JP5369824B2 - Movement schedule design device, movement schedule design program, movement schedule design method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a movement schedule design apparatus for designing a movement schedule of a program for preventing degradation of service performance. <P>SOLUTION: The movement schedule design apparatus includes a CPU amount distribution unit 2 for distributing a CPU amount available by a computer before or after movement by a number of programs, a movement time calculation unit 3 for calculating a band of a network for each program based on the CPU amount of each program and calculating a movement time for each program, a start time decision unit 4 for selecting a program of a calculated shortest movement time, taking a time preceding a movement completion schedule time by a movement time for the selected program as a movement start time for the selected program, and calculating a remaining data amount of an unselected remaining program, and a repetitive control unit 5 for taking the remaining program as a program having the remaining data amount, and exerting control so that each of the units performs processing repetitively. <P>COPYRIGHT: (C)2011,JPO&amp;INPIT

Description

  The present invention relates to a movement schedule design apparatus, a movement schedule design program, and a movement schedule design method for designing a schedule when a program moves between servers.

  The virtualized data center operates by virtualizing a server providing a service with software and starting a plurality of virtual servers (virtual instances: hereinafter simply referred to as instances) on the same physical server. Such a virtualized data center can have a scale of 1000 or more physical servers and 5000 to 7000 instances.

  In the virtual data center, physical servers are stored in racks and managed in rack units. Each physical server is directly connected to a switch in the rack (hereinafter referred to as an intra-rack SW (SW: switch device)), and the intra-rack SW is connected to an upper rack switch (herein referred to as an inter-rack SW). A network between all physical servers is constructed by connecting with a switch having a multi-stage configuration such as being connected. At that time, the bandwidth that can be used when passing through the inter-rack SW is narrower than the bandwidth of the intra-rack SW.

  Since the bandwidth of the SW between racks is narrow, if some instances of a service consisting of multiple instances are placed on a physical server in another rack and operated, data communication and switches such as other services and storage Service performance may be reduced due to contention for bandwidth. Therefore, an instance of the same service is designed to be placed on each physical server under the SW in the same rack.

  In addition, in the operation management of the virtual data center, the rearrangement of the instances is performed in order to adjust the increase of the capacity of the instances and the increase / decrease in the size / quantity of the instances. Also in this case, rearrangement design is performed so that instances of the same service do not cross between racks.

  When moving a service to each physical server in a different rack, even if all instances in the service are requested to be moved at the same time, the resource amount of each instance (allocated CPU (Central Processing Unit) performance, allocated memory amount, etc.) ), The movement time of the instances differs, so that during the service movement work, there are instances where instances operating in the rack before movement and instances operating in the rack after movement are mixed.

  The following documents are disclosed as conventional techniques.

JP-T-2004-508616 JP 2004-110791 A

  16 shows a WEB instance (indicated as WEB in FIG. 16), an APP instance (APP: application (indicated as APP in FIG. 16)), and a DB instance (DB: Database (in FIG. 16, as DB). The operation when data in the main storage device (hereinafter referred to as memory) used in the notation)) is moved to each physical server in the rack B will be described. Data communication generated in the process is performed via a service system NET constituted by switches SW-A, SW-C, and SW-B, and data movement of each instance is performed via a management system network. .

  FIG. 16A shows an initial state before movement. Here, when the movement of data in the memory of each instance (hereinafter simply referred to as movement) is started at the same time, the movement process is first completed in ascending order of the resource amount. In this example, the movement of the APP instance is completed (see FIG. 16B), and then the movement of the WEB instance is completed (see FIG. 16C).

  When there is a service provision request from the user in the state of FIG. 16B and FIG. 16C, in the conventional technology, between the instance that has been moved first and the instance that has not yet been moved, Since data communication that occurs during processing for providing a service starts on the service network, service performance may be degraded. Further, simply synchronizing the completion of the memory copy at the time of moving the service continuously copies the data in the memory while waiting for synchronization, and wastes resources.

  To solve this problem, wait until the migration of other instances is complete for the instances on the physical server after the migration, so that the instances in the service can be switched at the same time. It is also possible to switch after all instances have been moved. However, if there is a request from the user, even for the instance that has been moved, the memory of the instance that is running in the movement source physical server will continue to be updated. The movement process is continued. Therefore, resources are unnecessarily consumed while waiting for movement, and the movement of other instances is affected. Therefore, the entire movement time will increase.

  The present invention has been made in order to solve the above-described problems. For each instance constituting the service, the amount of memory of the instance, the amount of resources available at the time of movement, and the combination of instances that move at the same time are considered. , Move schedule design device, move schedule design program, move that can estimate the time required to move the instance and set the move start time of each instance so that the move completion time of all the instances that make up the service match The purpose is to provide a schedule design method.

  The movement schedule design apparatus is a movement schedule design apparatus that designs a schedule for moving each data in a memory when a plurality of programs are running to the memory of one or more other computers. A CPU amount distribution unit that distributes the CPU amount available on the source or destination computer by the number of programs to which data is to be moved, and distributes the allocated CPU amount to each of the programs; and the CPU Based on the CPU amount for each program distributed by the amount distribution unit, the network bandwidth used when moving each data in the memory is calculated for each program, and the calculated bandwidth for each program Based on the data amount for each program, the transfer for each program. A travel time calculation unit for calculating time, and a program for which the travel time calculated by the travel time calculation unit is the shortest, and the movement of the selected program from the scheduled time to complete the movement of all the programs to be moved A time that goes back by time is set as the movement start time of the selected program, and the remaining program data is moved within the shortest movement time with respect to the remaining programs other than the selected program. A start time determination unit that calculates the amount of unmoved data, which is the remaining amount of data when considered, for each remaining program, and each remaining program remains unmoved until there is no remaining program having the amount of unmoved data Assuming that the program has a data amount, the processing for the remaining program The CPU quantity distribution unit, said travel time calculation unit, and a repetition control section which controls to repetitively execute the start-time-determination unit.

  The movement schedule design program causes a computer to execute a process for designing a schedule for moving each data in a memory when each of a plurality of programs is running to the memory of one or more other computers. A migration schedule design program, wherein the CPU amount available on the source or destination computer is divided by the number of programs to which data is to be moved, and the allocated CPU amount is distributed to each of the programs. Then, based on the distributed CPU amount for each program, a network bandwidth used when moving each data in the memory is calculated for each program, and the calculated bandwidth for each program and the program are calculated. Based on the amount of data for each A step of calculating a travel time for each program, and a program that selects the program that has the shortest travel time, and a time that goes back by the travel time of the selected program from the scheduled time to complete the travel of all the programs to be moved As the movement start time of the selected program, and for the remaining programs other than the selected program, the remaining time when the data of the remaining program is considered to have moved within the shortest movement time. Calculating the amount of unmoved data that is the amount of data for each remaining program, and the remaining programs each having the amount of unmoved data until there is no remaining program having the amount of unmoved data. As above, the process for the remaining program is repeated for each step. To execute a step of controlling so as to execute, to a computer.

  In the movement schedule design method, a computer executes a process of designing a schedule for moving each data in a memory when a plurality of programs are running to the memory of one or more other computers. A migration schedule design method comprising the steps of distributing the amount of CPU available in a source or destination computer by the number of programs to which data is to be moved and distributing the allocated CPU amount to each of the programs. Then, based on the distributed CPU amount for each program, a network bandwidth used when moving each data in the memory is calculated for each program, and the calculated bandwidth for each program and the program are calculated. Based on the amount of data for each program Selecting a program having the shortest travel time calculated, and selecting a time that is retroactive from the scheduled time to complete the movement of all the programs to be moved. And the remaining data amount when the remaining program data is considered to have moved within the shortest movement time with respect to the remaining programs other than the selected program. Calculating a certain amount of unmoved data for each of the remaining programs, and assuming that each remaining program is a program having each amount of unmoved data until there is no remaining program having the amount of unmoved data. Repeat the above steps for each program. And controlling, the computer executes.

  Since it is possible to design a schedule with the movement completion time of the program to be moved at the same time, even if a user request occurs while the program movement work process is being performed according to the designed schedule, service performance degradation is prevented can do.

It is a figure for demonstrating the difference of the quality of the service type | system | group net by the conventional instance movement method and the instance movement method in this Embodiment. It is a figure for demonstrating an example of the function of the instance movement management system which concerns on this Embodiment. It is a figure which shows an example of the data which the movement schedule design apparatus concerning this Embodiment uses. It is a figure for demonstrating an example of the schedule design method which concerns on this Embodiment (initial state). It is a figure for demonstrating an example of the schedule design method which concerns on this Embodiment (distribution of CPU amount). It is a figure for demonstrating an example of the schedule design method which concerns on this Embodiment (calculation of movement band consumption). It is a figure for demonstrating an example of the schedule design method which concerns on this Embodiment (correction | amendment of movement band consumption). It is a figure for demonstrating an example of the schedule design method which concerns on this Embodiment (correction | moving CPU consumption for a movement). It is a figure for demonstrating an example of the schedule design method which concerns on this Embodiment (calculation of transfer time). It is a figure for demonstrating an example of the schedule design method which concerns on this Embodiment (movement start time determination of DB instance). It is a figure for demonstrating an example of the schedule design method which concerns on this Embodiment (movement start time determination of AP instance). It is a figure for demonstrating an example of the schedule design method which concerns on this Embodiment (movement start time determination of a WEB instance). It is a flowchart which shows an example of a process of the movement schedule design apparatus which concerns on this Embodiment. It is a figure which shows an example of the hardware constitutions of the movement schedule design apparatus which concerns on this Embodiment. It is a figure which shows an example of the hardware constitutions inside the movement schedule design apparatus which concerns on this Embodiment. It is a figure for demonstrating operation | movement when moving the memory data in each physical server installed in the conventional predetermined rack to each physical server in another rack.

  First, based on FIG. 1, the difference in quality between service-related nets (networks for providing services to users via the inter-rack SW) between the conventional instance migration method and the instance migration method according to the present embodiment will be described. . In this embodiment, a WEB instance that is a front end that receives a request from a user and transmits a processing result to the user, a DB instance in which data to be provided to the user is stored, and a request from the user Assume that one service is configured by each program of an application instance (hereinafter referred to as an APP instance) that performs processing by acquiring data from a DB instance. Each instance moves from a physical server connected to a predetermined intra-rack SW to a physical server connected to another intra-rack SW.

  In the conventional instance movement method (see FIG. 1A), when the movement start time of an instance set in advance is reached, all instances for providing a service simultaneously start the movement process. Here, for example, when the movement of the DB instance is completed, the DB instance operates on the physical server of the movement destination rack, and the other instances operate on the physical server of the movement source rack. When there is a request from the user in this state, the bandwidth of the service system net is consumed for the transaction that occurs between the DB instance and other instances.

  After that, when the movement of the APP instance is completed, similarly, the bandwidth of the service system net is consumed for the transaction that occurs between the WEB instance and other APP instances and DB instances.

  In this way, the occurrence of bandwidth consumption due to transactions between instances is caused by the running instances straddling the inter-rack SW. Therefore, in order to suppress this bandwidth consumption, this embodiment proposes a method for designing an instance movement schedule such that the movement completion times of all instances are the same time (see FIG. 1B). In this way, when the movement completion time is the same, the processing for the user request is completed in the movement source rack during the instance movement process, and therefore, no extra bandwidth is consumed in the service system net. Further, when the instance movement process is completed, the process is immediately performed in the movement destination rack.

  FIG. 2 illustrates an instance migration management system according to the present embodiment. The instance movement management system 100 includes a movement schedule design device 10, an instance management function unit 20, and a service movement management function unit 30.

  The service movement management function unit 30 receives a movement instruction for each service from the user 200, and instructs the movement schedule design apparatus 10 to design a movement schedule for the instance related to the service. In addition, the service movement management function unit 30 receives an instance movement completion notification from the instance management function unit 20 related to the service.

  The migration schedule design apparatus 10 designs a migration schedule for instances related to services for which a migration instruction has been received by the service migration management function unit 30.

  The instance management function unit 20 acquires the movement start time of each instance from the movement schedule design device 10 and moves the predetermined instance when the current time reaches the movement start time of the predetermined instance. The instance management function unit 20 notifies the service movement management function unit 30 of the completion of movement when all instances have been moved.

  In the present embodiment, the instance management function unit 20 and the service movement management function unit 30 are both computers having an arithmetic processing unit and a storage device.

  Here, the movement schedule design apparatus 10 will be further described. The movement schedule design device 10 includes a movement schedule design unit 1 that is a functional unit that performs a process for designing a movement schedule of an instance. The movement schedule design unit 1 includes a CPU amount distribution unit 2 and a movement time calculation unit 3. The start time determination unit 4 and the repetition control unit 5 are provided.

  The CPU amount distribution unit 2 apportions the CPU amount available on the physical server before or after the movement by the number of instances to be moved, and distributes the apportioned CPU amount to each instance.

  Based on the CPU amount for each instance distributed by the CPU amount distribution unit 2, the movement time calculation unit 3 determines the bandwidth of the network (management network) used for moving each data in the memory for each instance. The movement time for each instance is calculated based on the calculated bandwidth for each instance and the amount of data in the memory for each instance.

  The start time determination unit 4 selects the instance having the shortest travel time calculated by the travel time calculation unit 3 and goes back the travel time of the selected instance from the scheduled time to complete the movement of all the instances to be moved. Let the time be the movement start time of the selected instance. In addition, the start time determination unit 4 determines the remaining time when the remaining instance data is considered to have moved within the shortest travel time with respect to the remaining instances other than the selected instance (herein referred to as “remaining instances”). Is calculated for each remaining program.

  The iterative control unit 5 determines that the remaining program is an instance having the unmoved data amount until there is no remaining program having the unmoved data amount, and performs processing for the remaining program as the CPU amount distributing unit 2 and the moving time calculating unit. 3. Control to make the start time determination unit 4 repeatedly execute.

  The movement schedule design unit 1 uses a movement service configuration table 11, a physical server available resource amount table 12, a CPU resource versus transfer rate table 13, and an inter-rack available bandwidth table 14 that are preset data tables. Predetermined items of the movement schedule resource estimation table 15 and the movement schedule table 16 are calculated.

  FIG. 3 shows an example of a data table used by the movement schedule design unit 1.

  The movement service configuration table 11 includes an identifier of a service to be moved (service name), an identifier of an instance related to the service (instance name), an identifier of a rack before movement (rack before movement), and a physical server belonging to the rack before movement. FIG. 3A is a table having a correspondence relationship between each item of an identifier (physical server before movement), an identifier of a rack after movement (rack after movement), and an identifier of a physical server belonging to the rack after movement (physical server after movement) (FIG. 3A )reference). The physical server available resource amount table 12 includes an identifier of the physical server (physical server name), an available CPU amount of the physical server (available CPU amount), an available bandwidth of the physical server port (available port) (Refer to FIG. 3B). Note that the CPU amount that can be used is, for example, a physical server with a CPU clock frequency of 500 MHz, and when the amount of CPU that is used is 450 MHz, the amount of CPU that can be used is 50 (= 500-450). It is an indicator based on numbers.

  The CPU resource versus transfer rate table 13 has a physical server name and a moving speed per CPU performance unit of the physical server (see FIG. 3C). The CPU performance unit is an index indicating the performance of the CPU, such as 1 being 1 MHz of the CPU clock. For example, when the CPU performance unit is 1 per 1 MHz clock and the physical server having the CPU clock number of 500 MHz is used for 450 MHz, the available CPU amount is 50 (= 500-450). The moving speed per CPU performance unit is an index indicating the transfer speed on the network used for moving the instance when the CPU performance unit is 1, for example. As the moving speed per CPU performance unit, for example, a benchmark result considering the characteristics of a relay device such as a switch or a cable is used.

  The inter-rack usable bandwidth table 14 has an identifier (rack name) of the rack before and after the movement, and a bandwidth (usable switch bandwidth) available in the network between the racks before and after the movement (FIG. 3D). )reference).

  The migration schedule resource estimation table 15 includes an instance name, CPU consumption used when the instance moves (CPU consumption for migration), available CPU amount of the physical server (physical server available CPU), and movement of the instance. Bandwidth consumed when moving (consumption of bandwidth for movement), bandwidth that can be used at the port of the physical server after the movement, where the instance is stored (or scheduled to be stored) Available port bandwidth), available switch bandwidth, data amount of memory that has not been moved from the physical server before movement (unmoved memory), and time taken to move the instance (transfer time) (See FIG. 3E). Note that the physical server available CPU is a physical server that has either a pre-migration physical server on which the instance is operating or a post-migration physical server on which the instance is scheduled to operate and that has a small amount of available CPU. Adopted.

  The migration schedule table 16 includes a service name, an instance name, a physical server name before migration, a physical server name after migration, and a memory amount used by the instance, and has a total memory amount (total memory amount) to be migrated by the instance. . Further, the movement schedule table 16 has a movement amount for each time zone indicating a movement amount (transfer amount of data in the memory) of each instance for each time zone (see FIG. 3F).

  It should be noted that the values of the movement CPU consumption, the movement bandwidth consumption, the unmoved memory, the transfer time, and the movement amount per time period in the movement schedule table 16 are set or updated as the processing proceeds. It will be.

  Next, an example of the schedule design method of the movement schedule design apparatus 10 will be described with reference to FIGS. In this example, the movement of an instance between a pre-movement rack 301 that is a rack before movement and a post-movement rack 302 that is a rack after movement will be described. In addition, two instances of a WEB instance (denoted as INS-A-web in the drawing) and an APP instance (denoted as INS-A-app in the drawing) are running on one physical server in the rack 301 before movement. It is assumed that a DB instance (indicated as INS-A-db in the drawing) is operating on another physical server in the rack 301 before movement. In this example, it is assumed that the movement schedule design apparatus 10 designs a schedule when each of these instances moves to three physical servers in the rack 302 after movement.

  First, an initial state in this example is shown in FIG. FIG. 4A shows the CPU usage and the memory usage of each physical server in the pre-movement rack 301 and the post-movement rack 302 as a bar graph, and the pre-movement rack 301 and the post-movement rack 302. FIG. 6 is a schematic diagram showing a used band of a management system network between and as a bar graph. 4B shows the current table state of the movement schedule resource estimation table 15, and FIG. 4C shows the current table state of the movement schedule table 16. Hereinafter, description will be made with reference to FIGS. In addition, with respect to (B) and (C) in each figure, the portion where the numerical value is written and the portion updated by the processing are surrounded by a broken line.

  As shown in FIG. 5, the CPU amount distribution unit 2 obtains an available CPU amount for each physical server before movement on which the instance to be moved operates, and distributes the obtained available CPU amount to each instance to be moved. . In this example, as shown in FIG. 5 (A), the CPU amount distribution unit 2 applies to the physical server on which the WEB instance and APP instance are operating, other than the CPU amount used for the operation of the WEB instance and APP instance. The remaining CPU amount is apportioned by 2, and these are used as the CPU amount to be used for the movement (in this example, since the WEB instance and the APP instance are in operation, the apportioned by 2, for example, the instance is 3 If you are running one, apportion by 3.) In addition, since the number of instances is only one DB instance in the physical server where the DB instance is operating, the CPU amount distribution unit 2 uses the remaining CPU amount other than the CPU amount used for the operation of the DB instance as it is. The amount of CPU used when moving.

  This processing is performed by the CPU amount distribution unit 2 apportioning the physical server available CPUs of the migration schedule resource estimation table 15 by the number of instances for each physical server and writing the apportioning result to the migration CPU consumption (FIG. 5). (See (B)).

  The movement time calculation unit 3 integrates the calculated movement CPU consumption for each instance and the movement speed per CPU performance unit in the CPU resource vs. transfer speed table 13 to obtain the bandwidth consumption of the management system net for each instance. (Movement bandwidth consumption in the migration schedule resource estimation table 15) is calculated. As a result of this processing, a value for the movement band consumption is newly calculated as shown in FIG.

  In this state, the total value of the bandwidth consumption for moving the instance is the bandwidth of the management network (note that the bandwidth of the management network is the available port bandwidth and the available bandwidth as shown in FIG. 6A). The smaller of the switch bandwidths is used). Thus, if the total bandwidth consumed for each instance (referred to as the total bandwidth here) exceeds the bandwidth currently available on the management network (referred to as the available bandwidth here), the travel time The calculation unit 3 corrects the bandwidth consumption for each instance so that the total bandwidth becomes a bandwidth that can be used at present.

  As shown in FIG. 7A, the movement time calculation unit 3 divides the available bandwidth by the added bandwidth so that the added bandwidth falls within the available bandwidth, and the value obtained by the division and the current moving bandwidth By integrating the consumption values (see FIG. 6B), the movement band consumption values are corrected (see FIG. 7B).

  Moreover, the movement time calculation unit 3 also corrects the movement CPU consumption here. As shown in FIG. 8B, the CPU consumption for movement is also corrected at the same ratio as the ratio of the corrected amount of band consumption for movement.

  The movement time calculation unit 3 calculates the movement transfer time of each instance at the present time based on the corrected value of the band consumption for movement. That is, the movement time calculation unit 3 calculates the current transfer time based on the unmoved memory in the movement schedule resource estimation table 15 and the movement band consumption after the correction process, and moves the movement schedule resource estimate. The calculated transfer time is written in Table 15. Here, FIG. 9B shows the migration schedule resource estimation table 15 in which the transfer time is written.

  Next, the start time determination unit 4 selects an instance having the shortest transfer time calculated by the movement time calculation unit 3. In this example, the transfer time of the DB instance is calculated as 120 seconds, and since this is the shortest, the DB instance is selected.

  In addition, the start time determination unit 4 determines the time that goes back the movement time of the selected instance from the scheduled movement end time as the movement start time of the selected instance, and stores the data for the movement amount for each time zone in the movement schedule table 16. Write. For example, assuming that the scheduled movement end time is 3:00 am, the start time determination unit 4 writes time information from 2:58 am to 3:00 am, which is a time that goes back 120 seconds, in the movement amount for each time zone. In addition, the start time determination unit 4 calculates the transfer amount for each instance from 2:58 am to 3:00 am based on the transfer time and the movement bandwidth consumption for each instance, and writes the transfer amount for each time zone ( (See FIG. 10C).

  In this example, the schedule in which the transfer of all the data amount (total memory amount) of the memory used by the DB instance has been completed can be created, and thus the migration schedule design for the DB instance ends. At this time, the unmoved memory amount of the WEB instance is 512 MB (= 768 MB-256 MB), and the unmoved memory amount of the APP instance is 256 MB (= 512 MB-256 MB).

  If the same processing as described above with reference to FIGS. 4 to 10 is performed on the remaining memory amounts of the WEB instance and the APP instance, the state shown in FIG. 11 is obtained. That is, it is calculated that the transfer time (60 seconds) of the APP instance is shorter than the transfer time (120 seconds) of the WEB instance, and the movement start time of the next instance is added to the movement amount for each time zone in the movement schedule table 16. Written. At this point in time, a schedule has been created in which all of the amount of memory data (total amount of memory) used by the APP instance has been transferred, so the movement schedule design for the APP instance ends.

  Finally, the same processing as that described with reference to FIGS. 4 to 10 is performed only on the WEB instance, whereby the movement start time of the WEB instance is also determined (see FIG. 12C). Further, the movement schedule design device 10 issues a movement instruction to the instance management function unit 20 in the format shown in FIG.

  Next, the process of the movement schedule design apparatus 10 will be described with reference to the flowchart of FIG.

  The iterative control unit 5 selects an instance (indicated as ins in FIG. 13) in which a memory that has not yet moved remains (S1). The iterative control unit 5 determines whether or not all instances have been moved (S2), and if all instances have been moved (S2, yes), the process ends. On the other hand, when the movement of all the instances has not been completed (S2, no), the CPU amount distribution unit 2 newly creates or updates the movement schedule estimation table 15 (S3). In the case of new creation, the CPU amount distribution unit 2 writes items other than the movement CPU consumption, the movement band consumption, and the transfer time. Next, the CPU amount distribution unit 2 distributes the available CPU amount for each instance (S4).

  The movement time calculation unit 3 calculates the bandwidth consumption for each instance from the distributed CPU amount (S5), and determines whether the total bandwidth consumption exceeds the bandwidth available at the physical server port (S6). . When the bandwidth available at the port is exceeded (S6, yes), the movement time calculation unit 3 corrects the bandwidth consumption for each instance so that the total bandwidth consumption becomes the bandwidth available at the port (S7). ). On the other hand, when the bandwidth available at the port is not exceeded (S6, no), the travel time calculation unit 3 skips S7 and proceeds to S8.

  Next, the travel time calculation unit 3 determines whether or not the total bandwidth consumption exceeds the bandwidth that can be used in the management network that is a network between racks (S8). When the bandwidth that can be used in the management network is exceeded (S8, yes), the travel time calculation unit 3 calculates the bandwidth consumption for each instance so that the total bandwidth consumption becomes the bandwidth that can be used in the management network. Correction is performed (S9). On the other hand, if the bandwidth available on the management network is not exceeded (S8, no), the travel time calculation unit 3 skips S9 and proceeds to S10.

  The travel time calculation unit 3 corrects the CPU amount for each instance according to the bandwidth as necessary (S10). The movement time calculation unit 3 calculates the movement time of the unmoved memory for each instance from the amount of unmoved memory and the corrected band assigned to each instance (S11).

  The start time determination unit 4 selects the instance that has the shortest travel time (S12), and calculates the travel amount per hour in the travel schedule table 16 (S13). At this time, the start time determination unit 4 updates the unmoved memory in the migration schedule resource estimation table 15 corresponding to the remaining instances other than the selected instance.

  After that, if there are instances in which the total amount of memory moved does not reach the total memory amount, the iterative control unit 5 assumes that the remaining instance is an instance having a data amount of unmoved memory from the above S1. Control is performed so that the process of S13 is repeatedly executed (loop from S13 to S1). In this way, the processes from S1 to S13 are repeated until there is no unmoved instance by the repetition control unit 5.

  The present invention can be applied to the following computer system. FIG. 14 is a diagram illustrating an example of a computer system to which the present invention is applied. A computer system 900 shown in this figure includes a main body 901 incorporating a CPU, a disk drive, and the like, a display 902 that displays an image according to an instruction from the main body 901, a keyboard 903 for inputting various information to the computer system 900, A mouse 904 for designating an arbitrary position on the display screen 902a of the display 902 and a communication device 905 for accessing an external database or the like and downloading a program or the like stored in another computer system are provided. The communication device 905 may be a network communication card, a modem, or the like.

  A program for executing the steps described above in the computer system constituting the movement schedule design apparatus as described above can be provided as a movement schedule design program. By storing this program in a recording medium readable by the computer system, the computer system of the movement schedule design apparatus can be executed. A program for executing the above steps is stored in a portable recording medium such as a disk 910 or downloaded from a recording medium 906 of another computer system by the communication device 905. A movement schedule design program for causing the computer system 900 to have at least a movement schedule design function is input to the computer system 900 and compiled. This program causes the computer system 900 to operate as a movement schedule design system having a movement schedule design function. Further, this program may be stored in a computer-readable recording medium such as a disk 910, for example. Here, examples of the recording medium readable by the computer system 900 include an internal storage device such as a ROM and a RAM, a portable storage such as a disk 910, a flexible disk, a DVD disk, a magneto-optical disk, and an IC card. It includes a medium, a database holding a computer program, or other computer systems and the database, and various recording media accessible by a computer system connected via communication means such as a communication device 905.

  Here, a hardware configuration of the above-described movement schedule design apparatus will be described with reference to FIG. The migration schedule design device is a CPU (Central Processing Unit) 951 that is an arithmetic processing device, a memory 952 that is a main storage device, a disk 910 and the like, a disk drive 953 that reads and writes data, and a nonvolatile storage device. A nonvolatile storage device 954 is provided, and an I / O device 955 for controlling data communication with the outside is provided. Note that each functional block in the migration schedule design apparatus described above includes the CPU 951, the memory 952, the nonvolatile storage device 954, the hardware resources of the I / O device 955, and the firmware stored in the nonvolatile storage device 954 in advance. Is achieved through collaboration.

  According to the present embodiment, when all instances of a service are moved under another switch, all instances are switched at the same time, so that it is possible to prevent deterioration in service performance.

  The present invention can be implemented in various other forms without departing from the spirit or main features thereof. Therefore, the above-described embodiment is merely an example in all respects and should not be interpreted in a limited manner. The scope of the present invention is shown by the scope of claims, and is not restricted by the text of the specification. Moreover, all modifications, various improvements, substitutions and modifications belonging to the equivalent scope of the claims are all within the scope of the present invention.

  DESCRIPTION OF SYMBOLS 1 movement schedule design part, 2 CPU amount distribution part, 3 movement time calculation part, 4 start time determination part, 5 iteration control part, 10 movement schedule design apparatus, 20 instance management function part, 30 service movement management function part, 100 instance Mobility management system.

Claims (5)

A movement schedule design device for designing a schedule for moving each data in a memory when a plurality of programs are running to the memory of one or more other computers,
Of the source computer or destination computer , the available CPU amount of the computer with less available CPU amount is divided by the number of programs to which data should be moved, and the divided CPU amount is distributed to each of the programs. CPU amount distribution unit
By multiplying the CPU amount for each program distributed by the CPU amount distribution unit by a predetermined value indicating the transfer rate of the network used for moving the program in a computer with a small amount of available CPU , The network bandwidth used when moving each data in the memory is calculated for each program, and the moving time for each program is calculated based on the calculated bandwidth for each program and the data amount for each program. A travel time calculation unit for calculating
The program with the shortest travel time calculated by the travel time calculation unit is selected, and the selected time is determined by going back the travel time of the selected program from the scheduled time to complete the movement of all the programs to be moved. The remaining data when it is assumed that each of the remaining program data has moved within the shortest movement time with respect to the remaining program which is a program other than the selected program, as well as the movement start time of the program A start time determination unit that calculates an unmoved data amount that is an amount for each remaining program;
Of the remaining program until said pulp programs that have a non-mobile data amount is eliminated, the processing intended for the program CPU quantity distribution unit, the travel time calculating unit, repetitively executes the start-time-determination unit A repetitive control unit for controlling
A moving schedule design apparatus comprising: In the movement schedule design device according to claim 1,
The travel time calculation unit further includes:
When the total bandwidth obtained by adding the calculated bandwidths for each program exceeds the bandwidth that can be used in the network, the bandwidth for each program is corrected so that the total bandwidth becomes a bandwidth that can be used in the network. A movement schedule design apparatus for calculating a movement time for each program based on the band for each program and the data amount for each program. A migration schedule design program that causes a computer to execute a process for designing a schedule for moving each data in a memory when each of the programs is running to the memory of one or more other computers. And
Of the source computer or destination computer , the available CPU amount of the computer with less available CPU amount is divided by the number of programs to which data should be moved, and the divided CPU amount is distributed to each of the programs. And steps to
Each of the data in the memory is obtained by multiplying the CPU amount for each distributed program by a predetermined value indicating the transfer rate of the network used for moving the program in a computer with a small amount of available CPU. Calculating the network bandwidth used when moving the program for each program, and calculating the travel time for each program based on the calculated bandwidth for each program and the amount of data for each program;
A program that has the shortest travel time is selected, and a time that is traced back by the travel time of the selected program from the scheduled time to complete the travel of all the programs to be moved is defined as the travel start time of the selected program. In addition, with respect to the remaining program that is a program other than the selected program, unmoved data that is the remaining data amount when it is assumed that the data of the remaining program has moved within the shortest moving time. Calculating an amount for each remaining program;
A step of the of the remaining programs, until said pulp programs that have a non-mobile data amount is eliminated, controls to repetitively execute the processing applied to the program to the respective step,
A schedule design program that allows a computer to execute a program. In the movement schedule design program according to claim 3,
The step of calculating the travel time for each program further includes the step of using the total bandwidth in the network when the total bandwidth obtained by adding the calculated bandwidth for each program exceeds a bandwidth that can be used in the network. A movement schedule design program for correcting a band for each program so as to be a band, and calculating a movement time for each program based on the corrected band for each program and the data amount for each program. A move schedule design method in which a computer executes a process for designing a schedule for moving each data in a memory when a plurality of programs are running to the memory of one or more other computers. And
Of the source computer or destination computer , the available CPU amount of the computer with less available CPU amount is divided by the number of programs to which data should be moved, and the divided CPU amount is distributed to each of the programs. And steps to
Each of the data in the memory is obtained by multiplying the CPU amount for each distributed program by a predetermined value indicating the transfer rate of the network used for moving the program in a computer with a small amount of available CPU. Calculating the network bandwidth used when moving the program for each program, and calculating the travel time for each program based on the calculated bandwidth for each program and the amount of data for each program;
A program that has the shortest travel time is selected, and a time that is traced back by the travel time of the selected program from the scheduled time to complete the travel of all the programs to be moved is defined as the travel start time of the selected program. In addition, with respect to the remaining program that is a program other than the selected program, unmoved data that is the remaining data amount when it is assumed that the data of the remaining program has moved within the shortest moving time. Calculating an amount for each remaining program;
A step of the of the remaining programs, until said pulp programs that have a non-mobile data amount is eliminated, controls to repetitively execute the processing applied to the program to the respective step,
A moving schedule design method in which a computer is executed.

Images