WO2012073369A1

WO2012073369A1 - Method of managing virtual machine, computer system and non-temporary computer readable medium

Info

Publication number: WO2012073369A1
Application number: PCT/JP2010/071592
Authority: WO
Inventors: 海斗山田
Original assignee: 株式会社日立製作所
Priority date: 2010-12-02
Filing date: 2010-12-02
Publication date: 2012-06-07

Abstract

One embodiment of the present invention manages a virtual machine within a computer system including a plurality of physical computers communicating via a network. The present management method measures the traffic among a plurality of physical computers within a network. The network traffic is predicted from the result of the measurement of cases where a virtual machine existing on one of the plurality of physical computers is transferred to each transfer destination candidate within the plurality of physical computers. A priority level is determined for the relevant transfer destination candidate on the basis of the predicted traffic. Hereby, it is possible for a system including a network to select a suitable transfer destination for a virtual machine.

Description

Method for managing virtual machine, computer system, and non-transitory computer-readable storage medium

The present invention relates to a method for managing a virtual machine, a computer system, and a non-transitory computer-readable storage medium, and more particularly to movement of a virtual machine between physical computers in the computer system.

In a system in which multiple host servers are connected via a network and virtual machines are running on each host server, the function to move an active virtual machine from one host server to another is quickly known. Yes. This function is generally called live migration.

In general, life migration avoids a situation in which the processing load of a virtual machine running on a certain host server increases, resulting in a shortage of resources used by the virtual machine, thereby causing the host server to become overloaded. Used for. In the life migration, the resource shortage can be solved by moving the virtual server from the current host server to another host server.

In addition, live migration is used in order to maintain the performance guarantee of the virtual server in accordance with SLA (Service Level Agreement) concluded with the user. As described above, the uses of live migration are diversified, and it is important to select an appropriate destination in consideration of system characteristics in live migration.

On the other hand, Patent Document 1 discloses a technique for selecting a destination in consideration of specifications required by a virtual machine and the state of a host server, for example, CPU, power supply, fan, and failure history during live migration. ing. The technique disclosed in Patent Document 1 determines the performance and reliability of the host server on which the virtual machine operates by determining the destination host server from the consistency between the specifications required by the virtual machine and the state of the actual host server. Sex can be maintained.

JP 2008-276320 A

In order to realize proper operation of the system, it is important to properly manage not only the operating status of each host server but also network traffic. However, the conventional technology does not consider network traffic in selecting a destination host server for a virtual machine.

The virtual machine communicates with other host servers via the network in the system. Therefore, moving a virtual machine changes network traffic. In system management, it is important to select virtual machines according to changes in network traffic due to movement and maintain appropriate network traffic.

Therefore, there is a demand for a technique for selecting an appropriate virtual machine migration destination for a system including a network in the migration of virtual machines in a plurality of host servers connected by a network.

One aspect of the present invention is a method for managing a virtual machine in a computer system including a plurality of physical computers that communicate via a network. This management method measures traffic between the plurality of physical computers in the network. The traffic of the network when a virtual machine existing in one of the plurality of physical computers is moved to each of movement destination candidates in the plurality of physical computers is predicted from the measurement result. The priority of the destination candidate is determined based on the predicted traffic.

According to the present invention, it is possible to select a destination of a virtual machine suitable for a system including the network in the movement of the virtual machine in a plurality of physical computers communicating via the network.

It is a figure which shows typically the structural example of the computer system of this embodiment. It is a figure which shows typically the structure of the host server of this embodiment. It is a figure which shows typically the structure of the host server of this embodiment. It is a figure which shows typically the structure of the host server of this embodiment. It is a figure which shows typically the structure of the physical switch of this embodiment. It is a figure which shows typically the hardware constitutions of the management server of this embodiment. It is a figure which shows typically the program and table which implement | achieve the function of the management server of this embodiment. It is a figure which shows the structural example of the host server information table of this embodiment. It is a figure which shows the structural example of the physical switch information table of this embodiment. It is a figure which shows the structural example of the traffic information table between virtual machines of this embodiment. It is a flowchart which shows the flow of the information registration process to the traffic information table between virtual machines by the traffic information acquisition part of this embodiment. It is a figure which shows the structural example of the traffic information table between virtual machines of this embodiment, and a host server. It is a flowchart which shows the flow of a process of the traffic calculation part between virtual machine host servers of this embodiment. It is a figure which shows the structural example of the physical switch prediction traffic table of this embodiment. It is a flowchart which shows the flow of preparation of the prediction traffic table by the traffic calculation part after a virtual machine movement of this embodiment. In this embodiment, it is a figure which shows the traffic before the movement of a virtual machine. In this embodiment, it is a figure which shows the traffic change accompanying the movement of a virtual machine. In this embodiment, it is a figure which shows the traffic change accompanying the movement of a virtual machine. It is a figure which shows the structural example of the movement destination candidate table of this embodiment.

Hereinafter, embodiments for carrying out the present invention will be described with reference to the drawings. For clarity of explanation, the following description and drawings are omitted and simplified as appropriate. Moreover, in each drawing, the same code | symbol is attached | subjected to the same element and the duplication description is abbreviate | omitted as needed for clarification of description.

This embodiment is characterized in that a virtual machine moves between host servers (physical computers) that communicate via a network. In the present embodiment, network traffic after movement of a virtual machine is predicted, and the priority of the destination host server is determined using the predicted traffic. Thereby, it is possible to select an appropriate destination of the virtual machine in the system including the network.

FIG. 1 is a block diagram schematically showing a configuration example of a computer system of this embodiment. The computer system includes a plurality of host servers A1a to C1c (a plurality of physical computers), a physical switch 2, and a management server 3. The host servers A1a to C1c communicate with each other via a network including the physical switch 2. The management server 3 is a physical computer.

The physical switch 2 has a plurality of ports. Specifically, the port A21a to the port E21e are provided. FIG. 1 shows an example of a physical switch 2, and the number of ports depends on the computer system. Host server A1a to host server C1c are connected to port A21a to port C21c, respectively.

In the example of FIG. 1, the ports 11a to 11c of the host server A1a to the host server C1c are connected to the corresponding ports A21a to C21c of the physical switch 2, respectively. In the example of FIG. 1, the ports of the host server A1a to the host server C1 are connected to the ports of the physical switch 2 by wiring, but may be connected via other devices that do not become bottlenecks. The correspondence between the ports of the host servers A1a to C1 and the ports of the physical switch 2 is fixed, and when two host servers that perform communication are determined, the port through which communication data passes is also determined.

A plurality of virtual machines A12a to D12d are operating on the host server A1a to the host server C1c. Specifically, the virtual machine A12a operates on the host server A1a, the virtual machine B12b and the virtual machine C12c operate on the host server B1b, and the virtual machine D12d operates on the host server C1c.

Host server A1a to host server C1c each execute virtualization software, and a virtual machine operates on the virtualization software (see FIGS. 2A to 2C). The virtual machine is software and includes an operating system (OS) in addition to the configuration information of the virtual machine. In this example, the virtual machine is a virtual server, and includes an application program of the virtual server and data for operating the program in addition to the OS.

The management server 3 has a port 31 and is connected to the network 4 through the port. The management server 3 communicates with the physical switch 2 via the network 4 and further communicates with the host servers A1a to C1c via the network 4 and the physical switch 2. The management server 3 may be connected to the physical switch 2 without going through the network 4. The management server 3 manages the virtual machines A12a to D12d.

The management server 3 of the present embodiment has a feature in controlling the movement of the virtual machine. This embodiment is particularly useful for virtual machine live migration (moving a running virtual machine), but can also be applied to other virtual machine movement methods.

In the present embodiment, as an example of movement of a virtual machine, movement of the virtual machine C12c from the host server B1b to another host server will be described. In this example, one of the host server A1a and the host server C1c is selected. The management server 3 refers to the traffic on the network connecting the host servers A1a to C1c and selects an appropriate destination host server. This makes it possible to select an appropriate destination host server for the entire system including the physical switch 2.

In the preferred configuration, the management server 3 determines the destination host server by referring to the specifications of the destination host server in addition to the traffic. Therefore, the configuration of the host server A1a to the host server C1c will be described with reference to FIGS. 2A to 2B.

As shown in FIG. 2A, the host server A1a that is a physical computer includes three CPUs 13a to 13c, a memory 14a that is a main storage device, and device information 15a. The device information 15a is typically stored in a secondary storage device (nonvolatile storage device) of the host server A1a. The device information 15a is information relating to the state of the host server A1a, and includes information on the operation state (use state) in addition to information on the hardware configuration such as the number of CPUs and memory capacity. The device information 15a may be stored in another storage device connected via a network. The device information 15a will be described later.

As described above, the virtualization software 16a is operating on the host server A1a, and the virtual machine A12a is operating thereon. For convenience of explanation, FIG. 2A shows the virtualization software 16a and the virtual machine A12a in the memory 14a. However, data (including programs) required for the processing of the host server A1a is typically secondary. It is stored in the memory 14a from the storage device. Each information (representing data) is stored in the storage area of each information in the memory 14a and the secondary storage device. The memory 14a, the secondary storage device, and the set thereof are storage devices. This is the same for other host servers.

The CPUs 13a to 13c execute predetermined programs stored in the memory 14a (operate according to instructions in the program), thereby realizing predetermined functions of the virtualization software 16a and the virtual machine A12a.

The virtualization software 16a has the function of the virtual switch 161a. The virtual machine A12a communicates with the other virtual machines B12b to D12d via the virtual switch 161a. The virtual switch 161a receives data from the virtual machine A 12a and transmits it to the physical switch 2 from the port 11a. Similarly, the virtual switch 161a receives data from the physical switch 2 via the port 11a and passes the data to the virtual machine A 12a.

2B and 2C schematically show the configurations of the host server B1b and the host server C1c, which are physical computers, respectively. These basic configurations are the same as those of the host server A1a, and differences from the host server A1a will be mainly described below. As shown in FIG. 2B, the host server B1b includes four CPUs 13d to 13g.

As shown in the memory 14b, the two virtual machines B12b and the virtual server C12c communicate data via the virtual switch 161b of the virtualization software 16b. Data communicated between them is not output from the virtual switch 161b to the port 11b, but is transmitted and received by the host server B1b. The device information 15b includes configuration information and status information of the host server B1b.

As shown in FIG. 2C, the host server C1c includes four CPUs 13h to 13k. As shown in the memory 14c, the virtual machine D12d performs data communication with other virtual machines via the virtual switch 161c of the virtualization software 16c. The device information 15c includes configuration information and status information of the host server C1c.

FIG. 3 schematically shows the configuration of the physical switch 2. The physical switch 2 includes a controller 22, device information 23, and traffic information 24. Although not shown, the physical switch 2 includes a switch circuit in addition to these. The controller 22 sets the data transfer destination by the switch circuit and controls the processing in the physical switch 2.

Controller 22 transmits and receives information in response to a request from an external device. Specifically, in response to a request from the controller 22 and the management server 3, information included in the device information 23 and the traffic information 24 is transmitted. The device information 23 and the traffic information 24 are typically stored in a nonvolatile storage device in the physical switch 2. The device information 23 and traffic information 24 will be described later.

The physical switch 2 in this example is configured by one unit, but the physical switch 2 can be configured by a plurality of units. Each unit has the configuration shown in FIG. For example, the physical switch 2 has two units, and the ports of the units are interconnected by wiring. Data communicated between the ports of different units passes through the ports that interconnect the units.

FIG. 4A is a block diagram schematically showing a hardware configuration of the management server 3 which is a physical computer. 4A, the management server 3 includes a port 31, a CPU 32 that is a processor, a memory 35 that is a main storage device, a secondary storage device 36, a display interface 34, and an input device interface 33. These devices are connected via a bus. The input device interface 33 is connected to an input device 37, and the display interface 34 is connected to a display 38.

The management server 3 is connected to the network 4 via the port 31 and communicates with other devices in the system. The management server 3 may have a plurality of ports. The administrator (user) uses the input device 37 and the display 38 to input and confirm necessary information. The input device 37 is typically a mouse and a keyboard. The management server 3 may include an output device different from the display 38. FIG. 4A illustrates the management server 3 including the input / output device, but the administrator may use the function of the management server 3 from another computer via the network 4.

The CPU 32 implements a predetermined function of the management server 3 by executing a program stored in the memory 35. FIG. 4B shows a program and a table for realizing the function of the management server 3. Specifically, the management server 3 includes a device information acquisition unit 351, a traffic information acquisition unit 352, a virtual machine-host server traffic calculation unit 353, a virtual machine migration traffic calculation unit 354, and a virtual machine migration selection unit 355. . These are programs.

In this example, the function of the management server 3 is indicated by five programs, but the number of programs may be any number. The functions of any two programs may be executed by one program, and the functions of one program may be executed by a plurality of programs. A part of the processing of a certain program may be executed by another program.

The CPU 32 can operate as these functional units by operating according to these programs. Therefore, the processing performed by the program is the processing of the CPU 32 and the management server 3 including the CPU 32. FIG. 4B shows a program in the memory 35 for convenience of explanation, but data (including a program) required for the processing of the management server 3 is typically transferred from the secondary storage device 36 to the memory 35. Stored. The operation of these programs (according to the CPU 32) will be described later. In this example, the processor of the management server 3 is configured by one CPU, but the processor can include a plurality of CPUs.

As shown in FIG. 4B, the management server 3 includes a physical switch prediction traffic table 361, a host server information table 362, a migration destination candidate table 363, a physical switch information table 364, an inter-virtual machine traffic information table 365, a virtual machine-host server. A traffic information table 366 is provided.

Details of each table will be described later. In this configuration example, information used for system operation management is stored in each table. However, the number of tables including necessary information and the information stored in each table depend on the system design. In the present embodiment, the information stored in the data storage area does not depend on the data structure and may be expressed in any data structure. For example, a data structure appropriately selected from a table, list, or database can store information.

FIG. 4B shows the table in the secondary storage device 36 for convenience, but typically the CPU 32 obtains data of these tables loaded in the memory 35. In the management server 3, the program and table (information) are stored in the storage area of each information in the memory 35 and the secondary storage device 36.

The secondary storage device 36 is a non-volatile storage device that stores the program and information necessary for realizing the predetermined function of the management server 3. The secondary storage device 36 is typically a hard disk device. The management server 3 may use a semiconductor storage medium such as a flash memory, or may use a storage device connected via a network. The memory 35, the secondary memory device 36, and the set thereof are storage devices, and may include other elements.

The management server 3 creates the tables 361 to 366 before selecting the destination of the virtual machine. In the following, the tables 361 to 366 and their creation by the management server 3 will be described. FIG. 5 shows a configuration example of the host server information table 362. The host server information table 362 has status information of the host servers A1a to C1c, and in this example, the status information includes hardware configuration information and usage status information.

In the example of FIG. 5, the host server information table 362 stores information on the number of CPUs of each host server, the CPU usage rate, the memory capacity, and the memory usage rate. The device information acquisition unit 351 in the management server 3 acquires information from each of the host servers A1a to C1c.

Specifically, the device information acquisition unit 351 requests the host server A1a to the host server C1c to transmit information registered in the host server information table 362 via the network 4 and the physical switch 2. For example, the device information acquisition unit 351 may transmit a request according to SNMP (Simple Network Management Protocol).

The virtualization software 16a to 16c or the management OS of the host server A1a to the host server C1c transmits the requested information to the management server 3 in response to a request from the device information acquisition unit 351. The device information 15a to 15c of the host server A1a to the host server C1c stores hardware configurations such as the number of CPUs and memory capacity. The host server A1a to the host server C1c may transmit the CPU usage rate and the memory usage rate when receiving the request to the management server 3.

The device information acquisition unit 351 registers information acquired from the host server A1a to the host server C1c in the host server information table 362. The device information acquisition unit 351 may acquire the device information at any timing before the selection of the destination of the virtual machine. Typically, in order to select a more appropriate destination, it is necessary to move the virtual server. Acquire device information after determination.

FIG. 6 shows a configuration example of the physical switch information table 364. The physical switch information table 364 has state information of the physical switch 2, and in this example, includes state information of the hardware configuration and information on its use state. In the example of FIG. 6, the physical switch information table 364 stores information on the bandwidth and usage rate of the main body of the physical switch 2 and the bandwidth and usage rate of each port. Bandwidth is the capacity of data transfer. The table 364 further stores external devices (connection destinations) of communication partners of each port.

The device information acquisition unit 351 in the management server 3 acquires information of the host server information table 362 from the physical switch 2. Specifically, the device information acquisition unit 351 requests the physical switch 2 to transmit information via the network 4. For example, the device information acquisition unit 351 may transmit a request according to SNMP (Simple Network Management Protocol).

In response to a request from the device information acquisition unit 351, the controller 22 of the physical switch 2 transmits the requested information to the management server 3. The controller 22 stores the hardware configuration information of the physical switch 2 and the connection relationship with other devices in the device information 23. In this example, the switch body, the bandwidth of each port, and the connection destination of the port are stored in the device information 23. The controller 22 can obtain this information in the negotiation with the connection destination. In response to a request from the device information acquisition unit 351, the controller 22 acquires bandwidth information from the device information 23 and transmits it.

The controller 22 updates the traffic information 24 periodically or in response to a predetermined event. Typically, the controller 22 stores the total communication amount of the switch body and each port in the traffic information 24. The total communication amount is, for example, a cumulative value of traffic (for example, the number of input / output packets) since the previous reset. The controller 22 transmits these values in response to a request from the device information acquisition unit 351. The device information acquisition unit 351 can periodically acquire this traffic information and calculate the bandwidth usage amount by dividing the traffic increase amount by time.

The device information acquisition unit 351 registers the bandwidth information acquired from the physical switch 22 and the calculated bandwidth usage information in the physical switch information table 364. Note that the controller of the physical switch 22 may calculate the bandwidth usage amount and transmit it to the device information acquisition unit 351. In the configuration, the device information acquisition unit 351 typically acquires device information and traffic information after determining the movement of the virtual server.

In the above configuration, the device information acquisition unit 351 automatically acquires necessary information from the host server A1a to the host server C1c and the physical switch 2. Unlike this, the administrator may obtain necessary information from the host server A1a to the host server C1c and the physical switch 2 by operating the management server 3 (or other computer). The administrator registers the acquired information in the host server information table 362 and the physical switch information table 364.

The information acquired by the device information acquisition unit 351 (information registered in the host server information table 362 and the physical switch information table 364) is preferably the above information, but may include information different from this. For example, the device information acquisition unit 351 may register the number of power supplies in the host server information or register the throughput in the physical switch information.

Next, the inter-virtual machine traffic information table 365 will be described. FIG. 7 shows a configuration example of the inter-virtual machine traffic information table 365. The inter-virtual machine traffic information table 365 indicates data communication traffic (band usage in this example) between each of the virtual machines A12a to D12d operating in the system. In this example, the host server of each transmission source virtual machine and the host server of the transmission destination virtual machine are also included in the inter-virtual machine traffic information table 365.

As shown in FIG. 7, this table 365 stores the bandwidth usage (traffic) of each communication direction between two virtual machines for each pair of the inter-virtual machine traffic information table 365. For example, the transmission amount (bandwidth usage) from the virtual machine C12c to the virtual machine D12d between the virtual machine A12a and the virtual machine B12b is 20 Mbps, and the transmission amount (bandwidth usage) from the virtual machine D12d to the virtual machine C12c. ) Is 30 Mbps.

In FIG. 7, the bandwidth usage between the virtual machine B 12b and the virtual machine C 12c is blank in both directions. Both the virtual machine B12b and the virtual machine C12c are operating on the host server B1b. As described above, the inter-virtual machine traffic information table 365 stores the bidirectional bandwidth usage between virtual machines operating on different host servers.

In this configuration example, the traffic information acquisition unit 352 (see FIG. 4B) creates a virtual machine traffic information table 365 and writes information in the table. The traffic information acquisition unit 352 updates the inter-virtual machine traffic information table 365 periodically or in response to a predetermined event. Typically, the creation or update of the inter-virtual machine traffic information table 365 is matched with the physical switch information table 364.

FIG. 8 is a flowchart showing a flow of information registration processing in the inter-virtual machine traffic information table 365 by the traffic information acquisition unit 352. The traffic information acquisition unit 352 executes Step 101 to Step 104 for all combinations of the two virtual machines.

The traffic information acquisition unit 352 first determines whether or not the virtual machine M and the virtual machine N exist on one host server (S101). Virtual machine M and virtual machine N are any two selected from virtual machine A to virtual machine D. When two virtual machines are operating on one host server (Y in S101), the traffic information acquisition unit 352 ends the process for the pair. In this example, the virtual machine B12b and the virtual machine C12c correspond to this.

When the two virtual machines are operating on different host servers (N in S101), the traffic information acquisition unit 352 acquires transmission traffic from the virtual machine M to the virtual machine N (S102), and further the virtual machine The transmission traffic from N to the virtual machine M is acquired (S103). The traffic information acquisition unit 352 writes the acquired two traffics in the inter-virtual machine traffic information table 365.

The traffic information acquisition unit 352 acquires traffic information from the physical switch 2 in accordance with, for example, the NetFlow protocol. The traffic information stored in the controller 22 of the physical switch 2 includes traffic information between virtual machines (transmission traffic from the virtual machine M to the virtual machine N and transmission traffic from the virtual machine N to the virtual machine M).

The traffic information acquisition unit 352 can acquire traffic information between virtual machines operating on different host servers from the physical switch 2 to create and update the inter-virtual machine traffic information table 365 shown in FIG. Traffic information between virtual machines (between virtual machine B 12b and virtual machine C 12c) on one host server in the table cannot be acquired from the physical switch 2. The specification of the traffic between these will be described later.

The information acquired by the traffic information acquisition unit is not limited to that of this embodiment as long as it corresponds to the physical switch information table 364. In this embodiment, since the bandwidth is registered in the physical switch information table, the bandwidth usage is acquired as the traffic information. For example, when the throughput is registered as the physical switch information, the traffic is Get the number of input / output packets as information.

As will be described later, the selection of the destination host server for the virtual machine uses traffic information between the virtual machine and each host server. Therefore, the inter-virtual machine traffic information table 365 may include information that directly indicates bi-directional transmission traffic between each virtual machine and each host server.

In this example, the traffic information acquisition unit 352 acquires traffic information between virtual machines from the physical switch 2, but may acquire traffic information from a different component. For example, according to the NetFlow protocol, the traffic information acquisition unit 352 may acquire traffic information from the virtualization software 16a to 16c.

The example of the inter-virtual machine traffic information table 365 has bi-directional traffic between two virtual machines, but may have bi-directional total traffic. In the traffic information acquisition example from the physical switch 2, it is assumed that traffic information between virtual machines (between the virtual machine B 12b and the virtual machine C 12c) on one host server cannot be directly acquired. When the virtualization software 16b has this information, the traffic information acquisition unit 352 may acquire it from the virtualization software 16b or the management OS.

Next, the virtual machine-host server traffic information table 366 will be described. The virtual machine host server traffic calculator 353 (see FIG. 4B) creates and updates this table 366. FIG. 9 shows an example of the configuration of the virtual machine-host server traffic information table 366. This table has traffic (band usage in this example) between each virtual machine and each host server.

As shown in FIG. 9, the total bidirectional traffic in the communication between the virtual machine and the host server is shown. The total traffic is the sum of the transmission traffic from the virtual machine to the host server and the transmission traffic from the host server to the virtual machine.

The virtual machine-host server traffic information table 366 includes a bandwidth usage (traffic) between each virtual machine (for example, the virtual machine A11a) and an external host server (for example, the host server B1b and the host server C1c), It has information on bandwidth usage (traffic) between each virtual machine and the host server (for example, host server A1a) where the virtual server exists.

The virtual machine performs data communication with a virtual machine operating on another host server, and also performs virtual network communication with another virtual machine operating on the same host server. Communication on one host server appears to the two virtual machines as network communication, but in reality, data is transmitted and received within the same host server (without going through the network).

When one of a plurality of virtual machines operating on one host server is moved to another host server, new network communication occurs between the source host server and the destination host server. . This causes a significant increase in network traffic and has a significant impact on system network traffic.

On the other hand, the communication on the network between the destination host server and the virtual machine before the movement changes to transmission / reception of data on the host server due to the movement of the virtual machine. Since there is no network traffic due to this communication, a large change occurs in the network traffic.

In addition to the traffic between the virtual machine and the external host server, consider the traffic between the virtual machine and the host server on which it resides, so that the traffic after moving the virtual machine to another host server is accurate. Can be predicted. Depending on the design, the traffic after movement may be predicted only from the traffic between the virtual machine and the external host server. However, for accurate traffic prediction, this example is used between the virtual machine and the current host server. Identify traffic and use it to predict post-movement traffic.

In this preferred configuration, the traffic between the moving virtual machine C12c and each of the host server A1a to the host server C1c is measured and calculated, and the traffic after moving the virtual server C12c is predicted from the traffic. By identifying the current traffic between the moving virtual machine and each host server, it is possible to accurately predict the traffic after moving the virtual machine. Depending on the design, the traffic after moving the virtual machine may be predicted from the traffic information specifying only the host server as the subject of transmission and reception.

The virtual machine host server traffic calculator 353 calculates traffic on the same host server (data amount transmitted and received) from the acquired information. The processing flow of the virtual machine host server traffic calculation unit 353 will be described with reference to the flowchart of FIG. The virtual machine host server traffic calculation unit 353 executes Step 201 to Step 207 for all combinations of virtual machines and host servers. In this example, there are 12 combinations between 4 virtual machines and 3 host servers.

The virtual machine host server traffic calculator 353 first determines whether or not the virtual machine host server and the communication partner host server are the same (S201). If they are different host servers (N in S201), steps 202 to 204 are executed. For example, in the example of the virtual machine A12a, steps 202 to 204 are executed for the communication between the virtual machine A12a and the host server B1b and the communication between the virtual machine A12a and the host server C1c.

The virtual machine host server traffic calculation unit 353 acquires transmission traffic from the virtual machine to the host server virtual machine in the current combination (S202). When a plurality of virtual machines are operating on the host server, the transmission traffic is the total transmission traffic to the plurality of virtual machines. With reference to the inter-virtual machine traffic information table 365, this transmission traffic can be calculated.

Specifically, the inter-virtual machine host server traffic calculation unit 353 determines all (one or more) virtual machines from the target virtual machine (for example, the virtual machine A12a) to which the target host server (for example, the host server B1b) exists. The transmission traffic to each of the machines (for example, virtual machine B12b and virtual machine C12c) is acquired. Furthermore, the sum total of those transmission traffic is calculated. This value is the transmission traffic from the virtual machine to the virtual machine of the host server.

The virtual machine host server traffic calculation unit 353 acquires transmission traffic from the virtual machine on the host server to the virtual machine in the current combination (S203). Similar to step 202, the transmission traffic can be calculated by referring to the inter-virtual machine traffic information table 365.

Specifically, the traffic calculation unit 353 between the virtual machine host servers from all (one or more) virtual machines (for example, the virtual machine B12b and the virtual machine C12c) in which the target host server (for example, the host server B1b) exists. Each transmission traffic to the target virtual machine (for example, virtual machine A12a) is acquired. Furthermore, the sum total of those transmission traffic is calculated. This value is the transmission traffic from the virtual machine existing in the target host server to the target virtual machine.

The virtual machine host server traffic calculation unit 353 calculates the sum of the two calculated transmission traffics, and registers it as the traffic between the virtual machine and the host server in the virtual machine-host server traffic information table 366 (S204). ).

Returning to step 201, when the host server of the virtual machine and the host server of the communication partner are the same (Y in S201), the virtual machine host server traffic calculation unit 353 executes step 205 to step 207. For example, in the example of the virtual machine A12a, Steps 205 to 207 are executed for communication between the virtual machine A12a and the host server A1a.

The virtual machine host server traffic calculation unit 353 acquires traffic (overall traffic) transmitted and received by the target virtual machine (S205). For example, the virtual machine host server traffic calculation unit 353 can acquire the information from the virtual machine according to SNMP. If the virtualization software has that information, it may obtain that information from it.

The virtual machine host server traffic calculation unit 353 acquires traffic (external traffic) with the target virtual machine transmitted and received by the physical switch 2 (S206). This traffic can be calculated with reference to the inter-virtual machine traffic information table 365. The inter-virtual machine traffic information table 365 has traffic between the virtual machine and an external host server (a host server other than the host server on which the virtual machine exists), and the sum of these traffic is the external traffic. is there.

The virtual machine host server traffic calculation unit 353 then calculates a value obtained by subtracting external traffic from the total traffic. Further, the calculated value is registered in the virtual machine-host server traffic information table 366 as traffic between the virtual machine and the host server on which the virtual machine is operating (S207).

In this example, only the host server B1b has traffic on one host server. That is, the traffic between the virtual machine and the host server on which the virtual machine is operating is only the traffic due to the communication between the virtual machine B12b and the virtual machine C1c.

As shown in FIG. 9, the bandwidth usage between the virtual machine A12a and the host server A1a is 0 Mbps, and the bandwidth usage between the virtual machine D12d and the host server C1ac is also 0 Mbps. On the other hand, the bandwidth usage between the virtual machines B12b and C12c and the host server B1b is 300 Mbps.

This processing example assumes that traffic information between virtual machines on one host server cannot be obtained directly. Therefore, calculation of traffic between the virtual machine and the host server uses different methods for communication between the virtual machine and an external host server and communication between the virtual machine and the host server on which the virtual machine exists. As described above, when the virtualization software has traffic information in the host server, the calculation method in steps 202 to 204 may be applied to all cases.

Next, the physical switch prediction traffic table 361 will be described. FIG. 11 shows a configuration example of the physical switch prediction traffic table 361. The traffic calculation unit 354 after moving the virtual machine (see FIG. 4B) creates this table 361. The physical switch prediction traffic table 361 has a traffic prediction value after moving the virtual machine.

11, the physical switch predicted traffic table 361 indicates the predicted traffic after the movement of the virtual machine C12c, as shown in FIG. Specifically, the column “Move to Host Server A” stores predicted values of bandwidth usage in the physical switch main body and ports A21a to E21e when the virtual machine C12c is moved to the host server A1a. . Similarly, the column “Move to Host Server C” stores predicted values of bandwidth usage in the physical switch body and ports A21a to E21e when the virtual machine C12c is moved to the host server C1c.

The after-virtual machine movement traffic calculation unit 354 can create the physical switch prediction traffic table 361 by referring to the physical switch information table 364 and the virtual machine-host server traffic information table 366. FIG. 12 is a flowchart illustrating a flow of processing (creation of the predicted traffic table 361) in which the traffic calculation unit 354 after moving the virtual machine calculates the load of the physical switch 2 after moving the virtual machine.

Before the description of the flowchart of FIG. 12, the traffic change accompanying the movement of the virtual machine C1c will be described with reference to FIGS. FIG. 13 shows current traffic before moving the virtual machine C1c. Traffic of 400 Mbps, 300 Mbps, and 500 Mbps exists between the virtual machine C1c, the virtual machine A1a, the virtual machine B1b, and the virtual machine D1d, respectively.

FIG. 14 shows predicted traffic when the virtual machine C1c is moved from the host server B1b to the host server A1a. FIG. 15 shows predicted traffic when the virtual machine C1c is moved from the host server B1b to the host server C1c. Each of the traffic between the virtual machines and each of the traffic between the virtual machines and the host server are not changed by the movement of the virtual machine C1c.

On the other hand, the traffic between the host server A1a to the host server C1c and the physical switch 2 (traffic between host servers) changes due to the movement of the virtual machine C1c. The host servers A1a to C1c communicate with external devices (physical switch 2 and other host servers) via ports 11a to 11c. Similarly, the physical switch 2 performs data communication via the port A21a to the port E21e.

In traffic control, traffic at ports (band usage in this example) is an important factor. In the physical switch 2, the traffic of the main body is also an important factor. If the bandwidth usage is large with respect to the bandwidth of the port (switch body), the communication speed becomes slow. Therefore, it is preferable that the bandwidth usage in each element is within a specified range.

In a preferred configuration, the traffic in the physical switch 2 is referred to when selecting the destination host server of the virtual machine. Thereby, appropriate measurement and prediction of the communication network of the host server A1a to the host server C1c can be performed efficiently. In the following, as a preferable configuration, a configuration that refers to the traffic of the main body of the physical switch 2 and the traffic of each port will be described. Selection of the destination host server may refer to other traffic information, for example, traffic at the port of the host server.

Typically, the bandwidth of the pair of corresponding ports of the physical switch 2 and the host server A1a to the host server C1c is set to a small bandwidth in the port pair by negotiation between them. In this way, since the bandwidths of the corresponding ports are set to the same value, appropriate traffic on the network can be realized after the movement by using the traffic of the physical switch 2 for selecting the movement destination.

The creation of the predicted traffic table 361 by the traffic calculation unit 354 after moving the virtual machine will be described with reference to FIG. The post-virtual machine movement traffic calculation unit 354 repeats Step 301 to Step 306 for all entries in the physical switch information table 364 for each of all the destination host server candidates.

In this example, the after-virtual machine movement traffic calculation unit 354 executes steps 301 to 306 for each of all entries in the physical switch information table 364 when the virtual machine C12c is moved to the host server A1a. . Further, Step 301 to Step 306 are executed for each of all entries in the physical switch information table 364 when the virtual machine C12c is moved to the host server C1c.

In step 301 of FIG. 12, the after-virtual machine movement traffic calculation unit 354 identifies the part of the target entry (traffic calculation target). When the target entry is a switch body, the after-virtual machine movement traffic calculation unit 354 calculates the predicted load (traffic) of the switch body according to the following equation.
Current bandwidth used-(V-Hd traffic) + (V-Hc traffic)

In this equation, V is a virtual machine, Hd is a migration destination candidate host server, and Hc is a current host server of the virtual machine. This formula subtracts the traffic between the target virtual machine (virtual machine C12c in this example) and the destination host server (host server A1a or host server C1c) from the current bandwidth used by the switch body. Furthermore, it shows that the traffic between the target virtual machine and the current host server (host server B1b in this example) is added.

When the target entry is a port (the port in S301), the after-virtual machine movement traffic calculation unit 354 specifies the connection destination of the port (S303). The connection destination is any one of the migration destination candidate host server Hd, the current host server Hc of the moving virtual machine, or another host server Ho. The traffic calculation unit 354 after moving the virtual machine can acquire this information from the physical switch information table 364.

When the connection destination of the target port is another host server Ho (Ho in S303), the after-virtual machine movement traffic calculation unit 354 registers the current bandwidth usage in the physical switch information table 364 in the predicted traffic table 361. (This step is not shown).

When the connection destination of the target port is the current host server Hc (Hc in S303), the after-virtual machine movement traffic calculation unit 354 calculates the predicted load (traffic) of the port according to the following equation (S304).
Current bandwidth used-(V-Hd traffic)
+ (Traffic between V-Hc)-(traffic between V-Ho)

This formula subtracts the traffic between the target virtual machine (virtual machine C12c in this example) and the migration destination candidate host server (one of the host server A1a and the host server C1c) from the current used bandwidth of the port. The traffic between the target virtual machine and the current host server (host server B1b in this example) is added, and the target virtual machine and other host servers (the other of the host server A1a and the host server C1c) Shows that the traffic between is subtracted.

When the connection destination of the target port is the migration destination candidate host server Hd (Hd in S303), the post-virtual machine movement traffic calculation unit 354 calculates the predicted load (traffic) of the port according to the following equation (S305). ).
Current bandwidth used-(V-Hd traffic)
+ (Traffic between V-Hc) + (traffic between V-Ho)

This formula subtracts the traffic between the target virtual machine (in this example, virtual machine C12c) and the destination host server (one of host server A1A or host server C1C) from the current bandwidth used by the port. The traffic between the target virtual machine and the current host server (host server B1B in this example) is added, and the target virtual machine and other host servers (the other of the host server A1A and the host server C1C) Shows that the traffic between is added.

The after-virtual machine movement traffic calculation unit 354 registers the value calculated in step 304 and step 305 in the predicted traffic table 361 as the traffic predicted value of the port of the entry (S306).

The selection of the destination host server of the virtual machine refers to the predicted traffic table 361. In this example, the virtual machine destination selection unit 355 (see FIG. 4B) selects a destination host server for the virtual machine. In a preferred configuration, the virtual machine destination selection unit 355 determines the destination by referring to the status information of the destination host server and the expected traffic after moving the virtual machine.

This makes it possible to select a more appropriate host server from the perspective of both the host server alone and traffic. Depending on the design, the destination host server may be determined without referring to the status information of the destination host server. In the following, processing for selecting a destination from both viewpoints will be described.

FIG. 16 shows a configuration example of the destination candidate table 363. The virtual machine migration destination selection unit 355 creates the table 363. The virtual machine migration destination selection unit 355 refers to the host server information table 362, acquires the CPU usage rate and memory usage rate of the migration destination candidate host server, and registers these values in the migration destination candidate table 363.

In this example, the host server usage status stores the usage status of the host server before moving the virtual machine. For example, in the migration destination candidate table 363 in FIG. 16, the CPU usage rate of the host server C1c is 25%, and the memory usage rate is 30%. Since these values are current values, they do not depend on the migration destination candidate of the virtual machine C12. The virtual machine destination selection unit 355 may calculate a predicted value of the usage state (CPU and memory usage rate) after moving the virtual machine C12c, and register the predicted value in the destination candidate table 363.

The virtual machine destination selection unit 355 further registers the predicted traffic at the two ends of the physical switch and its ports. FIG. 16 shows only the ports A21a to C21c related to the host servers A1a to C1c. The virtual machine migration destination selection unit 355 can acquire the physical switch body and the bandwidth of each port from the physical switch information table 364. Further, the predicted traffic (predicted usage value) can be acquired from the physical switch predicted traffic table 361.

The virtual machine destination selection unit 355 determines the priority of the destination candidate based on the status information of the destination candidate host server and the predicted traffic. In the example of FIG. 16, the priority of the host server A1a is HIGH, and the priority of the host server C1c is LOW. The virtual machine destination selection unit 355 selects the host server A1a as the destination host server for the virtual machine C12c.

For example, the virtual machine migration destination selection unit 355 first determines the priority of each host server according to the usage state of the host server. Since the CPU usage rate and the memory usage rate of the host server C1c are small, the provisional priority is determined to be HIGH. Since the CPU usage rate and the memory usage rate of the host server A1a are slightly high, the provisional priority is determined to MIDDLE.

The virtual machine destination selection unit 355 further determines the final priority with reference to the predicted traffic. When the virtual machine C12c moves to the host server A1a, the bandwidth usage of the switch body, the port A21a, and the port B21b decreases before the movement (see FIGS. 6 and 11). Therefore, the virtual machine migration destination selection unit 355 increases the priority of the host server A1a from MIDDLE to HIGH.

When the virtual machine C12c moves to the host server C1c, the predicted bandwidth usage of the switch body and the port C21c exceeds the bandwidth. Therefore, the virtual machine migration destination selection unit 355 lowers the priority of the host server C1c from HIGH to LOW. By determining the priority of the destination candidate according to the status information of the host server and then correcting the priority according to the predicted traffic, it is possible to select an appropriate destination from the viewpoint of both the host server and the traffic. Note that the virtual machine migration destination selection unit 355 may correct the priority determined according to the predicted traffic based on information of the host server.

As described above, when the predicted traffic includes a bandwidth usage exceeding the bandwidth in the physical switch 2 (main unit or port), the virtual machine destination selection unit 355 sets the priority of the destination candidate to the lowest or It is preferable to exclude the host server from the destination candidates. Thereby, it is possible to avoid the occurrence of a communication error and a significant decrease in the communication speed due to the reason. The virtual machine destination selection unit 355 can use a prescribed threshold value calculated from the bandwidth as a value to be compared with the bandwidth usage. This threshold value may be a value of a prescribed ratio (for example, 90%) of the bandwidth in addition to the bandwidth.

Priority can be determined using various methods according to the design. For example, the virtual machine migration destination selection unit 355 determines the first priority value of each migration destination candidate host server according to the usage status of the migration destination candidate host server. Further, the second priority value of the destination candidate host server is determined according to the predicted traffic.

The virtual machine migration destination selection unit 355 determines the migration candidate having the second highest priority value as the migration destination host server in the migration destination candidate host server having the highest first priority value. . Typically, the state of a destination host server is more important than traffic in selecting a destination candidate. Therefore, it is preferable to use the status information of the host server in preference to the predicted traffic.

As in the above example, if the predicted traffic contains a bandwidth usage that exceeds any bandwidth (threshold value calculated from the bandwidth) in the physical switch 2 (main unit or port), the mobile operation candidate is excluded. Is preferred. Alternatively, as in the above example, the virtual machine migration destination selection unit 355 may lower the value of the first priority of the migration destination candidate including the bandwidth usage exceeding the threshold by the specified level.

For example, the first priority value is determined by preliminarily defining a plurality of CPU usage rates and memory usage rates, and assigning a first priority value to each category. For example, when the first priority can take three values, 0% to 100% are divided into three sections, and one of HIGH, MIDDLE, and LOW is assigned to each section. The virtual machine migration destination selection unit 355 determines the lower priority as the first priority value of the migration candidate in the first priority values of the CPU usage rate and the memory usage rate of the migration destination candidate. .

The determination of the second priority value is performed by, for example, calculating the variation in the predicted bandwidth usage rate (predicted bandwidth usage / bandwidth) in the physical switch 2 and determining the value according to the variation value. A plurality of categories of variation values are defined in advance, and a second priority value is assigned to each category. The virtual machine destination selection unit 355 determines the value of the category including the predicted bandwidth usage rate as the second priority value of the destination candidate.

The virtual machine migration destination selection unit 355 may use other values instead of or in addition to the above-mentioned usage rate as the status information of the host server referred to in the priority determination. For example, the number of CPUs, the number of power supplies, or the failure rate of the host server may be used.

The virtual machine destination selection unit 355 may use the maximum value of the predicted bandwidth usage rate as the value of the predicted traffic to be referred to in the destination selection. The virtual machine migration destination selection unit 355 divides the range in which the predicted bandwidth usage rate can be divided into a plurality of sections, and assigns a second priority value to each section. The virtual machine destination selection unit 355 determines that the value of the section including the maximum predicted bandwidth usage rate of each destination candidate is the second priority value of the destination candidate.

The bandwidth usage rate used in the above example is a time average of bandwidth usage, but the virtual machine migration destination selection unit 355 may use the peak bandwidth usage rate as traffic information to be referenced in the migration destination selection. The priority can be determined based on the peak bandwidth usage rate by using the same method as the bandwidth usage rate.

As described above, the virtual machine migration destination selection unit 355 selects the migration destination candidate based on the set capacity (bandwidth in the above example) in addition to the traffic (band usage in the above example) of the main body and the port of the physical switch 2. It is preferable to determine the priority. As a result, it is possible to select a destination that can realize appropriate traffic for the specifications of the physical switch 2. Depending on the design, the priority of the destination candidate may be determined only from the traffic of the physical switch 2. For example, when the specification of the physical switch 2 is known, the virtual machine migration destination selection unit 355 may determine the priority using the traffic and the setting value.

In each of the above examples, the virtual machine destination selection unit 355 selects the final destination host server. Unlike this, the virtual machine move destination selection unit 355 does not automatically determine the move destination of the virtual machine, displays information necessary for the move destination selection on the display 38 to the administrator, and supports the move destination selection. Also good.

For example, the virtual machine destination selection unit 355 outputs the destination candidate table 363 shown in FIG. The administrator uses the input device 37 to input the identifier of the selected destination host server according to the displayed destination candidate table 363. The virtual machine destination selection unit 355 selects the host server indicated by the user input as the destination host server.

Preferably, as shown in FIG. 16, in addition to the priority of each destination candidate host server, the display image includes information used for determining the priority, in this example, the status information of the host server, the bandwidth of the physical switch Contains information including width (capacity) and predicted traffic. This provides more information useful to the administrator for destination selection.

Depending on the design, the display image may display only the identification information of each destination candidate host server and its priority. The image for selecting the destination host server may be displayed by any other method besides directly displaying the priority by a code (numerical value or character). For example, the host servers may be arranged in order from the top to the bottom in order of priority, or the color of the identification information of the host server may be changed according to the priority.

For example, if the priority can be two values, LOW and HIGH, the priority of the destination candidate when the predicted traffic exceeds the capacity of any port is LOW, and the priority of the other destination candidates is HIGH Suppose that The management server 3 displays the move destination candidates with the priority LOW in red, and displays the other move destination candidates in black. Thereby, the priority of each movement destination candidate can be shown.

Although the present invention has been described in detail with reference to the accompanying drawings, the present invention is not limited to such specific configurations, and various modifications and equivalents within the spirit of the appended claims Includes configuration. For example, at least a part of the processing of the management server may be executed by another physical computer including the host server. At least a part of the program may be realized by dedicated hardware. The program can be installed in each computer by a program distribution server or a computer-readable non-transitory storage medium, and can be stored in a nonvolatile storage device of each computer.

In the above configuration, all physical computers have a virtual environment, but the system can include a physical computer that does not have a virtual environment. The management server selects a physical computer having a virtual environment from all the physical computers as a movement destination candidate, and predicts traffic when the virtual machine is moved to each of the selected physical computers. The predicted traffic includes traffic between a virtual machine and a physical computer that does not have a virtual environment.

The management server may select a part as a destination candidate from a host server having a virtual environment. The management server predicts traffic when the virtual machine is moved to each of the migration destination candidate host servers, and determines the priority of each migration destination candidate using the prediction result. For example, the management server excludes host servers that are not appropriate to be included in the destination candidates from the status information of the host server.

The present invention can be used in a computer system that moves a virtual machine between physical computers via a network.

Claims

In a computer system including a plurality of physical computers communicating via a network, a method for managing virtual machines,
Measuring traffic between the plurality of physical computers in the network;
Predicting traffic of the network when moving a virtual machine existing in one of the plurality of physical computers to each of destination candidates in the plurality of physical computers from the measurement results;
Determining a priority of the destination candidate based on the predicted traffic;
A method involving that.
The prediction calculates traffic between the virtual machine and each of the plurality of physical computers from the measurement result of the traffic, and predicts the traffic of the network from the calculated traffic.
The method of claim 1.
In the calculation of the traffic between each of the plurality of physical computers, the traffic between the virtual machine and the physical computer in which the virtual machine exists is calculated from the total traffic of the virtual machine and the virtual machine. Subtract external traffic to and from external physical computers,
The method of claim 2.
The plurality of physical computers communicate with each other via a physical switch included in the network,
The prediction predicts traffic in each of the plurality of ports of the physical switch corresponding to each port of the plurality of physical computers,
The priority determination determines the priority of the destination candidate based on predicted traffic in each of the plurality of ports.
The method of claim 2.
The priority determination determines the priority of the destination candidate based on the capacity and the predicted traffic in each of the plurality of ports.
The method of claim 4.
In the priority determination, when the predicted traffic of the movement destination candidate exceeds a threshold determined from the capacity of the one port in at least one port of the plurality of ports, the priority of the movement destination candidate is lowered. ,
The method of claim 5.
Determining the priority further includes determining the priority of the destination candidate based on capacity and predicted traffic in the physical switch body;
The method of claim 5.
Displaying information indicating the priority on a display device;
According to the user input for the display, the destination physical computer of the virtual machine is selected from the destination candidates.
The method of claim 2 further comprising:
Selecting a physical computer having the highest priority among the migration destination candidates as a migration destination physical computer of the virtual machine;
The method of claim 2 further comprising:
The priority determination determines the priority of the destination candidate based on the predicted traffic and the status information of each destination candidate,
The method of claim 2.
Network,
A plurality of physical computers communicating with each other via the network;
A traffic measurement result acquisition unit for acquiring a measurement result of traffic between the plurality of physical computers in the network;
A traffic measurement information storage area for storing information representing the acquired measurement result;
A traffic prediction unit that predicts traffic of the network when moving a virtual machine existing in one of the plurality of physical computers to each of destination candidates in the plurality of physical computers from information in the traffic measurement area; ,
A predicted traffic information storage area for storing information representing the predicted traffic;
A priority determination unit that determines the priority of the destination candidate based on information in the predicted traffic information storage area;
Including computer system.
The traffic prediction unit calculates traffic between the virtual machine and the plurality of physical computers from the traffic measurement result, and predicts traffic of the network from the calculated traffic.
The computer system according to claim 11.
In a computer system including a plurality of physical computers communicating via a network and a management system for managing virtual machines of the physical computers, a program for causing the management system to execute processing for moving virtual machines is stored. A temporary computer-readable storage medium, wherein the processing includes:
Measuring traffic between the plurality of physical computers in the network;
Predicting traffic of the network when moving a virtual machine existing in one of the plurality of physical computers to each of destination candidates in the plurality of physical computers from the measured traffic;
Determining a priority of the destination candidate based on the predicted traffic;
A non-transitory computer-readable storage medium.
In the prediction, traffic between the virtual machine and the plurality of physical computers is calculated from the measurement result of the traffic, and traffic of the network is predicted from the calculated traffic.
The non-transitory computer readable storage medium according to claim 13.