is a corresponding switch port, the physical switch is added to the switch group with VM placement influence.
(Step 1403)
Using the bandwidth information included in the inter-VM communication relationship graph information 141 shown in FIG. 6, the network bandwidth of each port when the VM is placed on the candidate physical machine is recalculated.
(Step 1404)
The maximum value in the network bandwidth of each port calculated in step 1403 is set as the physical machine bandwidth penalty value. If the bandwidth penalty value exceeds the physical switch port or physical machine network bandwidth value, Since the candidate physical machine is unsuitable as a VM placement destination candidate, it is removed from the physical machine candidate list.
(Steps 1405a and 1405b)
If all the physical machines have been removed from the VM placement destination candidates in the step 1404, the processing is terminated as an error, and the process ends.
(Step 1406)
From the physical machines not deleted in step 1404, the physical machine having the smallest bandwidth penalty value calculated in step 1404 is selected, and the process ends.

(V-5) Logical Network Assignment Processing to New Physical Switch Next, logical network assignment processing to the new physical switch will be described with reference to FIG. FIG. 15 is a flowchart showing logical network assignment processing to a new physical switch.

If the logical network is not connected to the physical switch in the step 1309, it is necessary to connect the logical network to the physical switch in order to operate the business system. At this time, the physical switch suitable for the communication form of the application is selected as the placement destination by selecting the physical switch having the best conditions for the resources allocated to the logical network.
(Step 1501)
In accordance with the same procedure as in step 13053, all the logical networks connected to the additional VM (hereinafter referred to as “VM connection logical network group B”) are extracted from the inter-VM communication relationship graph 141.
(Step 1502)
With reference to the logical NW-physical NW correspondence table 133 shown in FIG. 10A, all logical networks that already have logical networks assigned to physical switches are extracted (hereinafter referred to as “allocated logical network group”).
(Step 1503)
It is determined whether or not one or more logical networks corresponding to the assigned logical network group exist.
(Step 1504)
When one or more logical networks corresponding to the assigned logical network group exist, a physical machine with a small number of physical switches used for communication in the system to be deployed is selected by the processing shown in the flowchart of FIG.
(Step 1505)
In step 1502, if there is no assigned logical network in the physical switch, a physical machine that is separated from the assigned logical network and has a low contention probability of using network resources is selected by the method shown in FIG.

(V-6) Selection of Physical Machine for Minimizing the Number of Physical Switches Used for Communication from Allocated Logical Network Group Next, the number of physical switches for communication used from the allocated logical network group is determined using FIG. A process for selecting a physical machine to be minimized will be described. FIG. 16 is a flowchart showing a physical machine selection process for minimizing the number of physical switches for communication used from the assigned logical network group.

The process shown in the flowchart of FIG. 16 is the same as the process in step 1504 of FIG. 15 in the case where there is a logical network already assigned to the physical switch (hereinafter referred to as “assigned network group”). This is a process of selecting a physical machine that minimizes the number of physical switches used for communication.
(Steps 1601a and 1601b)
Steps 1602 to 1606 are executed for each of the free physical machine groups 1352 extracted in step 13051 (hereinafter referred to as “physical machine candidates”) to minimize the number of physical switches used for communication in the system to be deployed. Select a physical machine.

(Steps 1602a and 1602b)
Steps 1603 to 1605 are executed for each of the assigned logical network groups in step 1502.

(Step 1603)
All physical switches to which the VM connection logical network is assigned (hereinafter referred to as a logical network assigned switch group 1653) are extracted by the same procedure as in step 13054.

(Step 1604)
Between the physical machine candidate and the logical network assigned switch group, the switch having the shortest number of transit switches (hereinafter referred to as “shortest reach switch”) and the number of transit switches to the shortest reach switch (hereinafter “shortest reach switch”) Number)).

This can be obtained by calculating the number of relay switches between each switch in the logical network assigned switch group and the physical machine candidate by the Dijkstra method (a kind of algorithm for the shortest path problem).

(Step 1605)
The number of physical switches used for communication in the system is evaluated using the indicators of steps 16051 to 16052 depending on the connection form (1: 1, 1: N, N: M, N: N) of the assigned logical network 1652. To do.

(Step 16051)
In the case of 1: 1 and N: N, the number of switches included in the logical network assigned switch group is set as the evaluation value (X). By this step, a plurality of servers having an equal communication relationship are arranged under the same network switch as much as possible, and the network traffic of the entire system is reduced.

(Step 16052)
In the case of N: 1 and N: M, the number of the shortest reaching switches is set as the evaluation value (X). This step selects a physical machine that is not far from the existing load balancing cluster.

(Step 1606)
The sum of the evaluation values (X) evaluated for the assigned logical network is taken as the evaluation value of the physical machine candidate.
(Step 1607)
The physical machine candidate with the smallest evaluation value is selected as the VM placement destination.
(Step 1608)
All the transit switches from the shortest arrival switch to the VM placement destination physical machine are associated with the VM connection logical network group B, and the association is additionally recorded in the logical NW-physical NW association table 133 and recorded. At this time, the terminal where the additional VM is connected to the logical network is associated with the port where the physical switch and the VM placement destination physical machine are connected, and the logical NW-physical NW correspondence table 133 is updated.

(V-7) Selection of Physical Machine with Low Network Resource Use Contention Probability Next, processing for selecting a physical machine with low network resource use contention probability will be described with reference to FIG. FIG. 17 is a flowchart illustrating a process of selecting a physical machine having a low competition probability for using network resources.

In the process shown in the flowchart of FIG. 17, in the process of step 1505 of FIG. 15, among the logical networks related to the additional VM, there is no one already assigned to the physical switch (allocated network group). This is a process of selecting a physical machine that is far from the assigned logical network and has a low contention probability of using network resources.
(Steps 1701a and 1701b)
Steps 1702 to 1704 are executed for each logical network in the VM connection logical network group A acquired in Step 1501.

(Step 1702)
With reference to the logical NW-physical NW correspondence table 133, an unassigned switch group to which no logical network is assigned and an assigned switch group to which one or more logical networks are assigned are extracted.

If there is an unassigned switch group, a physical switch and a physical machine with a low possibility of interference with other logical networks are selected by the method of steps 1703a-1703b, and if not, the method of steps 1704a-1704b is selected.

(Steps 1703a and 1703b)
For all switches in the unassigned switch group (hereinafter referred to as “switch candidates”), the switches are evaluated based on the indices of steps 17031 to 17033, and one switch and one physical machine are selected.

(Step 17031)
Using the same method as step 1604 above, the physical switch (hereinafter referred to as the “shortest reachable switch”) having the shortest number of transit switches between the switch candidate and the assigned switch group and the number of transit switches to the shortest reach switch (Hereinafter referred to as “the number of shortest reaching switches”).

(Step 17032)
The shortest number of switches is set as the evaluation value of the switch candidate, and the selection is performed in order from the largest evaluation value (hereinafter referred to as “selection switch”).

(Step 17033)
With reference to the cloud-based physical configuration graph information 133 shown in FIGS. 8A and 8B, it is confirmed whether or not the selection switch is directly connected to any physical machine in the free physical machine group selected in step 1303 above. To do. If there are one or more directly connected physical machines, select one of the physical machines and go to step 1705.
If the corresponding physical machine does not exist, the process returns to step 17032 to select the next physical switch.
(Steps 1704a and 1704b)
For all the physical switches among the switches assigned to the logical network, the switches are evaluated based on the indicators of steps 17041 and 17042, and the physical switch and the physical machine are selected.

(Step 17041)
The bandwidth of the logical network is calculated, the remaining bandwidth up to the maximum performance of the physical switch is calculated, and the evaluation value of the physical switch is selected in order from the largest evaluation value (hereinafter referred to as “selection switch”).

(Step 17042)
With reference to the cloud-based physical configuration graph information 133 shown in FIGS. 8A and 8B, it is confirmed whether or not the selection switch is directly connected to any physical machine in the free physical machine group selected in step 1303 above. To do. If there are one or more directly connected physical machines, select one of the physical machines and go to step 1705.

If the corresponding physical machine does not exist, the process returns to step 17032 to select the next physical switch.
(Step 1705)
One physical machine and one physical switch are selected for each logical network (each logical network in the VM connection logical network group A in steps 1701a to 1701b) by the above processing (steps 1703a to 1703b OR steps 1704a to 1704b). Is done. From these, one physical machine is selected by an arbitrary method.
(Step 1706)
A physical switch connected to the VM placement destination physical machine is associated with the VM connection logical network group, and the association is added to the logical NW-physical NW association table 133 and recorded. At this time, the terminal where the additional VM is connected to the logical network is associated with the port where the physical switch and the VM placement destination physical machine are connected, and the logical NW-physical NW correspondence table 133 is updated.
(VI) Features of VM provisioning according to the form of the logical network of this embodiment According to this embodiment, when it is necessary to additionally arrange a VM in the business system, in addition to the usage amount of the CPU and memory, The VM is arranged in consideration of the communication band between the VMs and communication characteristics (1: 1 communication, N: 1 communication, N: M communication, N: N communication).

If another VM that is in communication with the VM has already been placed on the physical machine, reallocate the bandwidth of the physical switch affected by placing the VM, check the necessary bandwidth, and place the VM At the same time, according to the rules for each communication characteristic, the distance on the network path and the number of network devices to be used are reduced. By reducing the number of network equipment used, the total amount of traffic that traverses between switches is reduced, reducing the possibility of network band interference.

When no VM in communication relationship with the VM is arranged on the physical machine, the VM is arranged so that the distance on the network path from other systems is increased, and the possibility that the network band interferes is reduced.

Also, according to the present embodiment, since 1: 1 communication is local, communication is arranged on the same physical machine as much as possible or on a physical machine with a short distance on the network path. N: N communication is a symmetric intra-cluster communication and does not clearly distinguish between the transmission source and the transmission destination. Therefore, the N: N communication is arranged so that the number of switches used in the entire cluster is reduced within the range satisfying the communication band. This reduces the number of switches used. In N: 1 and N: M communication, communication is asymmetric (for example, server-client communication), so the load may be concentrated on a specific VM, and the method of reducing the total number of switches used will result in insufficient network bandwidth. There is a case. Therefore, in the server-client communication given as an example, by suppressing the distance between the server-side and client-side VMs on the network, it is possible to avoid a decrease in network latency and to avoid an increase in traffic of the entire system. Become.

In addition, the communication characteristics described above can be derived as constraints from information on software (middleware, etc.) arranged on a business system, configuration information such as communication relationships between software, topology information between components, and the like. it can. In the present invention, a constraint condition can be arranged from such configuration information and topology information in the configuration information, and a VM placement destination can be deduced to generate a VM arrangement constraint condition.

10 ... Deployment Management Department
11 ... Virtual machine arrangement management unit 15 ... Logical configuration template 101 ... Logical configuration template reception unit
102 ... Inter-VM communication relation graph creation unit
103 ... Resource request transmission unit
111 ... Resource request reception part
113: Placement physical network selection unit
114: Placement physical machine selection unit
115... Virtual machine-physical machine association management unit
121 ... Logical configuration template management table
122 ... Communication amount calculation formula definition table by middleware
131 ... Cloud-based physical configuration graph information
132 ... Virtual machine-physical machine correspondence table 133 ... Logical network-physical network correspondence table
141 ... Inter-VM communication relation graph information

Claims (9)

A VM placement management system that uses a computer to provision a virtual machine (VM) system to one or more physical machines connected by a network,
A deployment manager that holds the logical information of the system to be built,
A virtual machine placement management unit that holds physical information of a system to be constructed and relationship information between a virtual machine and a physical machine, and provisions a virtual machine to a physical machine in response to a request from the deployment management unit;
The deployment management unit
A logical configuration template that associates virtual machine information with software running on it,
Information on the system to be constructed, a logical configuration template management table that holds the logical configuration template, and
A communication amount calculation definition table for each middleware that defines the relationship between the communication source and the communication destination of the software in the logical configuration template and the bandwidth information at the time of communication;
The virtual machine arrangement management unit
Physical configuration graph information that defines the physical switch that configures the network, the connection structure of the physical machines placed on the network, and the bandwidth information in the network;
A virtual machine-physical machine mapping table for associating virtual machines with physical machines,
A virtual machine-physical machine correspondence table for associating a logical network with a physical network;
Based on the logical configuration template, the logical configuration template management table, and the middleware traffic amount calculation definition table, the deployment management unit is configured to connect between VMs in which virtual machines are connected by a logical network. Generate communication relationship graph information, request provision to the physical machine of the virtual machine system from the deployment management unit,
The virtual machine arrangement management unit requests based on the physical configuration graph information, the virtual machine-physical machine correspondence table, the virtual machine-physical machine correspondence table, and the inter-VM communication relationship graph information. A VM placement management system characterized by provisioning a virtual machine system to a physical machine. When provisioning a new physical machine, the deployment management unit provisions the requested virtual machine system to the physical machine when there are unassigned physical switches and physical machines. The VM arrangement management system according to claim 1, wherein: When the deployment management unit allocates a physical machine to a new logical network, if there is an allocated logical network, the deployment management unit obtains an evaluation value of a physical switch connected to the physical machine and allocates the physical switch to the allocated logical network. 2. The VM placement management system according to claim 1, wherein the provisioning is performed by selecting the physical machine having the smallest evaluation value as a total of the evaluation values of the physical machines as the evaluation value of the physical machine. The evaluation value of the physical switch connected to the physical machine is that the connection form of the allocated logical network is a 1: 1 form in which virtual machines are connected in a one-to-one relationship, or a virtual machine is connected in N to N form. 4. The VM placement management system according to claim 3, wherein the evaluation value of the physical switch is the number of switches of the assigned logical network when N: N form (an integer of N ≧ 2). The evaluation value of the physical switch connected to the physical machine indicates that the connection form of the allocated logical network is N: 1 form (N ≧ 2 integer) in which virtual machines are connected N to 1, or a virtual machine Are connected in an N-to-M manner, and the virtual machines in the cluster to which the M virtual machines belong do not communicate with the virtual machines in the cluster to which they belong, but communicate with only the virtual machines in the cluster to which the N virtual machines belong. When the N: M unidirectional configuration is performed (N, an integer of M ≧ 2), the evaluation value of the physical switch is set to the number of shortest reaching switches of the physical machine of the allocated logical network. Item 4. The VM arrangement management system according to item 3. When a physical machine is newly assigned to a logical network, the deployment management unit, when there is no assigned logical network, sets an evaluation value of an unassigned physical switch of each logical network and an evaluation value of the assigned physical switch. The VM placement management system according to claim 1, wherein the provisioning is performed by selecting the physical machine based on the selection. The evaluation value of the unassigned physical switch is a physical machine and the number of the shortest reaching switches of the physical machine, and the provisioning is performed by selecting a physical switch having a larger evaluation value and a physical machine connected thereto. The VM arrangement management system according to claim 6 The evaluation value of the allocated physical switch is the bandwidth of the logical network and the remaining bandwidth up to the maximum performance of the physical switch, and selects the physical switch having the higher evaluation value and the physical machine connected thereto, and the provisioning The VM placement management system according to claim 6, wherein: A VM placement management system VM placement management system method for provisioning a virtual machine (VM) system to one or more physical machines connected by a network by a computer,
The VM placement management system includes:
A deployment manager that holds the logical information of the system to be built,
A virtual machine placement management unit that holds physical information of a system to be constructed and relationship information between a virtual machine and a physical machine, and provisions a virtual machine to a physical machine in response to a request from the deployment management unit;
The deployment management unit
A logical configuration template that associates virtual machine information with software running on it,
Information on the system to be constructed, a logical configuration template management table that holds the logical configuration template, and
A communication amount calculation definition table for each middleware that defines the relationship between the communication source and the communication destination of the software in the logical configuration template and the bandwidth information at the time of communication;
The virtual machine arrangement management unit
Physical configuration graph information that defines the physical switch that configures the network, the connection structure of the physical machines placed on the network, and the bandwidth information in the network;
A virtual machine-physical machine mapping table for associating virtual machines with physical machines,
Holds a virtual machine-physical machine mapping table that associates a logical network with a physical network,
Based on the logical configuration template, the logical configuration template management table, and the communication amount calculation definition table for each middleware, the deployment management unit is configured between VMs in which virtual machines are connected by a logical network. Generating communication relationship graph information and requesting the deployment management unit to provision a physical machine of a virtual machine system;
The virtual machine arrangement management unit makes a request based on the physical configuration graph information, the virtual machine-physical machine correspondence table, the virtual machine-physical machine correspondence table, and the inter-VM communication relationship graph information. Provisioning to a physical machine of the virtual machine system that has been performed.

Images