JP2016040876A - Arrangement determination device for virtual machine, method thereof and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To determine an optimal VM migration destination or addition destination with computational complexity also applicable to a carrier network while considering information of a service chain and a location of a user terminal.SOLUTION: A pattern of the service chain and a closest DC of the user terminal that utilizes the pattern are extracted from a service chain information table 21 and a user location table 22 by a pattern detection function part 14. The number of times of appearance for each pattern of the service chain and for each closest DC of the user terminal that utilizes the pattern is described in a pattern table 15. A high frequency pattern is extracted from the pattern table 15 by a VM arrangement determination function part 13. The number of VM providing a NW function included in the high frequency pattern is acquired from a NW function-VM number information table 23. The migration destination of the VM for each NW function included in the high frequency pattern is determined in accordance with a ratio of the number of times of appearance for each closest DC of the high frequency pattern.SELECTED DRAWING: Figure 7

Description

  The present invention relates to a network virtualization technique for determining a migration destination or addition destination of a virtual machine, particularly information on a service chain in an environment where a communication path passes through a plurality of virtual machines arranged on an arbitrary data center. The present invention also relates to a technique for determining a migration destination or addition destination of a virtual machine in consideration of the location of a user terminal.

  In recent years, NFV (Network Function Virtualization) / SDN (Software Defined Network) technologies have attracted attention as network (NW) virtualization technologies. The NFV / SDN technology is expected to reduce equipment costs, optimize resources, automate operations and improve efficiency, and can greatly reduce CAPEX (capital investment) / OPEX (operation costs). Furthermore, the NW can be controlled more flexibly by software, and it is expected to realize an NW service that has not been possible until now.

  As such a new NW control method, a service (function) chaining technique is attracting attention. The service chaining technology passes through a plurality of NW functions arranged as virtual machines (VMs) on a data center (DC) in traffic (packets) in units of user terminals (or their applications) by virtualization. As shown in FIG. 1, even if a plurality of VMs as a plurality of NW functions required by a user terminal are arranged on separate DCs (inner servers), as shown in FIG. It becomes possible to provide the NW function to the user terminal (or its application).

  In a network with tens of thousands or hundreds of thousands of subscribers (users) such as a carrier network, capacity and reliability are provided when providing network functions virtualized as virtual machines to user terminals. From this point of view, a form in which VMs are arranged not on one DC but on a plurality of DCs is conceivable. In this case, depending on the information of the service chain, the location of the user terminal, and the arrangement of the VM, as shown on the left side of FIG. It may be formed. Note that the service chain information refers to data of the order of NW functions that pass through, and the location of the user terminal refers to the DC with the shortest communication distance with the user terminal.

  As a method for solving this problem, as shown on the right side of FIG. 2, a method of migrating or adding a VM to a suitable DC based on the service chain information and the location of the user terminal is considered.

  On the other hand, as a method for selecting a server on which a VM is arranged in an environment where a plurality of servers on which a VM can be arranged exists, there is a method for determining a migration destination according to the resource status of the server as in Patent Document 1. However, such a method cannot solve the above-mentioned problem, and a method for arranging a virtual machine on a suitable server based on service chain information and a user terminal location has not been studied.

  Therefore, a method of migrating or adding a VM to a suitable DC based on service chain information and the location of a user terminal is proposed to solve the above problem.

  As a method of arranging a VM in a suitable DC based on service chain information and the location of a user terminal, it can be easily considered that all VM placement patterns are searched and optimal VM placement is performed. In an environment where tens of thousands or hundreds of thousands of users are subscribed and each user has service chain information, as shown in FIG. 3, the amount of calculation processing for calculating the optimum arrangement In addition, there is a possibility that the optimal VM arrangement in one user terminal may not be suitable for other user terminals, and it is almost impossible to calculate the optimal arrangement for all user terminals. .

  Although a method of extracting a high-frequency pattern of the service chain and performing optimal VM placement is also conceivable, as shown in FIG. 4, a route that makes a round trip to the WAN can be reduced, but a long-distance DC from the user terminal can be reduced. In this case, traffic from many user terminals passes through the WAN.

  For this reason, it is necessary to consider an arrangement method in which the calculation amount can be scaled even in the carrier network, and the communication path as a whole is as close to the optimum VM arrangement as possible.

  In the present invention, as shown in FIGS. 5 and 6, a table storing service chain information (hereinafter referred to as a service chain information table) and a table storing user terminal locations (hereinafter referred to as a user location table). The service chain pattern and the location of the user terminal that uses the service chain of that pattern are extracted, and for only the NW function corresponding to the high-frequency pattern, these functions exist in the same DC close to the user terminal. When a migration destination DC is selected as described above and a VM for a certain NW function needs to be added, the service chain pattern related to the NW function is similarly extracted. Select the DC to which the VM is added so that exists in the same DC close to the user terminal. To.

  By performing migration or addition of VMs only for NW functions that fall under the high-frequency pattern, the computational load is greatly reduced, ensuring scalability applicable to the carrier network, and effective for many user terminals. Realize high VM placement.

  According to the present invention, an inefficient communication path that makes a round trip to a DC that is physically separated by a service chain is reduced, so that the network provider can reduce the bandwidth used by the network provider, and the network is based on the service chain. For a user who performs communication, communication delay is reduced. Furthermore, even in a network having a large number of user terminals and service chain information such as a carrier network, it is possible to implement a VM placement method based on the above-described service chain.

Configuration diagram showing an overview of network (NW) virtualization technology Explanatory diagram showing an example of communication paths in service chaining technology Explanatory drawing which shows the subject of the method of searching all the arrangement patterns of VM and performing optimal VM arrangement | positioning Explanatory drawing which shows the subject of the method of extracting the high frequency pattern of a service chain and performing optimal VM arrangement | positioning Explanatory drawing showing the outline of the present invention system Explanatory drawing which shows the outline | summary of the method of performing optimal VM arrangement | positioning in consideration of both the service chain and user's location by this invention The block diagram which shows an example of the arrangement | positioning determination apparatus of the virtual machine of this invention Explanatory drawing which shows an example of each table which a data management server stores Explanatory drawing of user location table generation Explanatory drawing which shows the outline | summary of arrangement | positioning determination of the virtual machine by this invention Explanatory drawing which shows the outline | summary of arrangement | positioning determination of the virtual machine by this invention VM migration process flowchart Flow chart of the VM migration destination selection process in FIG. VM migration processing sequence diagram Explanatory drawing which shows the flow of VM migration processing VM addition processing flowchart Flowchart of VM addition destination selection process 1 in FIG. Flowchart of VM addition destination selection process 2 in FIG. VM addition process sequence diagram Explanatory drawing which shows the flow of VM addition processing

<Apparatus configuration of the embodiment>
FIG. 7 shows an example of a virtual machine arrangement determining apparatus according to the present invention together with a network configuration to which the present invention is applied. In the figure, 10 is an NW / VM control apparatus, 20 is a data management server, 30 is a switch, Reference numeral 40 denotes a data center (DC), and 50 denotes a user terminal. Details of each device will be described below.

  The NW / VM control device 10 includes an NW / VM control function unit 11, a VM state monitoring function unit 12, a VM placement determination function unit 13, a pattern detection function unit 14, and a service chain pattern table (hereinafter simply “pattern table”) 15. Have

  The NW / VM control function unit 11 includes a VM generation / deletion function for executing VM generation / deletion, a VM migration function for executing VM migration, and an NW control capable of changing the route information of the network of the switch 30 in accordance with them. It has a function.

  The VM state monitoring function unit 12 has a VM / server resource information transmission / reception function capable of requesting / acquiring resource information such as VM and DC CPU, memory, and hard disk capacity.

  The VM placement determination function unit 13 determines which DC to migrate or add to the VM based on the timer for periodically determining the placement of the VM and the service chain pattern information obtained from the pattern table 15. A VM migration destination / addition destination selection function, and a VM migration / addition possibility judgment function for judging whether or not to migrate or add a VM based on the resource information obtained from the VM state monitoring function unit 12 ing.

  The pattern detection function unit 14 extracts the service chain pattern and the location of the user terminal (nearest DC) that uses the service chain of the pattern from the service chain information table and the user location table stored in the data management server 20. In the pattern table 15, the service chain pattern and the number of appearances of each nearest DC of the user terminal that uses the service chain of the pattern are described.

  The VM arrangement determination function unit 13, the pattern detection function unit 14, and the pattern table 15 described above are novel functional configurations implemented in the present invention.

  The data management server 20 stores a service chain information table 21, a user location table 22 and an NW function-VM number information table 23.

  In the service chain information table 21, as shown in FIG. 8A, the order of the NW functions through which traffic from the user terminal passes corresponding to the identifier for each user terminal (hereinafter referred to as user ID) is NW function identifiers (hereinafter referred to as function IDs) are described in this order. Note that “¥ 0” in the figure indicates that there is no NW function to be followed.

  In the user location table 22, as shown in FIG. 8B, the identifier of the DC with the shortest communication distance to the user terminal (hereinafter, nearest DCID) is described corresponding to each user ID. . As shown in FIG. 9, the user location table 22 is obtained from a database storing the relationship between the user terminal and the IP address, the IP address and the accommodating station, and the accommodating station and the nearest DC (DC connected to the accommodating station). Can be sought.

  In the NW function-VM number information table 23, as shown in FIG. 8C, the number of VMs (currently) of the entire network that provides the NW function is described corresponding to each function ID. .

  The switch 30 has a flow transfer function unit 31. The flow transfer function unit 31 transfers a data flow (packet) based on the path information set by the NW / VM control function unit 11 of the NW / VM control device 10.

  The DC (collection of servers) 40 includes a virtual machine (VM) 41 that provides an NW function and a VM control function unit 42. The VM control function unit 42 creates and deletes VMs in the DC, collects data alerts of VMs and held server resources, and sends them to the VM state monitoring function unit 12 of the NW / VM control device 10. Send.

  10 and 11 show an outline of the virtual machine placement determination according to the present invention. The pattern detection function unit 14 uses the table in the data management server 20 to read before and after the continuous NW function from the service chain information for each user terminal. A pattern for each of the two combinations is extracted and described in the pattern table 15 together with the appearance frequency (number of times) for each nearest DC of the user terminal, and the VM placement determination function unit 13 has a high frequency pattern, here the number of appearances is 3 times. When the above pattern is recognized, the placement destination DCs of the two NW function VMs included in the high frequency pattern are selected so as to have the same ratio as the appearance frequency in the nearest DC, and the NW / VM control function unit 11 Run the migration.

<Embodiment 1: Case where pattern location is periodically extracted and VM is migrated>
Here, as an example of the embodiment, the NW / VM control apparatus 10 periodically detects a pattern and a location from the service chain information table 21 and the user location table 22 and automatically migrates the VM. Show the procedure. In order to simply show the following procedure, a flowchart of the VM migration processing is shown in FIGS. 12 and 13, a sequence diagram is shown in FIG. 14, and a flow is shown in FIG.

  First, it is assumed that the VM arrangement determination function unit 13 sets the threshold α of the high-frequency pattern and the time T from the migration to the next migration.

  1. When the timer time T has elapsed since the completion of the previous migration, the VM placement determination function unit 13 makes a pattern information update request to the pattern detection function unit 14.

  2. Upon receiving the pattern information update request, the pattern detection function unit 14 accesses the data management server 20, and from the service chain information table 21 and the user location table 22, the service chain pattern for each user terminal, here before and after the continuous NW function Two combinations of patterns are extracted, and the patterns are described in the pattern table 15 together with the appearance frequency for each nearest DC of the user terminal.

  3. Data is transmitted from the pattern table 15 to the VM arrangement determination function unit 13, and the VM arrangement determination function unit 13 extracts and holds a pattern having an appearance frequency equal to or higher than the threshold α as a high frequency pattern (1).

  4). 4. When there is a high-frequency pattern, the VM placement determination function unit 13 determines that migration is necessary, and Proceed to If there is no frequent pattern, the timer is reset to 0 and exited.

  5. Set variable n = 1.

  6). The VM arrangement determination function unit 13 refers to the NW function-VM number information table 23 for the n-th frequently occurring high frequency pattern, and confirms the VM numbers of the two NW functions included in the high frequency pattern ( 2).

  7). The VM placement determination function unit 13 selects the VM placement destination DC so that the VMs of the two NW functions have the same ratio as the appearance frequency of the DC of the high-frequency pattern (3).

  8). The VM placement determination function unit 13 confirms the resource with the VM state monitoring function unit 12 in order to check whether the resources (CPU, memory, hard disk capacity) for accommodating 1 VM are sufficient for the selected VM placement destination DC. Send a request.

  9. Receiving the resource confirmation request, the VM state monitoring function unit 12 transmits a resource state request to the corresponding DC.

  10. The VM control function unit 42 of the DC 40 receives the resource state request and transmits the DC resource state to the VM state monitoring function unit 12.

  11. The VM state monitoring function unit 12 that has received the DC resource state transmits the information to the VM arrangement determination function unit 13.

  12 The VM arrangement determination function unit 13 determines whether the DC resource is sufficient based on the received DC resource state information (4). If it is determined that the resources are sufficient, the selected DC is determined as the migration destination DC. Otherwise, VMs that cannot be accommodated are not relocated from the original DC.

  13. If migration destinations are determined for all high-frequency patterns, a virtual machine migration execution request is transmitted to the NW / VM control function unit 11. Otherwise, n = n + 1 and 6. Return to.

  14 The NW / VM control function unit 11 that has received the migration execution request executes the migration of the corresponding VM (5).

  The part showing the novelty of the present invention in this procedure is as follows. ~ 7. It is.

<Embodiment 2: Case where VM is arranged based on user location and service chain information when adding NW function resource>
Here, as another example of the embodiment, when a VM resource of a certain NW function is tight, an alert is detected, and a VM is added, an addition destination DC is determined based on service chain information. The procedure for placement is shown. In order to simply show the following procedure, a flowchart of the VM addition processing is shown in FIGS. 16 to 18, a sequence diagram is shown in FIG. 19, and a flow is shown in FIG. In FIG. 19, the flow of steps 10 and 11 in the following procedure is omitted.

  First, it is assumed that the threshold α of the high-frequency pattern is set by the VM arrangement determination function unit 13.

  1. Due to the VM resource shortage, the VM control function unit 42 of the DC 40 alerts the VM state monitoring function unit 12 of the NW / VM control device 10 of the resource state (1).

  2. The VM state monitoring function unit 12 that has received the alert makes a VM addition request to the VM placement determination function unit 13.

  3. Upon receiving the VM addition request, the VM placement determination function unit 13 requests the pattern table 15 for pattern information regarding the corresponding NW function.

  4). The pattern table 15 that has received the pattern information request transmits data to the VM arrangement determination function unit 13.

  5. The VM placement determination function unit 13 that has received the pattern data extracts and holds, as a high-frequency pattern, a pattern whose appearance frequency is equal to or higher than the threshold α among the service chain patterns including the corresponding NW function (2). If there is no frequent pattern, 9. Proceed to

  6). Set variable n = 1.

  7). Set variable m = 1.

  8). 10. From the high-frequency pattern, select the n-th highest frequency pattern, and select the m-th highest frequency DC as the VM addition destination DC (3). Proceed to This DC selection method is referred to as VM addition destination selection method 1 in FIGS.

  9. If there is no frequent pattern, set p = 1.

  10. 10. Select a DC having a large number of user terminals using the NW function of the VM to be added p-th as a VM addition destination DC (3); Proceed to This DC selection method is referred to as VM addition destination method 2 in FIGS.

  11. The VM arrangement determination function unit 13 inquires of the VM state monitoring function unit 12 whether there is a resource in the VM addition destination DC. After confirming the resource state of each DC, the VM state monitoring function unit 12 transmits the information to the VM arrangement determination function unit 13.

  12 The VM arrangement determination function unit 13 transmits a resource confirmation request to the VM state monitoring function in order to confirm whether the resources (CPU, memory, capacity) accommodating 1 VM are sufficient for the selected VM addition destination DC. (4).

  13. Receiving the resource confirmation request, the VM state monitoring function unit 12 transmits a resource state request to the corresponding DC.

  14 The DC VM control function unit 42 receives the resource state request and transmits the DC resource state to the VM state monitoring function unit 12.

  15. The DC VM state monitoring function unit 12 that has received the DC resource state transmits the information to the VM arrangement determination function unit 13.

  16. The VM arrangement determination function unit 13 determines whether the DC resource is sufficient based on the received DC resource state information. If it is determined that the resource is sufficient, a VM addition execution request is transmitted to the NW / VM control function unit 11; Proceed to Otherwise, if there is a high-frequency pattern (when VM addition destination selection method 1 is selected), m = m + 1 is set. Return to. In addition, if any DC in which the nth high-frequency pattern appears has insufficient resources, n = n + 1 is set. Return to. If there is no frequent pattern (when VM addition destination selection method 2 is selected), p = p + 1 and 10. Return to.

17. Upon receiving the addition execution request, the NW / VM control function unit 11 adds the VM to the addition destination DC (5), and ends. Wait until the VM state monitoring function unit 12 receives an alert.
18. The part showing the novelty of the present invention in this procedure is as follows. ~ 8. It is.

10: NW / VM control device, 11: NW / VM control function unit, 12: VM state monitoring function unit, 13: VM placement determination function unit, 14: Pattern detection function unit, 15: Service chain pattern table,
20: Data management server, 21: Service chain information table, 22: User location table, 23: NW function-VM number information table,
30: switch,
40: Data center (DC), 41: Virtual machine (VM), 42: VM control function unit,
50: User terminal.

JP 2005-115653 A

Claims (7)

Two or more data centers in which virtual machines providing arbitrary network functions are arranged and one or more user terminals are connected to each other via a communication line, and traffic from the user terminals passes through a plurality of arbitrary virtual machines. An apparatus for determining placement of a virtual machine in service chaining technology that provides a communication path to perform,
A service chain information table describing the order of network functions through which traffic from user terminals passes,
A user location table describing a data center having the shortest communication distance with the user terminal;
An NW function-VM number information table describing the current number of virtual machines providing each network function;
A service chain pattern table describing the number of appearances of each nearest data center of the user terminal using the service chain pattern and the service chain of the pattern;
A user terminal that extracts a service chain pattern and a nearest data center of a user terminal that uses the service chain of the pattern from the service chain information table and the user location table, and uses the service chain pattern and the service chain of the pattern A pattern detection function unit that describes the number of appearances for each nearest data center in the service chain pattern table;
A service chain pattern in which the number of appearances is equal to or greater than a predetermined threshold is extracted from the service chain pattern table as a high frequency pattern, and the number of virtual machines that provide a network function included in the high frequency pattern is the NW function-VM number. A VM placement determination function unit that obtains from the information table and determines a migration destination of a virtual machine of each network function included in the high-frequency pattern according to a ratio of the number of appearances of each high-frequency pattern for each nearest data center; An apparatus for determining placement of a virtual machine, comprising: Two or more data centers in which virtual machines providing arbitrary network functions are arranged and one or more user terminals are connected to each other via a communication line, and traffic from the user terminals passes through a plurality of arbitrary virtual machines. An apparatus for determining placement of a virtual machine in service chaining technology that provides a communication path to perform,
A service chain information table describing the order of network functions through which traffic from user terminals passes,
A user location table describing a data center having the shortest communication distance with the user terminal;
An NW function-VM number information table describing the current number of virtual machines providing each network function;
A service chain pattern table describing the number of appearances of each nearest data center of the user terminal using the service chain pattern and the service chain of the pattern;
A user terminal that extracts a service chain pattern and a nearest data center of a user terminal that uses the service chain of the pattern from the service chain information table and the user location table, and uses the service chain pattern and the service chain of the pattern A pattern detection function unit that describes the number of appearances for each nearest data center in the service chain pattern table;
When a request to add a virtual machine that provides a predetermined network function is received, a pattern whose frequency of occurrence is greater than or equal to a predetermined threshold in the service chain pattern including the predetermined network function from the service chain pattern table is frequently A VM placement determination function unit that, if extracted as a pattern and has a high-frequency pattern, determines a data center having a higher appearance frequency and a higher appearance frequency as an addition destination of the virtual machine having the predetermined network function; A virtual machine arrangement determining apparatus comprising at least a virtual machine. When a request to add a virtual machine that provides a predetermined network function is received, a pattern whose frequency of occurrence is greater than or equal to a predetermined threshold in the service chain pattern including the predetermined network function from the service chain pattern table is frequently If there is no high-frequency pattern extracted as a pattern, a VM placement determination function unit that determines a data center with many user terminals that use the predetermined network function as an addition destination of a virtual machine of the predetermined network function is provided. The virtual machine arrangement determining apparatus according to claim 2, wherein: Two or more data centers in which virtual machines providing arbitrary network functions are arranged and one or more user terminals are connected to each other via a communication line, and traffic from the user terminals passes through a plurality of arbitrary virtual machines. A method for determining placement of a virtual machine in a service chaining technology that provides a communication path to be performed,
A service chain information table describing the order of network functions through which traffic from user terminals passes,
A user location table describing a data center having the shortest communication distance with the user terminal;
An NW function-VM number information table describing the current number of virtual machines providing each network function;
Using a service chain pattern and a service chain pattern table that describes the number of appearances for each nearest data center of a user terminal that uses the service chain of the pattern,
A user terminal that extracts a service chain pattern and a nearest data center of a user terminal that uses the service chain of the pattern from the service chain information table and the user location table, and uses the service chain pattern and the service chain of the pattern A pattern detection process for describing the number of appearances for each nearest data center in the service chain pattern table;
A service chain pattern in which the number of appearances is equal to or greater than a predetermined threshold is extracted from the service chain pattern table as a high frequency pattern, and the number of virtual machines that provide a network function included in the high frequency pattern is the NW function-VM number. A VM placement determination step of determining a migration destination of a virtual machine of each network function included in the high-frequency pattern according to a ratio of the number of appearances of each high-frequency pattern for each nearest data center, obtained from the information table A method for determining the placement of a virtual machine, comprising: Two or more data centers in which virtual machines providing arbitrary network functions are arranged and one or more user terminals are connected to each other via a communication line, and traffic from the user terminals passes through a plurality of arbitrary virtual machines. A method for determining placement of a virtual machine in a service chaining technology that provides a communication path to be performed,
A service chain information table describing the order of network functions through which traffic from user terminals passes,
A user location table describing a data center having the shortest communication distance with the user terminal;
An NW function-VM number information table describing the current number of virtual machines providing each network function;
Using a service chain pattern and a service chain pattern table that describes the number of appearances for each nearest data center of a user terminal that uses the service chain of the pattern,
A user terminal that extracts a service chain pattern and a nearest data center of a user terminal that uses the service chain of the pattern from the service chain information table and the user location table, and uses the service chain pattern and the service chain of the pattern A pattern detection process for describing the number of appearances for each nearest data center in the service chain pattern table;
When a request to add a virtual machine that provides a predetermined network function is received, a pattern whose frequency of occurrence is greater than or equal to a predetermined threshold in the service chain pattern including the predetermined network function from the service chain pattern table is frequently If there is a high frequency pattern extracted as a pattern, a VM placement determination step for determining a data center having a higher appearance frequency and a higher appearance frequency as an addition destination of the virtual machine having the predetermined network function is at least A method for determining the placement of a virtual machine. When a request to add a virtual machine that provides a predetermined network function is received, a pattern whose frequency of occurrence is greater than or equal to a predetermined threshold in the service chain pattern including the predetermined network function from the service chain pattern table is frequently If there is no high-frequency pattern extracted as a pattern, it includes a VM placement determination step of determining a data center having many user terminals that use the predetermined network function as an addition destination of a virtual machine having the predetermined network function. The virtual machine arrangement determination method according to claim 5.   A program for causing a computer to execute each step of the method according to any one of claims 4 to 6.

Images