[go: up one dir, main page]

WO2011086824A1 - Migration management device, migration management system, migration management method, and migration management program - Google Patents

Migration management device, migration management system, migration management method, and migration management program Download PDF

Info

Publication number
WO2011086824A1
WO2011086824A1 PCT/JP2010/073071 JP2010073071W WO2011086824A1 WO 2011086824 A1 WO2011086824 A1 WO 2011086824A1 JP 2010073071 W JP2010073071 W JP 2010073071W WO 2011086824 A1 WO2011086824 A1 WO 2011086824A1
Authority
WO
WIPO (PCT)
Prior art keywords
migration
program
memory
ecpd
mpd
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/JP2010/073071
Other languages
French (fr)
Japanese (ja)
Inventor
輪 長池
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
NEC Corp
Original Assignee
NEC Corp
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by NEC Corp filed Critical NEC Corp
Priority to JP2011549894A priority Critical patent/JPWO2011086824A1/en
Publication of WO2011086824A1 publication Critical patent/WO2011086824A1/en
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00—Arrangements for program control, e.g. control units
    • G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/46—Multiprogramming arrangements
    • G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
    • G06F9/5005—Allocation of resources, e.g. of the central processing unit [CPU] to service a request
    • G06F9/5011—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resources being hardware resources other than CPUs, Servers and Terminals
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F11/00—Error detection; Error correction; Monitoring
    • G06F11/30—Monitoring
    • G06F11/34—Recording or statistical evaluation of computer activity, e.g. of down time, of input/output operation ; Recording or statistical evaluation of user activity, e.g. usability assessment
    • G06F11/3457—Performance evaluation by simulation
    • G06F11/3461—Trace driven simulation
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00—Arrangements for program control, e.g. control units
    • G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/46—Multiprogramming arrangements
    • G06F9/48—Program initiating; Program switching, e.g. by interrupt
    • G06F9/4806—Task transfer initiation or dispatching
    • G06F9/4843—Task transfer initiation or dispatching by program, e.g. task dispatcher, supervisor, operating system
    • G06F9/485—Task life-cycle, e.g. stopping, restarting, resuming execution
    • G06F9/4856—Task life-cycle, e.g. stopping, restarting, resuming execution resumption being on a different machine, e.g. task migration, virtual machine migration

Definitions

  • the present invention relates to a migration management apparatus, a migration management system, a migration management method, and a migration management program.
  • Patent Document 1 discloses a file block allocation control program. The program preferentially allocates blocks with high I / O access frequency to high-speed storage devices.
  • Patent Document 2 discloses an application migration method between different types of devices. In this method, each component of the application is classified into a hardware dependent part and an independent part. In this method, the migration destination device includes an alternative component of the hardware dependent unit, and reconfigures the application together with the migrated hardware independent unit.
  • Patent Document 3 discloses a cluster system for migrating virtual machines. The system allocates resources to each virtual machine according to the required performance level. When an application achieves performance beyond expectations, the system reduces the resources of the virtual machine that executes the application, and increases the resources of the virtual machine that executes the application whose performance is degraded.
  • Patent Document 4 discloses a processing server constituting a distributed batch system. In response to a batch migration request from the management server, the server executes a job scheduling daemon and outputs the predicted completion time of the batch job and whether the completion before deadline is possible. JP-A-11-184739 JP2002236659 JP 2006-244441 A JP 2002-342098 A
  • the technique of the above-mentioned patent document has a problem that when a program is migrated between platforms having a difference in instruction execution performance, the performance degradation before and after the migration cannot be reduced.
  • the program of Patent Document 1 does not solve the performance problem regarding migration between platforms.
  • the program does not consider the performance difference before and after the allocation. Therefore, if the performance of the memory that can be allocated is not sufficient, there is a possibility that a significant performance degradation will occur after the allocation.
  • the techniques of Patent Documents 2, 3, and 4 do not consider the performance degradation after migration based on the instruction execution performance between platforms, and do not solve the above-described problems.
  • An object of the present invention is to provide a migration management apparatus, a migration management system, a migration management method, and a migration management program that solve the above-described problems.
  • the migration management apparatus includes a memory capacity allocated to a migration program that operates on a migration source machine connected to the own apparatus via a network, and an effective instruction speed (ECPs) on the migration source machine.
  • ECPs effective instruction speed
  • the performance calculation means for calculating and the migration program to be migrated to its own device allocate the memory of the own device satisfying the MPd and having the capacity, and output an error report when the allocation is not possible Resource allocation means is provided.
  • a migration management program stored on a computer-readable medium is a memory capacity allocated to a migration program that operates on a migration source machine connected to the computer through a network. And an effective instruction speed (ECPs) on the migration source machine, and an effective instruction speed (ECPd) of the migration program description instruction on the ECPs and its own device and a memory speed (MPe) as a basis of the ECPd.
  • ECPs effective instruction speed
  • ECPd effective instruction speed
  • MPe memory speed
  • a computer has a memory capacity allocated to a migration program that operates on a migration source machine connected to its own device through a network, and an effective instruction on the migration source machine.
  • ECPs speed
  • ECPd effective instruction speed
  • MPe memory speed
  • the migration management apparatus can reduce performance degradation before and after migration when a program is migrated between platforms having a difference in instruction execution performance.
  • FIG. 1 shows the overall configuration of the migration management system 400.
  • FIG. 2 shows the configuration of the user program 130.
  • FIG. 3 shows the configuration of the kernel table 114.
  • FIG. 4 shows the configuration of the program configuration table 121.
  • FIG. 5 is an operation flowchart of the migration management apparatus 300.
  • FIG. 6 shows the configuration of the migration management apparatus 300.
  • FIG. 1 is a configuration diagram of a migration management system 400 according to this embodiment.
  • the migration management system 400 includes one or more migration management apparatuses 300 and one or more migration source machines 200 connected via the network 140.
  • the migration management apparatus 300 inputs and executes (migrates or migrates) a user program (migration program 132 in the user program 130) executed on the migration source machine 200.
  • the present invention prevents the deterioration of the instruction speed due to emulation or the like that occurs with migration by using a high-speed memory.
  • the migration management apparatus 300 includes a physical machine 100 and a kernel program 110.
  • the migration source machine 200 includes a physical machine 100, a kernel program 110, a program management unit 120, and a user program 130.
  • the physical machine 100 is a computer that operates under program control, and includes a central processing unit (CPU), a main memory or a secondary memory (memory), a communication unit, an input / output device, peripheral devices, and the like.
  • the kernel program 110 is a program that operates on the physical machine 100 and has a function of allocating hardware resources of the physical machine 100 to each program executed on the physical machine 100.
  • the kernel program 110 indicates an operating system (OS) kernel or the like.
  • OS operating system
  • the kernel program 110 indicates a virtual machine monitor or the like in a virtual computing environment.
  • the kernel program 110 of the migration management apparatus 300 includes a resource allocation unit 111, a performance calculation unit 112, and an emulator 113.
  • the kernel program 110 of the migration management apparatus 300 holds a kernel table 114 describing the characteristics of the physical machine 100 in the memory.
  • the kernel program 110 also has general OS functions other than those described above.
  • the resource allocation unit 111 allocates a migration memory to each user program 130 to be migrated.
  • the performance calculation unit 112 calculates a required memory speed (MPd) of the memory. This part is based on, for example, the effective instruction speed (ECPs) on the migration source machine 200, the effective instruction speed (ECPd) of the instruction describing the user program 130 in the own apparatus, and the memory speed (MPe) that is the basis of ECPd. Then, MPd is calculated.
  • the memory speed (MPe) on which ECPd is based is the memory speed used for the calculation or measurement of ECPd.
  • the memory speed indicates, for example, a memory bus width, a memory cycle time, a memory access time, or a data transfer speed (transfer rate) combining these.
  • the memory transfer rate is assumed to be the data transfer rate.
  • the effective instruction speed ECP indicates, for example, an instruction execution time.
  • the execution command speed ECPd of the migration management apparatus 300 can be measured by executing the evaluation program 131 on the apparatus. If the instruction set or the mounted device (instruction or the like) is not compatible between the migration source machine 200 and the migration destination migration management apparatus 300 and the migrated program is executed on the emulator 113, the ECPd is a performance by emulation. The value reflects the decline.
  • the emulator 113 dynamically emulates this instruction and the like when the user program 130 described by the instruction of the migration source machine 200 and using the device of the machine cannot be directly executed by the migration management apparatus 300.
  • the emulator 113 has, for example, a CPU state conversion function, a code conversion function, and a device emulation function.
  • the CPU state conversion function converts the CPU state of the migration source machine 200 to the state of the migration management apparatus 300.
  • the code conversion function converts the instruction of the migration source machine 200 into the instruction of the migration management apparatus 300.
  • the device emulation function imitates a device operation that the migration source machine 200 has and the migration management apparatus 300 does not have.
  • the resource allocation unit 111 and the performance calculation unit 112 may be realized as a hardware device controlled by the kernel program 110, or realized by the CPU of the physical machine 100 executing the migration management program 101 on the memory. May be.
  • the kernel program 110 of the migration source machine 200 includes a sending unit 118. In response to an instruction from the user or the like, the sending unit 118 requests the specific migration management apparatus 300 to migrate the migration program 132 executed on the migration source machine 200.
  • the kernel program 110 of the migration source machine 200 also has a general OS function other than the above.
  • the program management unit 120 operates according to instructions from the kernel program 110.
  • the same unit includes a CPU information management module for managing CPU information assigned to each user program 130 and a memory information management module for managing memory information assigned to each user program 130.
  • this unit includes a driver management module that manages information of the physical machine 100 on which each user program 130 operates.
  • the program management unit 120 holds a program configuration table 121 for managing the above-described information on a memory.
  • the CPU information management module has a function of storing CPU architecture information of all user programs 130 running on the kernel program 110 in the program configuration table 121.
  • the CPU architecture information includes CPU instruction set information, endian information, or extended instruction information.
  • the memory information management module has a function of storing the size and memory speed information of all storage device modules (hereinafter referred to as memory elements) used by the user program 130 running on the kernel program 110 in the program configuration table 121. .
  • the device driver management module has a function of storing platform information held by all user programs 130 running on the kernel program 110 in the program configuration table 121.
  • the platform information includes hardware information of a virtual or real device operated by the user program 130.
  • FIG. 2 shows the configuration of the user program 130.
  • the user program 130 is a program that operates on the physical machine 100 using the functions provided by the kernel program 110.
  • the user program 130 includes a migration program 132 and an evaluation program 131.
  • the migration program 132 is an application program that does not have the authority to operate the kernel area, or a single-function application component in a distributed execution environment.
  • the migration program 132 is a virtual machine in a virtual computing environment, for example.
  • the migration program 132 is originally executed on the migration source machine 200, but is migrated to the migration management apparatus 300.
  • the evaluation program 131 is created by a user or the like reflecting, for example, a usage pattern such as an instruction used by the migration program 132 or an access pattern of a memory element.
  • the evaluation program 131 is described by an instruction of a specific physical machine 100 such as the migration source machine 200.
  • the evaluation program 131 includes, for example, an evaluation code section in which the number of execution instructions is known, an execution time measurement section for the evaluation code, an execution time per instruction from the measured execution time and the number of execution instructions, that is, an effective instruction speed ECPd. It is comprised from the part which calculates.
  • the number of execution instructions may be configured such that the emulator 113 or the like measures when the evaluation program 131 is executed.
  • the evaluation program 131 may have a function of recording an instruction history or a system call history, and a function of measuring a memory access amount and a memory access time for each memory element used by the user program 130.
  • the ECPd can be accurately measured when the evaluation program 131 is prepared for each migration program 132.
  • the plurality of user programs 130 may share one evaluation program 131.
  • the program code and data format of the user program 130 are not necessarily compiled to operate on the physical machine 100 of the migration source machine 200. This is because the migration source machine 200 may function as a relay device for another migration source machine 200 in the migration of the user program 130.
  • the sending unit 118 of the migration source machine 200 transmits the migration program 132 and the evaluation program 131 to the migration management apparatus 300 via the communication unit.
  • a single computer may have the configurations of both the migration management apparatus 300 and the migration source machine 200. That is, a computer that functions as the migration management apparatus 300 (or migration source machine 200) in a certain migration may be configured to function as the migration source machine 200 (or migration management apparatus 300) in another migration. Further, the configuration and type of the physical machine 100 of the migration management apparatus 300 are not necessarily the same as the physical machine 100 of the migration source machine 200.
  • the kernel program 110 of the migration management apparatus 300 holds configuration information of hardware managed by the kernel program 110 among the hardware included in the physical machine 100 as a kernel table (FIG. 3) in a memory.
  • FIG. 3 kernel table
  • the kernel table 114 includes CPU and device information and memory resource information managed by the kernel program 110 of the migration management apparatus 300.
  • the CPU information is, for example, an ISA (M) 501 that specifies a native instruction set of the physical machine 100 and an Endian (M) 502 that specifies whether the data is big endian or little endian. If both ISA (M) 501 and Endian (M) 502 of the two physical machines 100 match, CPU emulation is not necessary for program transfer between the machines, but CPU emulation is necessary if they are different.
  • the device information is a Device (M) 500 that specifies a device configuration included in the physical machine 100.
  • Device (M) 500 is a platform name assigned to each device configuration, for example.
  • the Device (M) 500 may directly store individual device information. If the Device (M) 500 of the two physical machines 100 match, device emulation is not required for program transfer between the machines, but device emulation is required if they are different. Device emulation is to emulate information output of a device when an input / output command is issued to a device that the physical machine 100 does not have. For example, the CPU emulates device operation by accessing emulation information stored in the memory.
  • the memory resource information is, for example, TransRate (M) 503 (memory speed) that is the data transfer speed of the mounted memory, and Free 504 that indicates the free capacity of the memory.
  • the kernel table 114 includes an emulation capability 505 that is information for specifying a CPU and a device that can be emulated by the emulator 113.
  • the emulation capability 505 is, for example, a list of instruction set identification information (ISA), data format identification information (Endian), and device configuration identification information (Device).
  • ISA instruction set identification information
  • Endian data format identification information
  • Device device configuration identification information
  • the EmulationCapability 505 stores a plurality of lists.
  • the information stored in the kernel table 114 may be acquired and set by the kernel program 110 from its own management data or the physical machine 100, or set by the system administrator of the migration management apparatus 300 as a system parameter. Also good. In the example of FIG.
  • the kernel program 110 of the migration management apparatus 300 manages four types of memory elements.
  • the memory element in the first row has a transfer speed of 12800 [MB / s] and is currently 2560 [MB].
  • the kernel program 110 operates on a physical machine 100 having an instruction set “ARCH_A”, a little-endian CPU, and a device configuration specified by the name “Platform1”. Represents that.
  • FIG. 4 shows the program configuration table 121 in the embodiment of the present invention.
  • the program configuration table 121 includes information on memory resources being used by the user program 130 and information on CPUs and devices of the operable physical machine 100.
  • the resource allocation unit 111 compares the program corresponding row of the program configuration table 121 with the CPU and device information of the kernel table 114, and the compatibility of the program between the migration source and the migration destination is confirmed. Judge whether there is. If there is no compatibility, the resource allocation unit 111 executes the migrated program on the emulator 113.
  • the program configuration table 121 stores, for each identifier (program ID 601) of the user program 130, device configuration information (Device (P) 607) of the physical machine 100 in which the program can operate.
  • the program configuration table 121 further includes, for each program ID 601, identification information (ISA (P) 602) and data format identification information (Endian (P) 603) of an instruction set in which the program is described.
  • the program configuration table 121 further stores a memory speed (TransRate (P) 604) and an allocated amount (Assign 605) of a memory allocated for storing instructions and data of the program. These pieces of information are usually information about the migration source machine 200, and the program management unit 120 inputs these pieces of information from the kernel program 110 of the own apparatus and stores them in the program configuration table 121. The program management unit 120 may input such information from the system administrator of the migration source machine 200. Further, these pieces of information may be described in the user program 130 and read by the program management unit 120 from the user program 130.
  • the program configuration table 121 may store an instruction speed (P) 606.
  • the instruction speed (P) 606 is an effective instruction speed under the physical machine 100 having a configuration specified by Device (P) 607, ISA (P) 602, Endian (P) 603, and TransRate (P) 604.
  • the program management unit 120 inputs the output value of the evaluation program 131 executed on the migration source machine 200 and stores it in the instruction speed (P) 606.
  • the program management unit 120 may input the specification information of the physical machine 100 and the output value of the evaluation program 131 from the system administrator of the migration source machine 200 and store them in the command speed (P) 606.
  • FIG. 4 shows that the user program “UID1” operates on the big endian CPU with the instruction set “ARCH_B” and the physical machine 100 having the device configuration “Platform1”, and uses 96 MB of a storage device of 120 MB / s. Illustrated.
  • the program management unit 120 and the sending unit 118 may be realized by hardware by a circuit, or may be realized by software by executing a program on the computer or the kernel program 110.
  • FIG. 5 is a flowchart of the migration operation of the migration management apparatus 300 of this embodiment. In the following description example, the migration management apparatus 300 migrates the migration program 132 of the user program 130 executed on the migration source machine 200 to the own apparatus.
  • the resource allocation unit 111 of the migration management apparatus 300 receives a migration request specifying a specific migration program 132 being executed on the same machine from the sending unit 118 of the migration source machine 200 and starts the migration process.
  • the apparatus may receive a migration request specifying the migration source machine 200 and a specific migration program 132 from a user terminal connected to the apparatus.
  • the resource allocation unit 111 acquires the characteristic information of the migration target user program 130 stored in the program configuration table 121 from the program management unit 120 of the migration source machine 200 (S101).
  • the characteristic information acquired here is Device (P) 607, ISA (P) 602, Endian (P) 603, Assign 605, and command speed (P) 606.
  • the acquired command speed (P) 606 is referred to as ECPs hereinafter.
  • the resource allocation unit 111 acquires the acquired Device (P) 607, ISA (P) 602, Endian (P) 603, and Device (M) 500, ISA (M) 501, Endian (M) of the kernel table 114 of the own device. Compare 502. When even a part does not match, the same part determines that the migration management apparatus 300 cannot execute the program code of the user program 130 as it is (the instruction is not compatible between the migration source and the migration destination) (No in S102). If they are not compatible, the resource allocation unit 111 has a list (Device, ISA, Endian) that matches the acquired pair of Device (P) 607, ISA (P) 602, and Endian (P) 603 (EndianCapability 505 in the kernel table 114).
  • the same part determines whether or not the own apparatus can perform the migration source emulation (S103). When the emulation cannot be performed (cannot be found) (No in S103), the same part determines that migration is impossible, outputs an error report to that effect, and ends the migration process (S108).
  • the output destination of the error report is a migration request transmission source (user terminal or migration source machine 200) or the like. If emulation is possible (Yes in S103), the same unit obtains the evaluation program 131 of the corresponding user program 130 via the sending unit 118 of the migration source machine 200, and executes it using the emulator 113 (S104). .
  • the resource allocation unit 111 sets, for example, the acquired Device (P) 607, ISA (P) 602, and Endian (P) 603 as operation parameters of the emulator 113.
  • the emulator 113 performs emulation of Device (P) 607, ISA (P) 602, and Endian (P) 603.
  • the resource allocating unit 111 may select one that emulates the acquired Device (P) 607, ISA (P) 602, and Endian (P) 603 from among a plurality of emulators 113.
  • the same unit loads the evaluation program 131 and the emulator 113 into a memory having a known memory access speed value MPe, allocates a memory having the speed as a data area, and executes the evaluation program 131 while emulating the emulator 113. .
  • the resource allocation unit 111 executes the execution of the evaluation program 131 so as not to be interrupted by preemption or the like.
  • the executed evaluation program 131 calculates and outputs the effective instruction speed ECPd by dividing the time required for executing the evaluation code by the number of executed instructions.
  • the output ECPd is an effective instruction speed when the memory access speed is MPe.
  • the performance calculation unit 112 calculates the required memory speed MPd using the following equation 1 based on the effective instruction speed ECPd and the memory speed MPe output from the evaluation program 131, and the resource It outputs to the allocation part 111 (S105).
  • This equation is based on the premise that the performance degradation of ECPd occurs due to an increase in the number of memory accesses caused by emulation and the like, and the performance degradation due to other factors can be ignored compared to ECPs. Under this assumption, the performance calculation unit 112 calculates the required memory speed MPd required to compensate for the performance degradation using the equation (1).
  • the resource allocation unit 111 may calculate the estimated value of ECPd without executing the evaluation program 131.
  • the resource allocation unit 111 refers to the TransRate (M) 503 and the Free 504 in the kernel table 114, and whether the storage device satisfying the requested memory speed MPd exists in the own device with sufficient free space (the acquired Assign 605 or higher). Determine (S106). If there is no memory (request memory) that satisfies the condition (No in S106), the same part determines that migration is impossible, outputs an error report to that effect, and ends the migration process (S109). .
  • the source (user terminal or migration source machine 200) of the migration request that received the error report selects another migration management device 300 in the migration management system 400, for example, and resends the migration request. If the requested memory exists (Yes in S106), the resource allocation unit 111 acquires the migration program 132 from the sending unit 118 of the migration source machine 200, allocates the requested memory, and starts executing the migration program 132 on the emulator 113. To do. That is, the same part performs live migration of the user program 130 (S107).
  • the acquired Device (P) 607, ISA (P) 602, and Endian (P) 603 match the Device (M) 500, ISA (M) 501, and Endian (M) 502 in the kernel table 114 of the own device.
  • the migration management apparatus 300 of this embodiment can achieve the following effects.
  • the first effect is that the performance degradation of the migration program 132 migrated from the latter to the former can be suppressed regardless of whether the instructions of the physical machine 100 included in the migration management apparatus 300 and the migration source machine 200 are compatible. It is. The reason is that even when the effective instruction speed is reduced due to emulation or the like, performance is maintained by allocating a high-speed memory to the migration program 132.
  • the performance calculation unit 112 of the present embodiment uses a calculation formula that reflects the CPU load (L) in the calculation of the required memory speed MPd (S105 in FIG. 5).
  • the effective instruction speed ECPd decreases as a result of memory access contention among a plurality of CPUs and input / output data transfer devices.
  • the evaluation program 131 is executed to calculate the effective instruction speed ECPd, a speed that is too low is measured.
  • the performance calculation unit 112 of the present embodiment uses the following calculation formula that reflects the CPU load in calculating the required memory speed MPd.
  • the CPU usage rate L is a measured value of the kernel program 110, and is a real number between 0 (meaning 0%) and 1 (meaning 100%).
  • a is an estimated value of the maximum delay magnification of ECPd due to an increase in the CPU usage rate, and is given to the performance calculation unit 112 as a parameter by the system administrator or the like.
  • the performance calculation unit 112 can be configured to use various functions as func (L). Under the assumption that the memory access time increases exponentially as the CPU load increases, the performance calculation unit 112 may be configured to use, for example, the following equation (3) as func (L).
  • the performance calculation unit 112 can be configured to use, for example, the following equation (4) as func (L).
  • the performance calculation unit 112 may be configured to use an approximate expression obtained as a result of actual measurement of the memory contention effect accompanying an increase in CPU load, for example, a higher-order function of the CPU load L. If the CPU usage rate is greater than or equal to a predetermined value, the resource allocation unit 111 may wait for the CPU usage rate to decrease before executing the evaluation program 131 (S105).
  • the migration management apparatus 300 of the present embodiment can calculate the required memory speed MPd sufficient to prevent performance degradation under normal load conditions even when the effective instruction speed ECPd is measured in a situation where the CPU usage rate is high.
  • the performance calculation unit 112 calculates MPd by increasing the MPd according to the CPU load at the time of measurement.
  • the present invention has been described above with reference to preferred embodiments. However, the present invention is not necessarily limited to the above-described embodiments, and can be carried out by changing the form within the scope of the technical idea. it can.
  • the emulator 113 may execute only one of CPU emulation and device emulation.
  • the kernel table 114 and the program configuration table 121 may not include the Device (M) 500, the Device (P) 607, the ISA (M) 501, or the like, the ISA (P) 602, or the like.
  • the present invention can be applied to migration between physical machines 100 having code compatibility.
  • the present invention can reduce performance degradation due to an effective instruction speed difference caused by a microcode mounting difference or the like.
  • the migration source machine 200 includes the program management unit 120.
  • the computer described as the physical machine 100 in the present embodiment may be a virtual computer.
  • FIG. 6 shows the configuration of the migration management apparatus 300 of the third embodiment.
  • the apparatus includes a performance calculation unit 112 and a resource allocation unit 111.
  • the performance calculation unit 112 inputs the memory capacity allocated to the migration program 132 operating on the migration source machine 200 connected to the own apparatus through the network 140 and the effective instruction speed (ECPs) on the migration source machine 200. The same part calculates the required memory speed (MPd) based on the ECPs and the effective instruction speed (ECPd) of the migration program 132 description instruction on its own device and the memory speed (MPe) that is the basis of the ECPd.
  • the resource allocation unit 111 allocates the memory included in the own device that satisfies MPd and has the capacity for the migration program 132 to be migrated to the own device, and outputs an error report when the allocation cannot be performed.
  • the migration management apparatus 300 can reduce performance degradation before and after migration when migrating a program between platforms having instruction execution performance differences. The reason is that even when the effective instruction speed is reduced due to emulation or the like, performance is maintained by allocating a high-speed memory to the migration program 132.
  • the present invention has been described with reference to the embodiments (and examples), the present invention is not limited to the above embodiments (and examples). Various changes that can be understood by those skilled in the art can be made to the configuration and details of the present invention within the scope of the present invention. This application claims the priority on the basis of Japanese application Japanese Patent Application No. 2010-003829 for which it applied on January 12, 2010, and takes in those the indications of all here.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Software Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Hardware Design (AREA)
  • Quality & Reliability (AREA)
  • Debugging And Monitoring (AREA)

Abstract

Devices fail to reduce performance loss before and after migration when performing migration of a program between platforms having a difference in instruction execution performance. The disclosed device is provided with a performance calculation unit and a resource assignment unit. Into the performance calculation unit is input capacity of memory which has been assigned to a migration program which operates upon a source machine which is connected through a network to the device itself, and the actual instruction speed (ECPs) upon the source machine; whereupon, on the basis of ECPs, the actual instruction speed (ECPd) of the migration program description instructions upon the device itself, and the memory speed that forms the basis of the ECPd, the required memory speed (MPd) is calculated. The resource assignment unit performs assignment of memory that is provided to a device that satisfies MPd for the migration program that is to be migrated to the device and that has the memory capacity, and if the assignment cannot be performed, outputs an error report.

Description

マイグレーション管理装置、マイグレーション管理システム、マイグレーション管理方法、及びマイグレーション管理プログラムMigration management apparatus, migration management system, migration management method, and migration management program

　本発明はマイグレーション管理装置、マイグレーション管理システム、マイグレーション管理方法、及びマイグレーション管理プログラムに関する。 The present invention relates to a migration management apparatus, a migration management system, a migration management method, and a migration management program.

　特許文献１は、ファイルブロック割り当て用制御プログラムを開示する。同プログラムは、Ｉ／Ｏアクセスの頻度が高いブロックを高速な記憶装置に優先的に割り当てる。
　特許文献２は、異種デバイス間におけるアプリケーションのマイグレーション方法を開示する。同方法は、アプリケーションの各コンポーネントをハードウェア依存部・非依存部に分類する。同方法において、マイグレーション先デバイスは、ハードウェア依存部の代替コンポーネントを備え、マイグレーションされたハードウェア非依存部とともにアプリケーションを再構成する。
　特許文献３は、仮想マシンをマイグレーションするクラスタシステムを開示する。
同システムは、要求される性能レベルに応じて、各仮想マシンに対して資源割り当てを行う。或るアプリケーションが予想以上の性能を達成している場合、同システムは、このアプリケーションを実行する仮想マシンの資源を減らして、性能低下しているアプリケーションを実行する仮想マシンの資源を増加させる。
　特許文献４は、分散バッチシステムを構成する処理サーバを開示する。同サーバは、管理サーバからバッチ移行要求を受けて、ジョブスケジューリングデーモンを実行してバッチジョブの予測完了時間及びデッドライン前の完了可／不可を出力する。
特開平１１−１８４７３９ 特開２００２−２３６５９２ 特開２００６−２４４４８１ 特開２００２−３４２０９８ Patent Document 1 discloses a file block allocation control program. The program preferentially allocates blocks with high I / O access frequency to high-speed storage devices.
Patent Document 2 discloses an application migration method between different types of devices. In this method, each component of the application is classified into a hardware dependent part and an independent part. In this method, the migration destination device includes an alternative component of the hardware dependent unit, and reconfigures the application together with the migrated hardware independent unit.
Patent Document 3 discloses a cluster system for migrating virtual machines.
The system allocates resources to each virtual machine according to the required performance level. When an application achieves performance beyond expectations, the system reduces the resources of the virtual machine that executes the application, and increases the resources of the virtual machine that executes the application whose performance is degraded.
Patent Document 4 discloses a processing server constituting a distributed batch system. In response to a batch migration request from the management server, the server executes a job scheduling daemon and outputs the predicted completion time of the batch job and whether the completion before deadline is possible.
JP-A-11-184739 JP2002236659 JP 2006-244441 A JP 2002-342098 A

　上記特許文献の技術は、命令実行性能差があるプラットフォーム間でプログラムのマイグレーションを行う際、マイグレーション前後の性能低下を低減できないという課題がある。
　特許文献１のプログラムは、プラットフォーム間のマイグレーションについての性能問題を解決しない。また、当該プログラムは、割り当てに際して割り当て前後の性能差を考慮しない。したがって、割当可能なメモリの性能が十分でない場合、割当後に大幅な性能低下を発生させる可能性がある。
　特許文献２、３及び４の技術は、プラットフォーム間の命令実行性能に基づくマイグレーション後の性能低下を考慮しておらず、上述の課題を解決しない。本発明の目的は上述の課題を解決する、マイグレーション管理装置、マイグレーション管理システム、マイグレーション管理方法、及びマイグレーション管理プログラムを提供することである。 The technique of the above-mentioned patent document has a problem that when a program is migrated between platforms having a difference in instruction execution performance, the performance degradation before and after the migration cannot be reduced.
The program of Patent Document 1 does not solve the performance problem regarding migration between platforms. In addition, the program does not consider the performance difference before and after the allocation. Therefore, if the performance of the memory that can be allocated is not sufficient, there is a possibility that a significant performance degradation will occur after the allocation.
The techniques of Patent Documents 2, 3, and 4 do not consider the performance degradation after migration based on the instruction execution performance between platforms, and do not solve the above-described problems. An object of the present invention is to provide a migration management apparatus, a migration management system, a migration management method, and a migration management program that solve the above-described problems.

　本発明の一実施の形態に係るマイグレーション管理装置は、自装置とネットワークで接続された移行元マシン上で動作する移行プログラムに割り当てられたメモリの容量及び前記移行元マシン上における実効命令速度（ＥＣＰｓ）を入力して、前記ＥＣＰｓ及び自装置上に於ける前記移行プログラム記述命令の実効命令速度（ＥＣＰｄ）及び前記ＥＣＰｄの基礎となるメモリ速度（ＭＰｅ）に基づいて、要求メモリ速度（ＭＰｄ）を算出する性能算出手段と、自装置へ移行される前記移行プログラムに対し前記ＭＰｄを満たし、かつ前記容量を持った自装置の備えるメモリ割り当てを行い、前記割り当てが出来ないときはエラー報告を出力する資源割り当て手段を備える。
　本発明の一実施の形態に係るコンピュータ読み取り可能な媒体に格納されたマイグレーション管理プログラムは、コンピュータに、自装置とネットワークで接続された移行元マシン上で動作する移行プログラムに割り当てられたメモリの容量及び前記移行元マシン上における実効命令速度（ＥＣＰｓ）を入力して、前記ＥＣＰｓ及び自装置上に於ける前記移行プログラム記述命令の実効命令速度（ＥＣＰｄ）及び前記ＥＣＰｄの基礎となるメモリ速度（ＭＰｅ）に基づいて、要求メモリ速度（ＭＰｄ）を算出する性能算出処理と、自装置へ移行される前記移行プログラムに対し前記ＭＰｄを満たし、かつ前記容量を持った自装置の備えるメモリ割り当てを行い、前記割り当てが出来ないときはエラー報告を出力する資源割り当て処理を実行させる。
　本発明の一実施の形態に係るマイグレーション管理方法は、コンピュータが、自装置とネットワークで接続された移行元マシン上で動作する移行プログラムに割り当てられたメモリの容量及び前記移行元マシン上における実効命令速度（ＥＣＰｓ）を入力して、前記ＥＣＰｓ及び自装置上に於ける前記移行プログラム記述命令の実効命令速度（ＥＣＰｄ）及び前記ＥＣＰｄの基礎となるメモリ速度（ＭＰｅ）に基づいて、要求メモリ速度（ＭＰｄ）を算出し、自装置へ移行される前記移行プログラムに対し前記ＭＰｄを満たし、かつ前記容量を持った自装置の備えるメモリ割り当てを行い、前記割り当てが出来ないときはエラー報告を出力する。 The migration management apparatus according to an embodiment of the present invention includes a memory capacity allocated to a migration program that operates on a migration source machine connected to the own apparatus via a network, and an effective instruction speed (ECPs) on the migration source machine. ) To determine the required memory speed (MPd) based on the ECPs, the effective instruction speed (ECPd) of the migration program description instruction on the device, and the memory speed (MPe) on which the ECPd is based. The performance calculation means for calculating and the migration program to be migrated to its own device allocate the memory of the own device satisfying the MPd and having the capacity, and output an error report when the allocation is not possible Resource allocation means is provided.
A migration management program stored on a computer-readable medium according to an embodiment of the present invention is a memory capacity allocated to a migration program that operates on a migration source machine connected to the computer through a network. And an effective instruction speed (ECPs) on the migration source machine, and an effective instruction speed (ECPd) of the migration program description instruction on the ECPs and its own device and a memory speed (MPe) as a basis of the ECPd. ) Based on the performance calculation processing for calculating the required memory speed (MPd), and the allocation of the memory included in the own device having the capacity and satisfying the MPd for the migration program to be migrated to the own device, If the allocation is not possible, the resource allocation process that outputs an error report is executed. That.
In the migration management method according to an embodiment of the present invention, a computer has a memory capacity allocated to a migration program that operates on a migration source machine connected to its own device through a network, and an effective instruction on the migration source machine. Enter the speed (ECPs), and based on the ECPs and the effective instruction speed (ECPd) of the migration program description instruction on the device and the memory speed (MPe) on which the ECPd is based, the required memory speed ( MPd) is calculated, and the memory allocation of the self-device that satisfies the MPd and has the capacity is performed for the migration program to be migrated to its own device, and an error report is output when the allocation is not possible.

　本発明にかかるマイグレーション管理装置は、命令実行性能差があるプラットフォーム間でプログラムのマイグレーションを行う際、マイグレーション前後の性能低下を低減することができる。 The migration management apparatus according to the present invention can reduce performance degradation before and after migration when a program is migrated between platforms having a difference in instruction execution performance.

図１は、マイグレーション管理システム４００の全体構成を示す。FIG. 1 shows the overall configuration of the migration management system 400. 図２は、ユーザプログラム１３０の構成を示す。FIG. 2 shows the configuration of the user program 130. 図３は、カーネル表１１４の構成を示す。FIG. 3 shows the configuration of the kernel table 114. 図４は、プログラム構成表１２１の構成を示す。FIG. 4 shows the configuration of the program configuration table 121. 図５は、マイグレーション管理装置３００の動作フローチャートである。FIG. 5 is an operation flowchart of the migration management apparatus 300. 図６は、マイグレーション管理装置３００の構成を示す。FIG. 6 shows the configuration of the migration management apparatus 300.

　１００　　物理マシン
　１０１　　マイグレーション管理プログラム
　１１０　　カーネルプログラム
　１１１　　資源割り当て部
　１１２　　性能算出部
　１１３　　エミュレータ
　１１４　　カーネル表
　１１８　　送出部
　１２０　　プログラム管理部
　１２１　　プログラム構成表
　１３０　　ユーザプログラム
　１３１　　評価プログラム
　１３２　　移行プログラム
　１４０　　ネットワーク
　２００　　移行元マシン
　３００　　マイグレーション管理装置
　４００　　マイグレーション管理システム
　５００　　Ｄｅｖｉｃｅ（Ｍ）
　５０１　　ＩＳＡ（Ｍ）
　５０２　　Ｅｎｄｉａｎ（Ｍ）
　５０３　　ＴｒａｎｓＲａｔｅ（Ｍ）
　５０４　　Ｆｒｅｅ
　５０５　　ＥｍｕｌａｔｉｏｎＣａｐａｂｉｌｉｔｙ
　６０１　　プログラムＩＤ
　６０２　　ＩＳＡ（Ｐ）
　６０３　　Ｅｎｄｉａｎ（Ｐ）
　６０４　　ＴｒａｎｓＲａｔｅ（Ｐ）
　６０５　　Ａｓｓｉｇｎ
　６０６　　命令速度（Ｐ）
　６０７　　Ｄｅｖｉｃｅ（Ｐ） DESCRIPTION OF SYMBOLS 100 Physical machine 101 Migration management program 110 Kernel program 111 Resource allocation part 112 Performance calculation part 113 Emulator 114 Kernel table 118 Sending part 120 Program management part 121 Program structure table 130 User program 131 Evaluation program 132 Migration program 140 Network 200 Migration origin machine 300 Migration management device 400 Migration management system 500 Device (M)
501 ISA (M)
502 Endian (M)
503 TransRate (M)
504 Free
505 EmulationCapability
601 Program ID
602 ISA (P)
603 Endian (P)
604 TransRate (P)
605 Assign
606 Command speed (P)
607 Device (P)

〈第一の実施形態〉
　図１は、本実施の形態のマイグレーション管理システム４００の構成図である。マイグレーション管理システム４００は、ネットワーク１４０を経由して接続された、一台以上のマイグレーション管理装置３００と一台以上の移行元マシン２００を包含する。マイグレーション管理装置３００は、移行元マシン２００で実行されている利用者のプログラム（ユーザプログラム１３０内の移行プログラム１３２）を入力して実行する（マイグレート、または移行する）。
　本発明は、マイグレーションに伴って発生するエミュレーション等による命令速度の劣化を、高速なメモリを使用することで防止する。
　図１において、マイグレーション管理装置３００は、物理マシン１００、カーネルプログラム１１０を備える。移行元マシン２００は、物理マシン１００、カーネルプログラム１１０、プログラム管理部１２０、ユーザプログラム１３０を備える。
　物理マシン１００は、プログラム制御により動作するコンピュータであり、Ｃｅｎｔｒａｌ　Ｐｒｏｃｅｓｓｉｎｇ　Ｕｎｉｔ（ＣＰＵ）、メインメモリまたは二次メモリ（メモリ）、通信部、入出力装置、周辺機器等を備える。
　カーネルプログラム１１０は物理マシン１００上で動作するプログラムであり、物理マシン１００の有するハードウェア資源を当該物理マシン１００上で実行される各プログラムに割り当てる機能を有する。前記カーネルプログラム１１０は、Ｏｐｅｒａｔｉｎｇ　Ｓｙｓｔｅｍ（ＯＳ）のカーネル等を指すが、例えば仮想計算環境下においては仮想マシンモニタ等を指す。
　マイグレーション管理装置３００のカーネルプログラム１１０は、資源割り当て部１１１、性能算出部１１２と、エミュレータ１１３を備える。マイグレーション管理装置３００のカーネルプログラム１１０は、物理マシン１００の特性を記述したカーネル表１１４をメモリ上に保持する。カーネルプログラム１１０は、上記以外の一般的なＯＳ機能も有する。
　資源割り当て部１１１は、移行される各ユーザプログラム１３０に移行用のメモリを割り当てる。
　性能算出部１１２は当該メモリの要求メモリ速度（ＭＰｄ）を算出する。同部は、例えば、移行元マシン２００上における実効命令速度（ＥＣＰｓ）、及び自装置におけるユーザプログラム１３０を記述した命令の実効命令速度（ＥＣＰｄ）及びＥＣＰｄの基礎となるメモリ速度（ＭＰｅ）に基づいて、ＭＰｄを算出する。ＥＣＰｄの基礎となるメモリ速度（ＭＰｅ）は、ＥＣＰｄの算定または測定等に用いたメモリ速度である。
　ここで、メモリ速度は、例えば、メモリバス幅、メモリサイクルタイム、メモリアクセスタイム、あるいはこれらを組み合わせたデータ転送速度（転送レート）等を指す。以降メモリ転送速度は、データ転送速度であるとする。
　実効命令速度ＥＣＰは、例えば命令の実行時間を指す。マイグレーション管理装置３００の実行命令速度ＥＣＰｄは、当該装置上で評価プログラム１３１を実行することで測定可能である。移行元マシン２００と移行先のマイグレーション管理装置３００との間に命令セットまたは実装デバイス（命令等）の互換性がなく、移行されたプログラムがエミュレータ１１３上で実行される場合、ＥＣＰｄはエミュレーションによる性能低下を反映した値となる。
　エミュレータ１１３は、移行元マシン２００の命令で記述され、かつ当該マシンのデバイスを使用するユーザプログラム１３０が、マイグレーション管理装置３００で直接実行不可能な場合に、この命令等を動的にエミュレーションして実行する。エミュレータ１１３は、例えば、ＣＰＵ状態変換機能、コード変換機能、デバイスエミュレーション機能を有する。
　ＣＰＵ状態変換機能は、移行元マシン２００のＣＰＵ状態をマイグレーション管理装置３００の状態へ変換する。コード変換機能は、移行元マシン２００の命令をマイグレーション管理装置３００の命令に変換する。デバイスエミュレーション機能は、移行元マシン２００が備え、マイグレーション管理装置３００が備えないデバイス動作を模倣する。
　資源割り当て部１１１、及び性能算出部１１２は、カーネルプログラム１１０から制御されるハードウェア装置として実現されても良いし、物理マシン１００のＣＰＵがメモリ上のマイグレーション管理プログラム１０１を実行することで実現されても良い。
　移行元マシン２００のカーネルプログラム１１０は送出部１１８を備える。送出部１１８は、利用者等の指示を受けて、特定のマイグレーション管理装置３００に対して、移行元マシン２００で実行されている移行プログラム１３２の移行を依頼する。移行元マシン２００のカーネルプログラム１１０は、上記以外の一般的なＯＳ機能も有する。
　プログラム管理部１２０は、カーネルプログラム１１０の指示で動作する。同部は、各ユーザプログラム１３０に割り当てられたＣＰＵ情報を管理するＣＰＵ情報管理モジュールと、各ユーザプログラム１３０に割り当てられたメモリ情報を管理するメモリ情報管理モジュールを備える。さらに同部は、各ユーザプログラム１３０が動作する物理マシン１００の情報を管理するドライバ管理モジュールを備える。プログラム管理部１２０は、上述した情報を管理する為のプログラム構成表１２１をメモリ上に保持する。
　ＣＰＵ情報管理モジュールは、カーネルプログラム１１０上で稼動している全てのユーザプログラム１３０のＣＰＵアーキテクチャ情報をプログラム構成表１２１に格納する機能を有する。ＣＰＵアーキテクチャ情報は、ＣＰＵ命令セット情報、エンディアン情報、あるいは拡張命令情報を包含する。
　メモリ情報管理モジュールは、カーネルプログラム１１０上で稼動しているユーザプログラム１３０が使用している全ての記憶装置モジュール（以降メモリ要素）のサイズ及びメモリ速度情報をプログラム構成表１２１に格納する機能を有する。
　デバイスドライバ管理モジュールは、カーネルプログラム１１０上で稼動している全てのユーザプログラム１３０が持つプラットフォーム情報をプログラム構成表１２１に格納する機能を有する。プラットフォーム情報は、ユーザプログラム１３０が操作する仮想あるいは実デバイスのハードウェア情報を包含する。
　図２は、ユーザプログラム１３０の構成を示す。ユーザプログラム１３０は、カーネルプログラム１１０の提供する機能を使用して物理マシン１００上で動作するプログラムである。ユーザプログラム１３０は、移行プログラム１３２と評価プログラム１３１を包含する。
　移行プログラム１３２は、カーネル領域を操作する権限をもたないアプリケーションプログラム、あるいは分散実行環境下における単機能なアプリケーションコンポーネントである。移行プログラム１３２は、例えば仮想計算環境下では仮想マシンである。移行プログラム１３２は、元々移行元マシン２００で実行されているが、マイグレーション管理装置３００に移行される。
　評価プログラム１３１は、例えば、移行プログラム１３２が使用している命令等の使用パターンやメモリ要素のアクセスパターンを反映して利用者等によって作成されるものである。評価プログラム１３１は、移行元マシン２００等の特定の物理マシン１００の命令で記述される。評価プログラム１３１は、例えば、実行命令数が既知である評価用コード部、評価用コードの実行時間計測部、計測された実行時間と実行命令数から一命令当たりの実行時間、即ち実効命令速度ＥＣＰｄを算出する部分とから構成される。なお、実行命令数は、評価プログラム１３１実行時にエミュレータ１１３等が計測するような構成としても良い。
　さらに、評価プログラム１３１は、命令履歴あるいはシステムコール履歴を記録する機能と、ユーザプログラム１３０が使用する各メモリ要素へのメモリアクセス量及びメモリアクセス時間を測定する機能とを備えても良い。
　一般的に、評価プログラム１３１は移行プログラム１３２毎に用意された方が、ＥＣＰｄを正確に測定出来る。しかし、複数のユーザプログラム１３０の命令実行特性が類似している場合、当該複数のユーザプログラム１３０が１つの評価プログラム１３１を共用しても良い。
　ユーザプログラム１３０のプログラムコード及びデータフォーマットは必ずしも移行元マシン２００の物理マシン１００上で動作するようにコンパイルされているとは限らない。移行元マシン２００は、ユーザプログラム１３０の移行に於いて、別の移行元マシン２００の中継装置として機能することもあるからである。
　移行元マシン２００の送出部１１８は、移行プログラム１３２及び評価プログラム１３１を、通信部を介してマイグレーション管理装置３００に送信する。
　なお、一台のコンピュータが、マイグレーション管理装置３００と移行元マシン２００の両者の構成を備えていても良い。即ち、或るマイグレーションにおいてマイグレーション管理装置３００（または移行元マシン２００）として機能したコンピュータが、別のマイグレーションにおいて移行元マシン２００（またはマイグレーション管理装置３００）として機能出来る構成にしても良い。
　また、マイグレーション管理装置３００の物理マシン１００の構成や種類は、必ずしも移行元マシン２００の物理マシン１００と同一であるとは限らない。
　マイグレーション管理装置３００のカーネルプログラム１１０は、物理マシン１００の備えるハードウェアのうち、カーネルプログラム１１０の管理下にあるハードウェアの構成情報を、カーネルテーブル（図３）としてメモリに保有している。
　図３は、カーネル表１１４を示している。カーネル表１１４は、マイグレーション管理装置３００のカーネルプログラム１１０が管理するＣＰＵ及びデバイス情報とメモリ資源情報を包含する。ＣＰＵ情報は、例えば物理マシン１００のネイティブな命令セットを特定するＩＳＡ（Ｍ）５０１、データがビックエンディアンかリトルエンディアンかを特定するＥｎｄｉａｎ（Ｍ）５０２である。二つの物理マシン１００のＩＳＡ（Ｍ）５０１及びＥｎｄｉａｎ（Ｍ）５０２の両者が一致すれば、当該マシン間のプログラム移行に際してＣＰＵエミュレーションは不要であるが、一方でも異なればＣＰＵエミュレーションが必要である。
　デバイス情報は、物理マシン１００が備えるデバイス構成を特定するＤｅｖｉｃｅ（Ｍ）５００である。Ｄｅｖｉｃｅ（Ｍ）５００は、例えば、デバイス構成毎に付与されたプラットフォーム名である。プラットフォーム名に対応する具体的なデバイス構成は、別途、カーネルプログラム１１０等で定義されている。Ｄｅｖｉｃｅ（Ｍ）５００が、直接、個々のデバイス情報を格納しても良い。二つの物理マシン１００のＤｅｖｉｃｅ（Ｍ）５００が一致すれば、当該マシン間のプログラム移行に際してデバイスエミュレーションは不要であるが、異なればデバイスエミュレーションが必要である。
　デバイスエミュレーションは、物理マシン１００が備えていないデバイスに対して入出力命令が発行等されたときに、当該デバイスの情報出力等をエミュレートすることである。例えば、ＣＰＵが、メモリに格納されたエミュレート情報をアクセスしてデバイス動作をエミュレートする。
　メモリ資源情報は、例えば、実装されているメモリのデータ転送速度であるＴｒａｎｓＲａｔｅ（Ｍ）５０３（メモリ速度）、メモリの空き容量を示すＦｒｅｅ５０４である。
　更にカーネル表１１４は、エミュレータ１１３がエミュレーションできるＣＰＵ及びデバイスを特定する情報であるＥｍｕｌａｔｉｏｎＣａｐａｂｉｌｉｔｙ５０５を包含する。ＥｍｕｌａｔｉｏｎＣａｐａｂｉｌｉｔｙ５０５は、例えば命令セット特定情報（ＩＳＡ）、データ形式特定情報（Ｅｎｄｉａｎ）及びデバイス構成特定情報（Ｄｅｖｉｃｅ）のリストである。エミュレータ１１３が、複数の命令セット、データ形式またはデバイス構成のエミュレーションを可能なとき、ＥｍｕｌａｔｉｏｎＣａｐａｂｉｌｉｔｙ５０５は複数個のリストを格納する。
　カーネル表１１４に格納されている情報は、カーネルプログラム１１０が自身の管理データや物理マシン１００等から取得して設定しても良いし、マイグレーション管理装置３００のシステム管理者がシステムパラメータとして設定しても良い。
　図３の例は、マイグレーション管理装置３００のカーネルプログラム１１０は４種類のメモリ要素を管理し、例えば１行目のメモリ要素は１２８００［ＭＢ／ｓ］の転送速度を持ち、現在２５６０［ＭＢ］の空き容量が存在していることを表している。
　また、図３のカーネル表１１４は、例えばこのカーネルプログラム１１０は、命令セットが“ＡＲＣＨ＿Ａ”、リトルエンディアンのＣＰＵと、名称”Ｐｌａｔｆｏｒｍ１”で特定されるデバイス構成を有する物理マシン１００で動作していることを表している。
　図４は、本発明の形態におけるプログラム構成表１２１を示している。プログラム構成表１２１は、ユーザプログラム１３０が使用中のメモリ資源情報及び動作可能な物理マシン１００のＣＰＵ及びデバイスの情報を包含する。
　ユーザプログラム１３０のマイグレーションを行う場合、資源割り当て部１１１は、プログラム構成表１２１の当該プログラム対応行とカーネル表１１４のＣＰＵ及びデバイス情報を比較して、移行元と移行先間でプログラムの互換性があるか否かを判断する。互換性を持たない場合、資源割り当て部１１１は、マイグレーションした当該プログラムをエミュレータ１１３上で実行する。
　プログラム構成表１２１は、ユーザプログラム１３０の識別子（プログラムＩＤ６０１）毎に、当該プログラムが動作可能な物理マシン１００のデバイス構成情報（Ｄｅｖｉｃｅ（Ｐ）６０７）を格納する。プログラム構成表１２１は、更に、プログラムＩＤ６０１毎に、当該プログラムが記述されている命令セットの特定情報（ＩＳＡ（Ｐ）６０２）、データ形式特定情報（Ｅｎｄｉａｎ（Ｐ）６０３）も包含する。
　プログラム構成表１２１は、更に、当該プログラムの命令及びデータ格納用に割り当てられているメモリのメモリ速度（ＴｒａｎｓＲａｔｅ（Ｐ）６０４）と割当量（Ａｓｓｉｇｎ６０５）を格納する。これらの情報は、通常、移行元マシン２００に関する情報であり、プログラム管理部１２０は、これらの情報を自装置のカーネルプログラム１１０から入力してプログラム構成表１２１に格納する。プログラム管理部１２０は、これらの情報を移行元マシン２００のシステム管理者から入力しても良い。更に、これらの情報が、ユーザプログラム１３０に記載されていて、プログラム管理部１２０がユーザプログラム１３０から読み取っても良い。
　プログラム構成表１２１は命令速度（Ｐ）６０６を格納していても良い。命令速度（Ｐ）６０６は、Ｄｅｖｉｃｅ（Ｐ）６０７、ＩＳＡ（Ｐ）６０２、Ｅｎｄｉａｎ（Ｐ）６０３、ＴｒａｎｓＲａｔｅ（Ｐ）６０４で特定される構成の物理マシン１００下における実効命令速度である。
　移行元マシン２００が当該構成を持つ物理マシン１００である場合、プログラム管理部１２０は移行元マシン２００で実行された評価プログラム１３１の出力値を入力して、命令速度（Ｐ）６０６に格納する。プログラム管理部１２０は、当該物理マシン１００のスペック情報や評価プログラム１３１の出力値を、移行元マシン２００のシステム管理者から入力して命令速度（Ｐ）６０６に格納しても良い。
　図４は、ユーザプログラム“ＵＩＤ１”が命令セット“ＡＲＣＨ＿Ｂ”でビッグエンディアンのＣＰＵと、デバイス構成”Ｐｌａｔｆｏｒｍ１”を持つ物理マシン１００で動作し、１２０ＭＢ／ｓの記憶装置を９６ＭＢ使用していることを例示している。
　プログラム管理部１２０及び送出部１１８は、回路によってハードウェア的に実現されても良いし、プログラムが計算機上あるいはカーネルプログラム１１０上で実行されることにより、ソフトウェア的に実現されても良い。
　図５は、本実施の形態のマイグレーション管理装置３００のマイグレーション動作のフローチャートである。以下の説明例に於いて、マイグレーション管理装置３００は、移行元マシン２００上で実行されているユーザプログラム１３０の移行プログラム１３２を、自装置へマイグレートする。
　マイグレーション管理装置３００の資源割り当て部１１１は、移行元マシン２００の送出部１１８から同マシン上で実行中の特定の移行プログラム１３２を指定したマイグレーション要求を受信してマイグレーション処理を開始する。同装置は、同装置に接続された利用者端末から、移行元マシン２００及び特定の移行プログラム１３２を指定したマイグレーション要求を受信しても良い。
　資源割り当て部１１１は、移行元マシン２００のプログラム管理部１２０から、プログラム構成表１２１に格納されている移行対象のユーザプログラム１３０の特性情報を取得する（Ｓ１０１）。ここで取得される特性情報は、Ｄｅｖｉｃｅ（Ｐ）６０７、ＩＳＡ（Ｐ）６０２、Ｅｎｄｉａｎ（Ｐ）６０３、Ａｓｓｉｇｎ６０５、命令速度（Ｐ）６０６である。取得した命令速度（Ｐ）６０６は、以降ＥＣＰｓとして参照される。
　資源割り当て部１１１は、取得したＤｅｖｉｃｅ（Ｐ）６０７、ＩＳＡ（Ｐ）６０２、Ｅｎｄｉａｎ（Ｐ）６０３と、自装置のカーネル表１１４のＤｅｖｉｃｅ（Ｍ）５００、ＩＳＡ（Ｍ）５０１、Ｅｎｄｉａｎ（Ｍ）５０２を比較する。一部でも一致しない場合、同部はマイグレーション管理装置３００がそのままユーザプログラム１３０のプログラムコードを実行できない（移行元と移行先とで命令等の互換性がない）と判定する（Ｓ１０２でＮｏ）。
　互換性がない場合、資源割り当て部１１１は、取得したＤｅｖｉｃｅ（Ｐ）６０７、ＩＳＡ（Ｐ）６０２、Ｅｎｄｉａｎ（Ｐ）６０３の対と一致する（Ｄｅｖｉｃｅ，ＩＳＡ，Ｅｎｄｉａｎ）リストがカーネル表１１４のＥｎｄｉａｎＣａｐａｂｉｌｉｔｙ５０５の中に発見できるか判定する。即ち、同部は、自装置が移行元のエミュレーションを行えるかを判定する（Ｓ１０３）。エミュレーションが行えない（発見できない）場合（Ｓ１０３でＮｏ）、同部は、マイグレーションは不可能であると判断し、その旨のエラー報告を出力して、マイグレーション処理を終了する（Ｓ１０８）。エラー報告の出力先は、マイグレーション要求の送信元（利用者端末または移行元マシン２００）等である。
　エミュレーションが可能である場合（Ｓ１０３でＹｅｓ）、同部は移行元マシン２００の送出部１１８を経由して、該当するユーザプログラム１３０の評価プログラム１３１取得し、エミュレータ１１３を用いて実行する（Ｓ１０４）。
　実行に先立ち、資源割り当て部１１１は、例えば、エミュレータ１１３の動作パラメータに、取得したＤｅｖｉｃｅ（Ｐ）６０７、ＩＳＡ（Ｐ）６０２、Ｅｎｄｉａｎ（Ｐ）６０３を設定する。これにより、エミュレータ１１３は、Ｄｅｖｉｃｅ（Ｐ）６０７、ＩＳＡ（Ｐ）６０２、Ｅｎｄｉａｎ（Ｐ）６０３のエミュレーションを行うようになる。資源割り当て部１１１は、複数のエミュレータ１１３の中から、取得されたＤｅｖｉｃｅ（Ｐ）６０７、ＩＳＡ（Ｐ）６０２、Ｅｎｄｉａｎ（Ｐ）６０３のエミュレーションを行うものを選択しても良い。
　同部は、評価プログラム１３１及びエミュレータ１１３をメモリアクセス速度が既知の値ＭＰｅのメモリにロードすると共に、データ領域としても当該速度のメモリを割り当てて、エミュレータ１１３でエミュレーションしながら評価プログラム１３１を実行する。資源割り当て部１１１は、プリエンプション等で評価プログラム１３１の実行が中断されないように実行する。
　実行された評価プログラム１３１は、評価用コードの実行に要した時間を、実行命令数で除して、実効命令速度ＥＣＰｄを算出し出力する。出力されたＥＣＰｄはメモリアクセス速度がＭＰｅである場合の実効命令速度である。
　資源割り当て部１１１の要求を受けて、性能算出部１１２は評価プログラム１３１が出力した実効命令速度ＥＣＰｄ及びメモリ速度ＭＰｅに基づいて、下記の式１等を用いて要求メモリ速度ＭＰｄを算出し、資源割り当て部１１１に出力する（Ｓ１０５）。

Figure JPOXMLDOC01-appb-M000001
　当該式は、ＥＣＰｄはＥＣＰｓに比べエミュレーション等に起因するメモリアクセス回数増により性能低下が生じること、及び他の要因による性能低下分は無視されうることを前提としている。この前提の下、性能算出部１１２は、式（１）を用いて当該性能低下を補償する為に要求される要求メモリ速度ＭＰｄを算出する。
　エミュレータ１１３の設計情報や過去の実績等により、エミュレーションによるメモリアクセス増割合が既知である場合、資源割り当て部１１１は評価プログラム１３１を実行せずＥＣＰｄの推定値を計算等しても良い。
　資源割り当て部１１１は、カーネル表１１４のＴｒａｎｓＲａｔｅ（Ｍ）５０３及びＦｒｅｅ５０４を参照し、自装置内に要求メモリ速度ＭＰｄを満たす記憶装置は十分な空き容量（取得したＡｓｓｉｇｎ６０５以上）を持って存在するか判定する（Ｓ１０６）。当該条件を満たすメモリ（要求メモリ）が存在しない場合（Ｓ１０６でＮｏ）、同部は、マイグレーションは不可能であると判断し、その旨のエラー報告を出力してマイグレーション処理を終了する（Ｓ１０９）。
　エラー報告を受信したマイグレーション要求の送信元（利用者端末または移行元マシン２００）は、例えば、マイグレーション管理システム４００内の別のマイグレーション管理装置３００を選択して、マイグレーション要求を送信し直す。
　要求メモリが存在する場合（Ｓ１０６でＹｅｓ）、資源割り当て部１１１は、移行元マシン２００の送出部１１８から移行プログラム１３２を取得し、要求メモリを割り当ててエミュレータ１１３上で移行プログラム１３２の実効を開始する。即ち、同部はユーザプログラム１３０のライブマイグレーションを実行する（Ｓ１０７）。
　なお、取得したＤｅｖｉｃｅ（Ｐ）６０７、ＩＳＡ（Ｐ）６０２、Ｅｎｄｉａｎ（Ｐ）６０３と、自装置のカーネル表１１４のＤｅｖｉｃｅ（Ｍ）５００、ＩＳＡ（Ｍ）５０１、Ｅｎｄｉａｎ（Ｍ）５０２が一致する場合（図５のｓ１０２でＹｅｓ）、資源割り当て部１１１は評価プログラム１３１を、エミュレータ１１３を使用せずに直接実行する。性能算出部１１２は、この結果から要求メモリ速度ＭＰｄを算出する。
　本実施の形態のマイグレーション管理装置３００は、以下の効果を達成できる。第一の効果は、マイグレーション管理装置３００及び移行元マシン２００が備える物理マシン１００の命令等の互換性の有無に依らず、後者から前者へマイグレーションされた移行プログラム１３２の性能低下を抑えることができることである。
　その理由は、エミュレーション等に起因して実効命令速度が低下するような場合でも、移行プログラム１３２に高速なメモリを割り当てることで、性能維持を図るからである。移行後、移行プログラム１３２実行中の命令及びデータアクセスが速くなり、このアクセス高速化が、結果的にエミュレーションによる性能低下を相殺する。
　第二の効果は、移行プログラム１３２の動作特性を反映したメモリ割り当てができることである。
　その理由は、移行対象の移行プログラム１３２が持つ、特徴的な処理を含んだ評価プログラム１３１を、移行先であるマイグレーション管理装置３００で事前に実行することで、移行後の当該プログラム実行時の性能低下を的確に予測するからである。
＜第二の実施形態＞
　次に、本発明の第二の実施形態について説明する。本実施の形態の性能算出部１１２は、要求メモリ速度ＭＰｄの算出（図５のＳ１０５）に於いて、ＣＰＵ負荷（Ｌ）を反映する計算式を用いる。
　マルチプロセッサシステム等に於いてシステム内の平均ＣＰＵ使用率が高い場合、複数のＣＰＵや入出力データ転送装置間等のメモリアクセス競合により、結果的に実効命令速度ＥＣＰｄが低下する。このような場合に評価プログラム１３１を実行して実効命令速度ＥＣＰｄを算出すると、低すぎる速度が測定されてしまう。
　このような測定誤算を補正するために、本実施の性能算出部１１２は、要求メモリ速度ＭＰｄの算出に於いて、ＣＰＵ負荷を反映する下記計算式を用いる。
Figure JPOXMLDOC01-appb-M000002
 　ここで、ＣＰＵ使用率Ｌはカーネルプログラム１１０の測定値であり０（０％を意味する）以上１（１００％を意味する）以下の実数である。ｆｕｎｃ（Ｌ）は、ｆｕｎｃ（０）＝１、ｆｕｎｃ（１）＝１／ａ（ａは１以上の実数）であって、ＣＰＵ使用率Ｌの増加に伴って増加する関数である。なお、ａはＣＰＵ使用率の増加によるＥＣＰｄの最大遅延倍率の見積もり値であり、システム管理者等により、パラメータとして性能算出部１１２に与えられる。
　性能算出部１１２は、ｆｕｎｃ（Ｌ）として様々な関数を使用するように構成できる。ＣＰＵ負荷の上昇に伴いメモリアクセス時間が指数関数的に増加するとの前提の下、性能算出部１１２は、例えば以下の式（３）をｆｕｎｃ（Ｌ）として使用するように構成され得る。
Figure JPOXMLDOC01-appb-M000003
 　ＣＰＵ負荷の上昇に伴いメモリアクセス時間が線形に増加するとの前提の下、性能算出部１１２は、例えば以下の式（４）をｆｕｎｃ（Ｌ）として使用するように構成され得る。
Figure JPOXMLDOC01-appb-M000004
 　性能算出部１１２は、ＣＰＵ負荷の上昇に伴うメモリ競合影響を実測した結果の近似式、例えば、ＣＰＵ負荷Ｌの高次関数を使用するように構成されても良い。
　なお、ＣＰＵ使用率が所定値以上の場合は、資源割り当て部１１１がＣＰＵ使用率の低下を待ち合わせてから、評価プログラム１３１を実行（Ｓ１０５）するようにしても良い。
　本実施形態のマイグレーション管理装置３００は、ＣＰＵ使用率が高い状況で実効命令速度ＥＣＰｄを測定した場合でも、通常の負荷状況で性能低下を防止するのに十分な要求メモリ速度ＭＰｄを算出できる。その理由は、性能算出部１１２が、測定時のＣＰＵ負荷に応じてＭＰｄを増加補正して算出するからである。
　以上、好ましい実施の形態を挙げて本発明を説明したが、本発明は必ずしも上記実施の形態に制限されるものではなく、その技術的思想の範囲内においてその形態を変化させて実施することができる。
　例えば、エミュレータ１１３は、ＣＰＵエミュレーションまたはデバイスエミュレーションの一方だけを実行するものであっても良い。この場合、カーネル表１１４及びプログラム構成表１２１は、Ｄｅｖｉｃｅ（Ｍ）５００、Ｄｅｖｉｃｅ（Ｐ）６０７またはＩＳＡ（Ｍ）５０１等、ＩＳＡ（Ｐ）６０２等を包含しなくても良い。
　更に、例えば、本発明はコード互換性がある物理マシン１００間のマイグレーションにも適用できる。このようなケースにおいて、本発明はマイクロコードの実装差等に起因して発生する実効命令速度差による性能低下を低減できる。
　また、本実施形態では移行元マシン２００がプログラム管理部１２０を備えているが、別途プログラム管理部１２０専用の物理マシン１００を用意し、これによって複数の移行元マシン２００の集中管理を行っても良い。
　さらに、本実施の形態における物理マシン１００と記載しているコンピュータが仮想計算機であっても良い。
＜第三の実施形態＞
　図６は、第３の実施形態のマイグレーション管理装置３００の構成を示す。同装置は、性能算出部１１２と資源割り当て部１１１を備える。性能算出部１１２は、自装置とネットワーク１４０で接続された移行元マシン２００上で動作する移行プログラム１３２に割り当てられたメモリの容量及び移行元マシン２００上における実効命令速度（ＥＣＰｓ）を入力する。同部は、ＥＣＰｓ及び自装置上に於ける移行プログラム１３２記述命令の実効命令速度（ＥＣＰｄ）及びＥＣＰｄの基礎となるメモリ速度（ＭＰｅ）に基づいて、要求メモリ速度（ＭＰｄ）を算出する。資源割り当て部１１１は、自装置へ移行される移行プログラム１３２に対しＭＰｄを満たし、かつ容量を持った自装置の備えるメモリ割り当てを行い、割り当てが出来ないときはエラー報告を出力する。
　マイグレーション管理装置３００は、命令実行性能差があるプラットフォーム間でプログラムのマイグレーションを行う際、マイグレーション前後の性能低下を低減することができる。その理由は、エミュレーション等に起因して実効命令速度が低下するような場合でも、移行プログラム１３２に高速なメモリを割り当てることで、性能維持を図るからである。
　以上、実施形態（及び実施例）を参照して本願発明を説明したが、本願発明は上記実施形態（及び実施例）に限定されものではない。本願発明の構成や詳細には、本願発明のスコープ内で当業者が理解し得る様々な変更をすることができる。
　この出願は、２０１０年１月１２日に出願された日本出願特願２０１０−００３８２９を基礎とする優先権を主張し、その開示の全てをここに取り込む。 <First embodiment>
FIG. 1 is a configuration diagram of a migration management system 400 according to this embodiment. The migration management system 400 includes one or more migration management apparatuses 300 and one or more migration source machines 200 connected via the network 140. The migration management apparatus 300 inputs and executes (migrates or migrates) a user program (migration program 132 in the user program 130) executed on the migration source machine 200.
The present invention prevents the deterioration of the instruction speed due to emulation or the like that occurs with migration by using a high-speed memory.
In FIG. 1, the migration management apparatus 300 includes a physical machine 100 and a kernel program 110. The migration source machine 200 includes a physical machine 100, a kernel program 110, a program management unit 120, and a user program 130.
The physical machine 100 is a computer that operates under program control, and includes a central processing unit (CPU), a main memory or a secondary memory (memory), a communication unit, an input / output device, peripheral devices, and the like.
The kernel program 110 is a program that operates on the physical machine 100 and has a function of allocating hardware resources of the physical machine 100 to each program executed on the physical machine 100. The kernel program 110 indicates an operating system (OS) kernel or the like. For example, the kernel program 110 indicates a virtual machine monitor or the like in a virtual computing environment.
The kernel program 110 of the migration management apparatus 300 includes a resource allocation unit 111, a performance calculation unit 112, and an emulator 113. The kernel program 110 of the migration management apparatus 300 holds a kernel table 114 describing the characteristics of the physical machine 100 in the memory. The kernel program 110 also has general OS functions other than those described above.
The resource allocation unit 111 allocates a migration memory to each user program 130 to be migrated.
The performance calculation unit 112 calculates a required memory speed (MPd) of the memory. This part is based on, for example, the effective instruction speed (ECPs) on the migration source machine 200, the effective instruction speed (ECPd) of the instruction describing the user program 130 in the own apparatus, and the memory speed (MPe) that is the basis of ECPd. Then, MPd is calculated. The memory speed (MPe) on which ECPd is based is the memory speed used for the calculation or measurement of ECPd.
Here, the memory speed indicates, for example, a memory bus width, a memory cycle time, a memory access time, or a data transfer speed (transfer rate) combining these. Hereinafter, the memory transfer rate is assumed to be the data transfer rate.
The effective instruction speed ECP indicates, for example, an instruction execution time. The execution command speed ECPd of the migration management apparatus 300 can be measured by executing the evaluation program 131 on the apparatus. If the instruction set or the mounted device (instruction or the like) is not compatible between the migration source machine 200 and the migration destination migration management apparatus 300 and the migrated program is executed on the emulator 113, the ECPd is a performance by emulation. The value reflects the decline.
The emulator 113 dynamically emulates this instruction and the like when the user program 130 described by the instruction of the migration source machine 200 and using the device of the machine cannot be directly executed by the migration management apparatus 300. Execute. The emulator 113 has, for example, a CPU state conversion function, a code conversion function, and a device emulation function.
The CPU state conversion function converts the CPU state of the migration source machine 200 to the state of the migration management apparatus 300. The code conversion function converts the instruction of the migration source machine 200 into the instruction of the migration management apparatus 300. The device emulation function imitates a device operation that the migration source machine 200 has and the migration management apparatus 300 does not have.
The resource allocation unit 111 and the performance calculation unit 112 may be realized as a hardware device controlled by the kernel program 110, or realized by the CPU of the physical machine 100 executing the migration management program 101 on the memory. May be.
The kernel program 110 of the migration source machine 200 includes a sending unit 118. In response to an instruction from the user or the like, the sending unit 118 requests the specific migration management apparatus 300 to migrate the migration program 132 executed on the migration source machine 200. The kernel program 110 of the migration source machine 200 also has a general OS function other than the above.
The program management unit 120 operates according to instructions from the kernel program 110. The same unit includes a CPU information management module for managing CPU information assigned to each user program 130 and a memory information management module for managing memory information assigned to each user program 130. Further, this unit includes a driver management module that manages information of the physical machine 100 on which each user program 130 operates. The program management unit 120 holds a program configuration table 121 for managing the above-described information on a memory.
The CPU information management module has a function of storing CPU architecture information of all user programs 130 running on the kernel program 110 in the program configuration table 121. The CPU architecture information includes CPU instruction set information, endian information, or extended instruction information.
The memory information management module has a function of storing the size and memory speed information of all storage device modules (hereinafter referred to as memory elements) used by the user program 130 running on the kernel program 110 in the program configuration table 121. .
The device driver management module has a function of storing platform information held by all user programs 130 running on the kernel program 110 in the program configuration table 121. The platform information includes hardware information of a virtual or real device operated by the user program 130.
FIG. 2 shows the configuration of the user program 130. The user program 130 is a program that operates on the physical machine 100 using the functions provided by the kernel program 110. The user program 130 includes a migration program 132 and an evaluation program 131.
The migration program 132 is an application program that does not have the authority to operate the kernel area, or a single-function application component in a distributed execution environment. The migration program 132 is a virtual machine in a virtual computing environment, for example. The migration program 132 is originally executed on the migration source machine 200, but is migrated to the migration management apparatus 300.
The evaluation program 131 is created by a user or the like reflecting, for example, a usage pattern such as an instruction used by the migration program 132 or an access pattern of a memory element. The evaluation program 131 is described by an instruction of a specific physical machine 100 such as the migration source machine 200. The evaluation program 131 includes, for example, an evaluation code section in which the number of execution instructions is known, an execution time measurement section for the evaluation code, an execution time per instruction from the measured execution time and the number of execution instructions, that is, an effective instruction speed ECPd. It is comprised from the part which calculates. The number of execution instructions may be configured such that the emulator 113 or the like measures when the evaluation program 131 is executed.
Further, the evaluation program 131 may have a function of recording an instruction history or a system call history, and a function of measuring a memory access amount and a memory access time for each memory element used by the user program 130.
Generally, the ECPd can be accurately measured when the evaluation program 131 is prepared for each migration program 132. However, when the instruction execution characteristics of the plurality of user programs 130 are similar, the plurality of user programs 130 may share one evaluation program 131.
The program code and data format of the user program 130 are not necessarily compiled to operate on the physical machine 100 of the migration source machine 200. This is because the migration source machine 200 may function as a relay device for another migration source machine 200 in the migration of the user program 130.
The sending unit 118 of the migration source machine 200 transmits the migration program 132 and the evaluation program 131 to the migration management apparatus 300 via the communication unit.
A single computer may have the configurations of both the migration management apparatus 300 and the migration source machine 200. That is, a computer that functions as the migration management apparatus 300 (or migration source machine 200) in a certain migration may be configured to function as the migration source machine 200 (or migration management apparatus 300) in another migration.
Further, the configuration and type of the physical machine 100 of the migration management apparatus 300 are not necessarily the same as the physical machine 100 of the migration source machine 200.
The kernel program 110 of the migration management apparatus 300 holds configuration information of hardware managed by the kernel program 110 among the hardware included in the physical machine 100 as a kernel table (FIG. 3) in a memory.
FIG. 3 shows the kernel table 114. The kernel table 114 includes CPU and device information and memory resource information managed by the kernel program 110 of the migration management apparatus 300. The CPU information is, for example, an ISA (M) 501 that specifies a native instruction set of the physical machine 100 and an Endian (M) 502 that specifies whether the data is big endian or little endian. If both ISA (M) 501 and Endian (M) 502 of the two physical machines 100 match, CPU emulation is not necessary for program transfer between the machines, but CPU emulation is necessary if they are different.
The device information is a Device (M) 500 that specifies a device configuration included in the physical machine 100. Device (M) 500 is a platform name assigned to each device configuration, for example. A specific device configuration corresponding to the platform name is separately defined in the kernel program 110 or the like. The Device (M) 500 may directly store individual device information. If the Device (M) 500 of the two physical machines 100 match, device emulation is not required for program transfer between the machines, but device emulation is required if they are different.
Device emulation is to emulate information output of a device when an input / output command is issued to a device that the physical machine 100 does not have. For example, the CPU emulates device operation by accessing emulation information stored in the memory.
The memory resource information is, for example, TransRate (M) 503 (memory speed) that is the data transfer speed of the mounted memory, and Free 504 that indicates the free capacity of the memory.
Further, the kernel table 114 includes an emulation capability 505 that is information for specifying a CPU and a device that can be emulated by the emulator 113. The emulation capability 505 is, for example, a list of instruction set identification information (ISA), data format identification information (Endian), and device configuration identification information (Device). When the emulator 113 can emulate a plurality of instruction sets, data formats, or device configurations, the EmulationCapability 505 stores a plurality of lists.
The information stored in the kernel table 114 may be acquired and set by the kernel program 110 from its own management data or the physical machine 100, or set by the system administrator of the migration management apparatus 300 as a system parameter. Also good.
In the example of FIG. 3, the kernel program 110 of the migration management apparatus 300 manages four types of memory elements. For example, the memory element in the first row has a transfer speed of 12800 [MB / s] and is currently 2560 [MB]. Indicates that there is free space.
3, for example, the kernel program 110 operates on a physical machine 100 having an instruction set “ARCH_A”, a little-endian CPU, and a device configuration specified by the name “Platform1”. Represents that.
FIG. 4 shows the program configuration table 121 in the embodiment of the present invention. The program configuration table 121 includes information on memory resources being used by the user program 130 and information on CPUs and devices of the operable physical machine 100.
When migrating the user program 130, the resource allocation unit 111 compares the program corresponding row of the program configuration table 121 with the CPU and device information of the kernel table 114, and the compatibility of the program between the migration source and the migration destination is confirmed. Judge whether there is. If there is no compatibility, the resource allocation unit 111 executes the migrated program on the emulator 113.
The program configuration table 121 stores, for each identifier (program ID 601) of the user program 130, device configuration information (Device (P) 607) of the physical machine 100 in which the program can operate. The program configuration table 121 further includes, for each program ID 601, identification information (ISA (P) 602) and data format identification information (Endian (P) 603) of an instruction set in which the program is described.
The program configuration table 121 further stores a memory speed (TransRate (P) 604) and an allocated amount (Assign 605) of a memory allocated for storing instructions and data of the program. These pieces of information are usually information about the migration source machine 200, and the program management unit 120 inputs these pieces of information from the kernel program 110 of the own apparatus and stores them in the program configuration table 121. The program management unit 120 may input such information from the system administrator of the migration source machine 200. Further, these pieces of information may be described in the user program 130 and read by the program management unit 120 from the user program 130.
The program configuration table 121 may store an instruction speed (P) 606. The instruction speed (P) 606 is an effective instruction speed under the physical machine 100 having a configuration specified by Device (P) 607, ISA (P) 602, Endian (P) 603, and TransRate (P) 604.
When the migration source machine 200 is the physical machine 100 having the configuration, the program management unit 120 inputs the output value of the evaluation program 131 executed on the migration source machine 200 and stores it in the instruction speed (P) 606. The program management unit 120 may input the specification information of the physical machine 100 and the output value of the evaluation program 131 from the system administrator of the migration source machine 200 and store them in the command speed (P) 606.
FIG. 4 shows that the user program “UID1” operates on the big endian CPU with the instruction set “ARCH_B” and the physical machine 100 having the device configuration “Platform1”, and uses 96 MB of a storage device of 120 MB / s. Illustrated.
The program management unit 120 and the sending unit 118 may be realized by hardware by a circuit, or may be realized by software by executing a program on the computer or the kernel program 110.
FIG. 5 is a flowchart of the migration operation of the migration management apparatus 300 of this embodiment. In the following description example, the migration management apparatus 300 migrates the migration program 132 of the user program 130 executed on the migration source machine 200 to the own apparatus.
The resource allocation unit 111 of the migration management apparatus 300 receives a migration request specifying a specific migration program 132 being executed on the same machine from the sending unit 118 of the migration source machine 200 and starts the migration process. The apparatus may receive a migration request specifying the migration source machine 200 and a specific migration program 132 from a user terminal connected to the apparatus.
The resource allocation unit 111 acquires the characteristic information of the migration target user program 130 stored in the program configuration table 121 from the program management unit 120 of the migration source machine 200 (S101). The characteristic information acquired here is Device (P) 607, ISA (P) 602, Endian (P) 603, Assign 605, and command speed (P) 606. The acquired command speed (P) 606 is referred to as ECPs hereinafter.
The resource allocation unit 111 acquires the acquired Device (P) 607, ISA (P) 602, Endian (P) 603, and Device (M) 500, ISA (M) 501, Endian (M) of the kernel table 114 of the own device. Compare 502. When even a part does not match, the same part determines that the migration management apparatus 300 cannot execute the program code of the user program 130 as it is (the instruction is not compatible between the migration source and the migration destination) (No in S102).
If they are not compatible, the resource allocation unit 111 has a list (Device, ISA, Endian) that matches the acquired pair of Device (P) 607, ISA (P) 602, and Endian (P) 603 (EndianCapability 505 in the kernel table 114). Determine if you can find it. That is, the same part determines whether or not the own apparatus can perform the migration source emulation (S103). When the emulation cannot be performed (cannot be found) (No in S103), the same part determines that migration is impossible, outputs an error report to that effect, and ends the migration process (S108). The output destination of the error report is a migration request transmission source (user terminal or migration source machine 200) or the like.
If emulation is possible (Yes in S103), the same unit obtains the evaluation program 131 of the corresponding user program 130 via the sending unit 118 of the migration source machine 200, and executes it using the emulator 113 (S104). .
Prior to execution, the resource allocation unit 111 sets, for example, the acquired Device (P) 607, ISA (P) 602, and Endian (P) 603 as operation parameters of the emulator 113. As a result, the emulator 113 performs emulation of Device (P) 607, ISA (P) 602, and Endian (P) 603. The resource allocating unit 111 may select one that emulates the acquired Device (P) 607, ISA (P) 602, and Endian (P) 603 from among a plurality of emulators 113.
The same unit loads the evaluation program 131 and the emulator 113 into a memory having a known memory access speed value MPe, allocates a memory having the speed as a data area, and executes the evaluation program 131 while emulating the emulator 113. . The resource allocation unit 111 executes the execution of the evaluation program 131 so as not to be interrupted by preemption or the like.
The executed evaluation program 131 calculates and outputs the effective instruction speed ECPd by dividing the time required for executing the evaluation code by the number of executed instructions. The output ECPd is an effective instruction speed when the memory access speed is MPe.
In response to the request from the resource allocation unit 111, the performance calculation unit 112 calculates the required memory speed MPd using the following equation 1 based on the effective instruction speed ECPd and the memory speed MPe output from the evaluation program 131, and the resource It outputs to the allocation part 111 (S105).
Figure JPOXMLDOC01-appb-M000001
 This equation is based on the premise that the performance degradation of ECPd occurs due to an increase in the number of memory accesses caused by emulation and the like, and the performance degradation due to other factors can be ignored compared to ECPs. Under this assumption, the performance calculation unit 112 calculates the required memory speed MPd required to compensate for the performance degradation using the equation (1).
If the memory access increase rate by emulation is known from the design information of the emulator 113 or past results, the resource allocation unit 111 may calculate the estimated value of ECPd without executing the evaluation program 131.
The resource allocation unit 111 refers to the TransRate (M) 503 and the Free 504 in the kernel table 114, and whether the storage device satisfying the requested memory speed MPd exists in the own device with sufficient free space (the acquired Assign 605 or higher). Determine (S106). If there is no memory (request memory) that satisfies the condition (No in S106), the same part determines that migration is impossible, outputs an error report to that effect, and ends the migration process (S109). .
The source (user terminal or migration source machine 200) of the migration request that received the error report selects another migration management device 300 in the migration management system 400, for example, and resends the migration request.
If the requested memory exists (Yes in S106), the resource allocation unit 111 acquires the migration program 132 from the sending unit 118 of the migration source machine 200, allocates the requested memory, and starts executing the migration program 132 on the emulator 113. To do. That is, the same part performs live migration of the user program 130 (S107).
The acquired Device (P) 607, ISA (P) 602, and Endian (P) 603 match the Device (M) 500, ISA (M) 501, and Endian (M) 502 in the kernel table 114 of the own device. In the case (Yes in s102 of FIG. 5), the resource allocation unit 111 directly executes the evaluation program 131 without using the emulator 113. The performance calculation unit 112 calculates the required memory speed MPd from this result.
The migration management apparatus 300 of this embodiment can achieve the following effects. The first effect is that the performance degradation of the migration program 132 migrated from the latter to the former can be suppressed regardless of whether the instructions of the physical machine 100 included in the migration management apparatus 300 and the migration source machine 200 are compatible. It is.
The reason is that even when the effective instruction speed is reduced due to emulation or the like, performance is maintained by allocating a high-speed memory to the migration program 132. After the migration, the instruction and data access during execution of the migration program 132 becomes faster, and this speeding up of the access offsets the performance degradation due to emulation.
The second effect is that memory allocation reflecting the operation characteristics of the migration program 132 can be performed.
This is because the migration program 132 to be migrated has an evaluation program 131 including characteristic processing executed in advance by the migration management apparatus 300 that is the migration destination, so that the performance at the time of execution of the program after migration is achieved. This is because the decrease is accurately predicted.
<Second Embodiment>
Next, a second embodiment of the present invention will be described. The performance calculation unit 112 of the present embodiment uses a calculation formula that reflects the CPU load (L) in the calculation of the required memory speed MPd (S105 in FIG. 5).
In a multiprocessor system or the like, when the average CPU usage rate in the system is high, the effective instruction speed ECPd decreases as a result of memory access contention among a plurality of CPUs and input / output data transfer devices. In such a case, if the evaluation program 131 is executed to calculate the effective instruction speed ECPd, a speed that is too low is measured.
In order to correct such a measurement miscalculation, the performance calculation unit 112 of the present embodiment uses the following calculation formula that reflects the CPU load in calculating the required memory speed MPd.
Figure JPOXMLDOC01-appb-M000002
 Here, the CPU usage rate L is a measured value of the kernel program 110, and is a real number between 0 (meaning 0%) and 1 (meaning 100%). func (L) is func (0) = 1, func (1) = 1 / a (a is a real number of 1 or more), and is a function that increases as the CPU usage rate L increases. Note that a is an estimated value of the maximum delay magnification of ECPd due to an increase in the CPU usage rate, and is given to the performance calculation unit 112 as a parameter by the system administrator or the like.
The performance calculation unit 112 can be configured to use various functions as func (L). Under the assumption that the memory access time increases exponentially as the CPU load increases, the performance calculation unit 112 may be configured to use, for example, the following equation (3) as func (L).
Figure JPOXMLDOC01-appb-M000003
 Under the assumption that the memory access time increases linearly as the CPU load increases, the performance calculation unit 112 can be configured to use, for example, the following equation (4) as func (L).
Figure JPOXMLDOC01-appb-M000004
 The performance calculation unit 112 may be configured to use an approximate expression obtained as a result of actual measurement of the memory contention effect accompanying an increase in CPU load, for example, a higher-order function of the CPU load L.
If the CPU usage rate is greater than or equal to a predetermined value, the resource allocation unit 111 may wait for the CPU usage rate to decrease before executing the evaluation program 131 (S105).
The migration management apparatus 300 of the present embodiment can calculate the required memory speed MPd sufficient to prevent performance degradation under normal load conditions even when the effective instruction speed ECPd is measured in a situation where the CPU usage rate is high. The reason is that the performance calculation unit 112 calculates MPd by increasing the MPd according to the CPU load at the time of measurement.
The present invention has been described above with reference to preferred embodiments. However, the present invention is not necessarily limited to the above-described embodiments, and can be carried out by changing the form within the scope of the technical idea. it can.
For example, the emulator 113 may execute only one of CPU emulation and device emulation. In this case, the kernel table 114 and the program configuration table 121 may not include the Device (M) 500, the Device (P) 607, the ISA (M) 501, or the like, the ISA (P) 602, or the like.
Further, for example, the present invention can be applied to migration between physical machines 100 having code compatibility. In such a case, the present invention can reduce performance degradation due to an effective instruction speed difference caused by a microcode mounting difference or the like.
In the present embodiment, the migration source machine 200 includes the program management unit 120. However, even if a physical machine 100 dedicated to the program management unit 120 is prepared separately and centralized management of a plurality of migration source machines 200 is thereby performed. good.
Furthermore, the computer described as the physical machine 100 in the present embodiment may be a virtual computer.
<Third embodiment>
FIG. 6 shows the configuration of the migration management apparatus 300 of the third embodiment. The apparatus includes a performance calculation unit 112 and a resource allocation unit 111. The performance calculation unit 112 inputs the memory capacity allocated to the migration program 132 operating on the migration source machine 200 connected to the own apparatus through the network 140 and the effective instruction speed (ECPs) on the migration source machine 200. The same part calculates the required memory speed (MPd) based on the ECPs and the effective instruction speed (ECPd) of the migration program 132 description instruction on its own device and the memory speed (MPe) that is the basis of the ECPd. The resource allocation unit 111 allocates the memory included in the own device that satisfies MPd and has the capacity for the migration program 132 to be migrated to the own device, and outputs an error report when the allocation cannot be performed.
The migration management apparatus 300 can reduce performance degradation before and after migration when migrating a program between platforms having instruction execution performance differences. The reason is that even when the effective instruction speed is reduced due to emulation or the like, performance is maintained by allocating a high-speed memory to the migration program 132.
Although the present invention has been described with reference to the embodiments (and examples), the present invention is not limited to the above embodiments (and examples). Various changes that can be understood by those skilled in the art can be made to the configuration and details of the present invention within the scope of the present invention.
This application claims the priority on the basis of Japanese application Japanese Patent Application No. 2010-003829 for which it applied on January 12, 2010, and takes in those the indications of all here.

Claims (13)

　自装置とネットワークで接続された移行元マシン上で動作する移行プログラムに割り当てられたメモリの容量及び前記移行元マシン上における実効命令速度（ＥＣＰｓ）を入力して、前記ＥＣＰｓ及び自装置上に於ける前記移行プログラム記述命令の実効命令速度（ＥＣＰｄ）及び前記ＥＣＰｄの基礎となるメモリ速度（ＭＰｅ）に基づいて、要求メモリ速度（ＭＰｄ）を算出する性能算出手段と、
　自装置へ移行される前記移行プログラムに対し前記ＭＰｄを満たし、かつ前記容量を持った自装置の備えるメモリ割り当てを行い、前記割り当てが出来ないときはエラー報告を出力する資源割り当て手段を備えるマイグレーション管理装置。 Input the capacity of the memory allocated to the migration program running on the migration source machine connected to the local device and the network, and the effective instruction speed (ECPs) on the migration source machine, and enter the ECPs and the local device. Performance calculating means for calculating a required memory speed (MPd) based on an effective instruction speed (ECPd) of the migration program description instruction and a memory speed (MPe) as a basis of the ECPd;
Migration management comprising resource allocation means for allocating memory included in the own device having the MPd and having the capacity for the migration program to be migrated to the own device and outputting an error report when the allocation is not possible apparatus. 　複数の移行プログラムのおのおの対応に定義された評価プログラムを格納する前記移行元マシンから、前記移行プログラム対応の評価プログラムを入力して、自装置上で前記評価プログラムを実行して、実行結果から前記ＥＣＰｄを算出し、当該ＥＣＰｄ及び前記評価プログラムに割り当てたメモリ速度から前記ＭＰｄを算出する前記性能算出手段を備える、請求項１のマイグレーション管理装置。 From the migration source machine that stores an evaluation program defined for each of a plurality of migration programs, the migration program compatible evaluation program is input, the evaluation program is executed on its own device, The migration management apparatus according to claim 1, further comprising: the performance calculation unit that calculates the ECPd and calculates the MPd from the memory speed assigned to the ECPd and the evaluation program. 　自装置上で、自装置の命令と異なる前記移行元マシンの命令で記述された前記移行プログラムと前記評価プログラムを実行することができるエミュレータと、
　前記エミュレータ上で前記評価プログラムを実行して、実行結果から前記ＥＣＰｄを算出する前記性能算出手段を備える、請求項２のマイグレーション管理装置。 On the own device, an emulator capable of executing the migration program described in the migration source machine instruction different from the instruction of the own device and the evaluation program;
The migration management apparatus according to claim 2, further comprising the performance calculation unit that executes the evaluation program on the emulator and calculates the ECPd from an execution result. 　前記評価プログラム実行時のＣＰＵ負荷に応じて、前記ＭＰｄを増加補正して算出する前記性能評価手段を備える、請求項２または３のマイグレーション管理装置。 4. The migration management apparatus according to claim 2, further comprising the performance evaluation unit that calculates the MPd by increasing the MPd according to a CPU load when the evaluation program is executed. 　ネットワークで接続された複数の請求項１乃至４のいずれかのマイグレーション管理装置と、一の前記マイグレーション管理装置から前記エラー報告を受信すると、他の前記マイグレーション管理装置に移行要求を送信する前記移行元マシンを包含するマイグレーション管理システム。 The migration source that transmits a migration request to another migration management device when receiving the error report from a plurality of migration management devices according to any one of claims 1 to 4 connected via a network and the one migration management device A migration management system that includes machines. 　コンピュータに、自装置とネットワークで接続された移行元マシン上で動作する移行プログラムに割り当てられたメモリの容量及び前記移行元マシン上における実効命令速度（ＥＣＰｓ）を入力して、前記ＥＣＰｓ及び自装置上に於ける前記移行プログラム記述命令の実効命令速度（ＥＣＰｄ）及び前記ＥＣＰｄの基礎となるメモリ速度（ＭＰｅ）に基づいて、要求メモリ速度（ＭＰｄ）を算出する性能算出処理と、
　自装置へ移行される前記移行プログラムに対し前記ＭＰｄを満たし、かつ前記容量を持った自装置の備えるメモリ割り当てを行い、前記割り当てが出来ないときはエラー報告を出力する資源割り当て処理を実行させるマイグレーション管理プログラムを格納するコンピュータ読み取り可能な記録媒体。 The computer inputs the memory capacity allocated to the migration program operating on the migration source machine connected to the own device via the network and the effective command speed (ECPs) on the migration source machine, and the ECPs and the own device. A performance calculation process for calculating a required memory speed (MPd) based on an effective instruction speed (ECPd) of the transition program description instruction and a memory speed (MPe) as a basis of the ECPd;
Migration for executing a resource allocation process that performs memory allocation for the migration program that is migrated to the own device and that satisfies the MPd and has the capacity and that outputs an error report when the allocation is not possible A computer-readable recording medium for storing a management program. 　前記コンピュータに、複数の移行プログラムのおのおの対応に定義された評価プログラムを格納する前記移行元マシンから、前記移行プログラム対応の評価プログラムを入力して、自装置上で前記評価プログラムを実行して、実行結果から前記ＥＣＰｄを算出し、当該ＥＣＰｄ及び前記評価プログラムに割り当てたメモリ速度から前記ＭＰｄを算出する前記性能算出処理を実行させる、前記マイグレーション管理プログラムを格納する請求項６のコンピュータ読み取り可能な記録媒体。 From the migration source machine that stores an evaluation program defined for each of a plurality of migration programs in the computer, the evaluation program corresponding to the migration program is input, and the evaluation program is executed on its own device, The computer-readable record according to claim 6, wherein the migration management program for calculating the ECPd from an execution result and executing the performance calculation process for calculating the MPd from the memory speed assigned to the ECPd and the evaluation program is stored. Medium. 　自装置上で、自装置の命令と異なる前記移行元マシンの命令で記述された前記移行プログラムと前記評価プログラムを実行することができるエミュレータを備えた前記コンピュータに、
　前記エミュレータ上で前記評価プログラムを実行して、実行結果から前記ＥＣＰｄを算出する前記性能算出処理を実行させる、前記マイグレーション管理プログラムを格納する請求項７のコンピュータ読み取り可能な記録媒体。 On the own device, the computer having an emulator capable of executing the migration program and the evaluation program described by the instructions of the migration source machine different from the instructions of the own device;
8. The computer-readable recording medium according to claim 7, wherein the migration management program for executing the performance calculation processing for executing the performance calculation processing for calculating the ECPd from an execution result by executing the evaluation program on the emulator. 　前記コンピュータに、前記評価プログラム実行時のＣＰＵ負荷に応じて、前記ＭＰｄを増加補正して算出する前記性能評価処理を実行させる、前記マイグレーション管理プログラムを格納する請求項７または８のコンピュータ読み取り可能な記録媒体。 The computer-readable computer program product according to claim 7 or 8, wherein the migration management program for causing the computer to execute the performance evaluation process for calculating the MPd by increasing the MPd according to a CPU load when the evaluation program is executed is stored. recoding media. 　コンピュータが、自装置とネットワークで接続された移行元マシン上で動作する移行プログラムに割り当てられたメモリの容量及び前記移行元マシン上における実効命令速度（ＥＣＰｓ）を入力して、前記ＥＣＰｓ及び自装置上に於ける前記移行プログラム記述命令の実効命令速度（ＥＣＰｄ）及び前記ＥＣＰｄの基礎となるメモリ速度（ＭＰｅ）に基づいて、要求メモリ速度（ＭＰｄ）を算出し、
　自装置へ移行される前記移行プログラムに対し前記ＭＰｄを満たし、かつ前記容量を持った自装置の備えるメモリ割り当てを行い、前記割り当てが出来ないときはエラー報告を出力するマイグレーション管理方法。 A computer inputs a capacity of a memory allocated to a migration program operating on a migration source machine connected to the device itself via a network and an effective instruction speed (ECPs) on the migration source machine, and the ECPs and the device itself Calculate the required memory speed (MPd) based on the effective instruction speed (ECPd) of the transition program description instruction above and the memory speed (MPe) that is the basis of the ECPd,
A migration management method for allocating a memory included in a self-device satisfying the MPd and having the capacity for the migration program to be migrated to the self-device, and outputting an error report when the assignment is not possible. 　前記コンピュータが、複数の移行プログラムのおのおの対応に定義された評価プログラムを格納する前記移行元マシンから、前記移行プログラム対応の評価プログラムを入力して、自装置上で前記評価プログラムを実行して、実行結果から前記ＥＣＰｄを算出し、当該ＥＣＰｄ及び前記評価プログラムに割り当てたメモリ速度から前記ＭＰｄを算出する、請求項６のマイグレーション管理方法。 The computer inputs an evaluation program corresponding to the migration program from the migration source machine storing an evaluation program defined in correspondence with each of a plurality of migration programs, and executes the evaluation program on its own device. The migration management method according to claim 6, wherein the ECPd is calculated from an execution result, and the MPd is calculated from a memory speed assigned to the ECPd and the evaluation program. 　自装置上で、自装置の命令と異なる前記移行元マシンの命令で記述された前記移行プログラムと前記評価プログラムを実行することができるエミュレータを備えた前記コンピュータが、
　前記エミュレータ上で前記評価プログラムを実行して、実行結果から前記ＥＣＰｄを算出する、請求項７のマイグレーション管理方法。 The computer having an emulator capable of executing the migration program and the evaluation program described in the instructions of the migration source machine different from the instructions of the own device on the own device,
The migration management method according to claim 7, wherein the evaluation program is executed on the emulator, and the ECPd is calculated from an execution result. 　前記コンピュータが、前記評価プログラム実行時のＣＰＵ負荷に応じて、前記ＭＰｄを増加補正して算出する、請求項７または８のマイグレーション管理方法。 The migration management method according to claim 7 or 8, wherein the computer calculates the MPd by increasing the MPd according to a CPU load when the evaluation program is executed.
PCT/JP2010/073071 2010-01-12 2010-12-15 Migration management device, migration management system, migration management method, and migration management program Ceased WO2011086824A1 (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
JP2011549894A JPWO2011086824A1 (en) 2010-01-12 2010-12-15 Migration management apparatus, migration management system, migration management method, and migration management program

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
JP2010003829 2010-01-12
JP2010-003829 2010-01-12

Publications (1)

Publication Number Publication Date
WO2011086824A1 true WO2011086824A1 (en) 2011-07-21

Family

ID=44304112

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/JP2010/073071 Ceased WO2011086824A1 (en) 2010-01-12 2010-12-15 Migration management device, migration management system, migration management method, and migration management program

Country Status (2)

Country Link
JP (1) JPWO2011086824A1 (en)
WO (1) WO2011086824A1 (en)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2005234917A (en) * 2004-02-20 2005-09-02 Hitachi Ltd Server determination method in case of failure
JP2007257097A (en) * 2006-03-22 2007-10-04 Nec Corp Virtual computer system and method and program for reconfiguring physical resource thereof
JP2007323245A (en) * 2006-05-31 2007-12-13 Nec Corp Computer system, performance measurement method, and management server device
JP2009537894A (en) * 2006-05-18 2009-10-29 マイクロソフト コーポレーション Deploy virtual machines to hosts based on workload characteristics

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2005234917A (en) * 2004-02-20 2005-09-02 Hitachi Ltd Server determination method in case of failure
JP2007257097A (en) * 2006-03-22 2007-10-04 Nec Corp Virtual computer system and method and program for reconfiguring physical resource thereof
JP2009537894A (en) * 2006-05-18 2009-10-29 マイクロソフト コーポレーション Deploy virtual machines to hosts based on workload characteristics
JP2007323245A (en) * 2006-05-31 2007-12-13 Nec Corp Computer system, performance measurement method, and management server device

Also Published As

Publication number Publication date
JPWO2011086824A1 (en) 2013-05-16

Similar Documents

Publication Publication Date Title
US10768960B2 (en) Method for affinity binding of interrupt of virtual network interface card, and computer device
US9378032B2 (en) Information processing method, information processing apparatus, recording medium, and system
US10897428B2 (en) Method, server system and computer program product for managing resources
US9081612B2 (en) Virtual machine control method and virtual machine
US8725912B2 (en) Dynamic balancing of IO resources on NUMA platforms
JP4557178B2 (en) Virtual machine management system, method and program thereof
US9229751B2 (en) Apparatus and method for managing virtual memory
US9934062B2 (en) Technologies for dynamically allocating hardware acceleration units to process data packets
KR102640232B1 (en) Method and apparatus for allocating resources in virtual environment
US10169102B2 (en) Load calculation method, load calculation program, and load calculation apparatus
WO2015196931A1 (en) Disk io-based virtual resource allocation method and device
CN118034917A (en) PCIe resource allocation method and device, electronic equipment and storage medium
US12141613B2 (en) Resource management for preferred applications
US9507624B2 (en) Notification conversion program and notification conversion method
WO2014118969A1 (en) Virtual computer system and data transfer control method for virtual computer system
CN110750336A (en) OpenStack virtual machine memory hot-expanding method
CN110162396A (en) Method for recovering internal storage, device, system and storage medium
WO2023173961A1 (en) Memory allocation method and related product
CN114546587A (en) A method for expanding and shrinking capacity of online image recognition service and related device
KR101924467B1 (en) System and method of resource allocation scheme for cpu and block i/o performance guarantee of virtual machine
WO2017166997A1 (en) Inic-side exception handling method and device
CN106330595B (en) Heartbeat detection method and device for distributed platform
CN109062702B (en) Computing resource allocation method, related device and readable storage medium
JP2010026828A (en) Method for controlling virtual computer
WO2011086824A1 (en) Migration management device, migration management system, migration management method, and migration management program

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 10843177

Country of ref document: EP

Kind code of ref document: A1

WWE Wipo information: entry into national phase

Ref document number: 2011549894

Country of ref document: JP

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 10843177

Country of ref document: EP

Kind code of ref document: A1