JP6024148B2 - Program placement method - Google Patents

Description

  The present invention relates to a program arrangement method in a computer system in which a plurality of computers are connected so as to communicate with each other through a network.

  Conventionally, a virtual machine (VM) technology in which hardware resources of a physical machine are divided into a plurality of pieces using software and independent, and a plurality of virtual machines are simultaneously operated on one physical machine is known. Yes. In addition, a technique has been proposed in which a virtual machine is executed in multiple on a plurality of physical machines to provide fault tolerance (FT) for hardware resources.

  For example, Patent Document 1 describes an FT system using a high-speed checkpoint method. In this FT system, a virtual machine (active side) is executed on one physical machine while taking a snapshot (memory state, disk state, processor and I / O context) at regular checkpoints, Is duplicated by transferring to the physical machine and saving it on the virtual machine (standby side). If the active virtual machine can no longer execute the business process due to a physical machine failure, failover (switch from standby to active) is performed, and the standby virtual machine becomes the active virtual machine. Take over the business process that was being executed in step 3, and execute it.

  In addition, as a method for optimally arranging a plurality of virtual machines on a plurality of physical machines, the following method has been proposed.

  For example, Patent Document 2 describes a method of optimally arranging a plurality of virtual machines in a plurality of physical machines based on the storage capacities of the physical machine and the virtual machine. Specifically, each of the plurality of virtual machines is divided into a plurality of physical machines based on the actual measurement data indicating the performance of each virtual machine every predetermined time, the storage capacity of each physical machine, and the storage capacity of each virtual machine. Calculate the combination of the virtual machine and physical machine, that is, the placement of the virtual machine on the physical machine, so that the total of the values indicating the performance of each virtual machine at each time when operating at any of the above is maximized . Then, according to the calculated arrangement, the files of each virtual machine are stored in the storage area of the physical machine corresponding to the virtual machine and the virtual machines are rearranged.

  Patent Document 3 describes a method of relocating a virtual machine on a physical machine affected by a disaster to another physical machine not affected by the disaster in order to avoid the influence of the disaster.

  Further, Patent Document 4 describes a method of arranging a plurality of virtual machines on a plurality of physical machines based on a network usage rate. Specifically, based on the measurement results of traffic between virtual machines, the expected value of the network usage rate when a given virtual machine is moved to another physical machine is calculated to determine the validity of the movement. Move virtual machines to other physical machines when appropriate.

  Patent Document 5 describes a method of arranging a plurality of virtual machines on a plurality of physical machines based on the loads of the physical machine and the virtual machine. Specifically, by grasping or analyzing changes in the load of physical machines and virtual machines using approximate equations such as higher-order equations, it is possible to detect in advance the shortage of physical machine resources, and to make appropriate virtual Realize machine placement and movement.

JP 2010-160660 A JP 2005-115653 A JP2011-209811A JP 2011-180889 A JP 2010-117760 A

  In a set of virtual machines (active side virtual machine and standby side virtual machine) converted to FT by the fast checkpoint method, the active side virtual machine has a higher load than the standby side virtual machine, and the standby side virtual machine The amount of communication from the active side virtual machine to the standby side virtual machine is larger than the amount of communication from the virtual machine to the active side virtual machine.

  For this reason, when arranging a plurality of virtual machines on a plurality of physical machines, it is necessary to determine the arrangement of the virtual machines in consideration of the above-described characteristics. However, in the case of an FT system, when a failure occurs in any physical machine, the standby virtual machine on the other physical machine performs the business processing performed on the active virtual machine on the physical machine. Failover takes over. For this reason, when a failover occurs, the processing load of a plurality of physical machines and the distribution state of the network communication load among the physical machines collapse, and the execution efficiency of the entire system decreases. Although it may be possible to perform the optimal placement calculation and relocation of the virtual machine again when a failover occurs, the result is that the active virtual machine that is running is moved to another physical machine, resulting in a running operation. There is concern about the impact on

  The object of the present invention is to solve the above-described problem, that is, when the failover occurs, the processing load of a plurality of physical machines and the distributed state of the network communication load among the physical machines are disrupted and the execution efficiency of the entire system is lowered. The object is to provide a program placement method to be solved.

A program arrangement method according to an aspect of the present invention includes:
A program arrangement method executed by a computer system in which a first computer and a plurality of second computers are connected to each other through a network,
The first computer is an arrangement pattern in which a plurality of program sets, each of which is a set of two programs that operate as one active and the other as a standby, are arranged in the plurality of second computers, The two programs are not arranged on the same second computer, and an arrangement pattern is determined in consideration of at least one of the processing load of each of the plurality of second computers and the communication load between the plurality of second computers. And
The first computer arranges the plurality of program sets on the plurality of second computers according to the determined arrangement pattern,
When a failure occurs in any one of the second computers and the first computer fails over, the active and standby of the two programs in the remaining program set that has not failed over A switching pattern for switching between the plurality of second computers after the failover is determined in consideration of at least one of the processing load of each of the plurality of second computers and the communication load between the plurality of second computers,
The first computer adopts a configuration in which the two programs of the program set that are not failed over are switched between active and standby according to the determined switching pattern.

A computer according to another embodiment of the present invention includes:
The first computer in a computer system in which a first computer and a plurality of second computers are connected to each other through a network,
An arrangement pattern in which a plurality of program sets, each of which is a set of two programs operating as one active and the other as a standby, are arranged in the plurality of second computers, and the two programs in the same program set are the same An arrangement pattern determining means for determining an arrangement pattern that is not arranged in the second computer and takes into account at least one of the processing load of each of the plurality of second computers and the communication load between the plurality of second computers;
Program placement means for placing the plurality of program sets on the plurality of second computers according to the determined placement pattern;
When a failure occurs in any of the second computers and a failover is performed, a switching pattern for switching between active and standby of the two programs of the remaining program set that has not been failed over, Switching pattern determining means for determining in consideration of at least one of the processing load of each of the plurality of second computers after the failover and the communication load between the plurality of second computers;
According to the determined switching pattern, a configuration is adopted in which switching means for switching between active and standby of the two programs of the program set in which the failover is not performed is employed.

  Since the present invention has the above-described configuration, even when a failover occurs, it is possible to prevent the processing load of a plurality of physical machines and the distribution state of network communication load among physical machines from being disrupted to reduce the execution efficiency of the entire system. can do.

It is a block diagram of a 1st embodiment of the present invention. It is a flowchart which shows an example of operation | movement of the 1st Embodiment of this invention. It is a figure which shows the example of arrangement | positioning of the program set in the 1st Embodiment of this invention. It is a figure which shows the state of each program and the value of processing load immediately after the failover in the 1st Embodiment of this invention. It is a figure which shows the state of each program, and the value of processing load after switching active and standby of the program set which has not failed over according to the switching pattern determined after failover in the 1st Embodiment of this invention. . It is a block diagram of the 2nd Embodiment of this invention. It is a flowchart which shows the procedure which measures the processing capacity of each physical machine in the 2nd Embodiment of this invention. It is a figure which shows an example of the table holding the measurement result of the processing capacity of each physical machine in the 2nd Embodiment of this invention. It is a flowchart which shows the procedure which measures the communication speed between the physical machines in the 2nd Embodiment of this invention. It is a figure which shows an example of the table holding the measurement result of the communication speed between the physical machines in the 2nd Embodiment of this invention. It is a block diagram which shows the structural example which measures the processing load and communication amount of each program in the 2nd Embodiment of this invention. It is a flowchart which shows an example of the procedure which measures the processing load and communication amount of each program in the 2nd Embodiment of this invention. It is a figure which shows an example of the table holding the measurement result of the processing load of each program in the 2nd Embodiment of this invention. It is a figure which shows an example of the table holding the measurement result of the communication amount of each program in the 2nd Embodiment of this invention. It is a flowchart which shows an example of the selection procedure of the arrangement | positioning pattern in the 2nd Embodiment of this invention. It is a figure which shows an example of the weighting factor used in the selection procedure of the arrangement | positioning pattern in the 2nd Embodiment of this invention. It is a flowchart which shows an example of the selection procedure of the switching pattern in the 2nd Embodiment of this invention. It is a flowchart which shows an example of the selection procedure of the arrangement | positioning pattern in the 3rd Embodiment of this invention. It is a figure which shows an example of the weighting coefficient used in the selection procedure of the arrangement | positioning pattern in the 3rd Embodiment of this invention.

Next, embodiments of the present invention will be described in detail with reference to the drawings.
[First embodiment]
Referring to FIG. 1, in a computer system according to a first embodiment of the present invention, a first computer 1100 and a plurality of second computers 1200 are connected via a network 1300 so that they can communicate with each other.

  A computer 1100 is an information processing apparatus (physical machine) such as a personal computer and has a function of controlling the entire computer system.

  The computer 1100 includes a communication interface unit (communication I / F unit) 1101, a storage unit 1102, and a processor 1103.

  The communication I / F unit 1101 includes a dedicated data communication circuit, and has a function of performing data communication with various devices such as a computer 1200 connected via the network 1300.

  The storage unit 1102 includes a storage device such as a hard disk or a memory, and has a function of storing processing information and programs 1104 necessary for various processes in the processor 1103. As processing information, there are a plurality of program sets 1109.

  Each program set 1109 is a set of two programs, one operating as active and the other as standby. By this program set 1109, an FT system including an active virtual machine and a standby virtual machine is generated.

  The program 1104 is a program that implements various processing units by being read and executed by the processor 1103, and can be read by an external device (not shown) or computer readable via a data input / output function such as the communication I / F unit 1101. It is read in advance from a storage medium (not shown) and stored in the storage unit 1102.

  The processor 1103 has a microprocessor such as a CPU and its peripheral circuits, and reads and executes the program 1104 from the storage unit 1102 to realize various processing units by cooperating the hardware and the program 1104. have. The main processing units realized by the processor 1103 include an arrangement pattern determination unit 1105, a program arrangement unit 1106, a switching pattern determination unit 1107, and a switching unit 1108.

  The arrangement pattern determination unit 1105 has a function of determining an arrangement pattern in which a plurality of program sets 1109 are arranged in a plurality of computers 1200. The arrangement pattern determination unit 1105 prevents the two programs of the same program set 1109 from being arranged on the same computer 1200 in determining the arrangement pattern. The arrangement pattern determination unit 1105 determines the arrangement pattern in consideration of at least one of the processing load of each of the plurality of computers 1200 and the communication load between the plurality of computers 1200. Of course, the arrangement pattern may be determined in consideration of various types of information other than the processing load and the communication load.

  The program placement unit 1106 has a function of placing a plurality of program sets 1109 on a plurality of computers 1200 according to the placement pattern determined by the placement pattern determination unit 1105. In other words, the program placement unit 1106 transmits an active program constituting a certain program set 1109 to the placement destination computer 1200 via the network 1300 to generate an active virtual machine on the computer 1200. Further, the program placement unit 1106 transmits a standby-side program constituting a certain program set 1109 to the placement destination computer 1200 via the network 1300 to generate a standby-side virtual machine on the computer 1200.

  When a failure occurs in any computer 1200 and a failover is performed, the switching pattern determination unit 1107 switches between switching between active and standby of the two programs of the remaining program set 1109 that has not been failed over It has a function to generate a pattern. In generating the switching pattern, the switching pattern determination unit 1107 determines the switching pattern in consideration of at least one of the processing load of each of the plurality of computers 1200 and the communication load between the plurality of computers 1200 after failover. Of course, the switching pattern may be determined in consideration of various types of information other than the processing load and the communication load.

  The switching unit 1108 has a function of switching between active and standby of the two programs of the program set 1109 not subjected to the failover according to the switching pattern determined by the switching pattern determination unit 1107. That is, the switching unit 1108 instructs the computer 1200 on which an active program of a certain program set 1109 is placed to change to standby through the network 1300 and at the same time, configures the standby program that configures the program set 1109. Is instructed to change to active through the network 1300.

  The computer 1200 is an information processing apparatus (physical machine) such as a personal computer and has a function of operating a plurality of virtual machines. Similar to the computer 1100, the computer 1200 includes a communication I / F unit, a storage unit, and a processor.

  Further, the computer 1100 and the computer 1200 may have operation input devices such as a keyboard and a mouse, and screen display devices such as an LCD and a PDP.

  FIG. 2 is a flowchart showing the procedure of the program arrangement method executed by the computer system according to the present embodiment. The operation of this embodiment will be described below with reference to FIGS.

  First, the arrangement pattern determining unit 1105 of the computer 1100 determines an arrangement pattern for arranging a plurality of program sets 1109 on a plurality of computers 1200 (step S1001). At this time, the arrangement pattern determination unit 1105 determines an arrangement pattern in which two programs, active and standby, of the same program set are arranged in different computers 1200, respectively. In addition, the arrangement pattern determination unit 1105 determines an arrangement pattern in consideration of at least one of the processing load of the plurality of computers 1200 and the communication load between the computers 1200. For example, when all the computers 1200 have the same performance, the arrangement pattern determination unit 1105 makes the processing load of each computer 1200 (the total load due to the running program) almost equal, and between any two computers 1200 Determining the layout pattern so that the communication load (the amount of traffic from the active and standby programs running on one side to the paired standby and active running on the other side, and vice versa) is almost equal To do. For these processing loads and communication loads, data measured in advance and stored in the storage unit 1102 may be used, or data actually measured using a plurality of computers 1200 may be used.

  Next, the program placement unit 1106 of the computer 1100 places a plurality of program sets 1109 on the plurality of computers 1200 according to the determined placement pattern (step S1002). As a result, a plurality of FT systems, each of which is a set of two virtual machines, active and standby, are generated on a plurality of computers 1200 and start operating.

  Thereafter, when a failure occurs in any of the plurality of computers 1200 and the operation of the active virtual machine operating on the computer 1200 is stopped, a failover is performed, and the processing of the stopped virtual machine is performed on the standby side. The virtual machine takes over and executes. As described above, since the virtual machine that has been operating as a standby until now operates as active, the processing load of the computer 1200 in which the newly activated virtual machine is arranged increases compared to before the failover. In addition, the communication load of the other computer 1200 that was communicating with the program on the computer 1200 in which the failure has occurred is lower than before the failover. As a result, the balance of processing load and communication load remains unbalanced as it is, and the execution efficiency of the entire system is reduced.

  Therefore, in the present embodiment, when a failover is performed, the following processing is promptly executed.

  First, when a failover is performed, the switching pattern determination unit 1107 of the computer 1100 determines a switching pattern for switching between active and standby of the two programs of the remaining program set 1109 that has not been failed over (step S1003). ). In determining the switching pattern, the switching pattern determining unit 1107 determines the switching pattern in consideration of at least one of the processing load of each of the plurality of computers 1200 after failover and the communication load between the computers 1200. For example, when the performance of all the computers 1200 is the same, the switching pattern determination unit 1107 has almost the same processing load (total load due to the operating program) of each computer 1200, and between any two computers 1200. A switching pattern is determined so that the communication load (the amount of traffic from the active and standby programs running on one side to the paired standby and active running on the other side and the amount of traffic in the opposite direction) is almost equal. To do. These processing loads and communication loads may be measured in advance and used in the data stored in the storage unit 1102 or may be data actually measured using a plurality of second computers 1200. good.

  Next, the switching unit 1108 of the computer 1100 switches between active and standby of the two programs of the program set 1109 that has not been failed over according to the determined switching pattern (step S1004). Normally, switching between active and standby can be performed at high speed in a short time of one second or less. As a result, it is possible to improve the balance of processing load and communication load without affecting business processing, and increase the execution efficiency of the entire system.

  Next, the effect of this embodiment will be described using a simplified example.

  FIG. 3 shows an example of a pattern in which a total of eight types of program sets are arranged on four computers 1200 having the same performance by the program arrangement unit 1106.

  Referring to FIG. 3, the computer 1200-1 includes active programs A1 to A3 of the first to third program sets and a standby program S6 of the sixth program set. The computer 1200-2 includes standby programs S1 to S3 of the first to third program sets and active programs A4, A7, and A8 of the fourth, seventh, and eighth program sets. ing. In the computer 1200-3, standby programs S4 and S7 of the fourth and seventh program sets and an active program A5 of the fifth program set are arranged. In the computer 1200-4, standby programs S5 and S8 of the fifth and eighth program sets and an active program A6 of the sixth program set are arranged.

  The numerical value added to the location of the program in FIG. 3 indicates the value of the processing load of the program (the unit is, for example, MIPS). The numerical value added to the location of the computer 120 indicates the total processing load of the programs arranged on the computer. In the arrangement pattern shown in FIG. 3, the processing load of the computer 1200-1 is 91, the processing load of the computer 1200-2 is 93, and the processing loads of the computers 1200-3 and 1200-4 are both 92. The processing load is almost equal.

  FIG. 4 shows the state of each program and the value of the processing load immediately after a failure occurs in the computer 1200-1 and a failover is performed. Programs in the first to third program sets on the computer 1200-2 are switched from standby to active due to failover. As a result, the total processing load of the computer 1200-2 increases from 93 to 180.

  FIG. 5 shows the state of each program and the value of the processing load after switching the active and standby of the program set in which the switching unit 1108 has not failed over in accordance with the switching pattern determined by the switching pattern determining unit 1107. Yes. In the example of FIG. 5, switching between active and standby is performed for the second and eighth program sets. As a result, the processing load on the computer 1200-2 decreases from 180 to 122, and the processing load on the computers 1200-3 and 1200-4 increases from 92 to 121. As a result, the processing loads of the three computers 1200 are almost equal.

  In the above example, the arrangement pattern and the switching pattern are determined in consideration of the processing load. However, the arrangement pattern and the switching pattern are determined in consideration of the communication load between the computers 1200, and both the processing load and the communication load are determined. It is obvious that the same operation and effect can be obtained even if the arrangement pattern and the switching pattern are determined in consideration.

  This embodiment can be variously added and changed as follows.

  In determining the arrangement pattern, the arrangement pattern determining unit 1105 determines the degree of distribution of at least one of the processing load of each computer 1200 after the arrangement and the communication load between the computers 1200 for each of the plurality of arrangement pattern candidates. An index value to be expressed may be calculated, and an arrangement pattern may be determined from the plurality of candidates based on the calculated index value.

  Alternatively, the arrangement pattern determination unit 1105 determines the degree of distribution of at least one of the processing load of each computer 1200 and the communication load between the computers 1200 after arrangement for each of a plurality of arrangement pattern candidates. Expected by switching between active and standby of the two programs of the remaining program set 1109 that are not failed over when a failure occurs in any computer 1200 and a failover occurs An index value representing the degree of distribution of at least one of the processing load of the computer 1200 and the communication load between the computers 1200 is calculated, and a plurality of the arrangement patterns are calculated based on the two types of index values. An arrangement pattern may be determined from the candidates.

  The switching pattern determination unit 1107 determines the degree of distribution of at least one of the processing load of the computer 1200 after switching and the communication load between the computers 1200 for each of the plurality of candidates for the switching pattern. An index value to be expressed may be calculated, and a switching pattern may be determined from the plurality of candidates based on the index value.

[Second Embodiment]
[Overview of this embodiment]
When a set of virtual machines converted into an FT by the high-speed checkpoint method is regarded as a set of program objects, this program set has the following characteristics.
(A) Although the dependency relationship between the program sets constituting the duplex is large, it is necessary to arrange them in separate physical machines. Therefore, it is important to be able to communicate at high speed between program sets.
(B) The load is asymmetric between program sets. The active program is more loaded than the standby program.
(C) Communication is asymmetric between program sets. The amount of communication from the active program to the standby program is large.
(D) When a physical machine fails, failover (switching from standby to active) can be performed at high speed in a short time of one second or less.
(E) Even when the physical machine has not failed, switching between active and standby can be performed at a high speed in a short time of one second or less by an instruction.

  This embodiment is configured in consideration of the characteristics of the program set as described above. In this embodiment, when a plurality of program sets that can switch between active and standby at high speed are arranged for a plurality of physical machines, the processing load of the physical machine and the network communication load between the physical machines are distributed, It is possible to increase the overall execution efficiency. In addition, when a single physical machine becomes unavailable and a failover occurs, it is possible to switch the active and standby of multiple program sets to redistribute the load and increase the overall execution efficiency. .

  More specifically, the management unit issues instructions to the processing capacity measurement unit and the communication speed measurement unit of the physical machine, and measures the processing capacity of the physical machine and the communication speed between the physical machines. In addition, each program set is arranged in a physical machine, and the processing load of the program and the communication amount between the program sets are measured. When multiple program sets are placed on multiple physical machines, the management unit calculates the processing load balancing rating and network load balancing rating for each placement pattern, and optimal placement is achieved so that the load is balanced. Select and execute. Also, when a physical machine becomes unusable and a failover occurs, and when switching between active and standby of multiple program sets, the management unit sets the processing load balancing rating and network load balancing for each switching pattern. Recalculate the rating, select and execute the best switch to balance the load.

[Configuration of this embodiment]
Referring to FIGS. 6 and 11, the present embodiment includes a plurality of physical machines 101 to 105, a network 115 connecting them, and program sets 401 to 402 operating on them. One physical machine includes a management unit 106. Other physical machines include communication speed measuring units 107 to 110 and processing capacity measuring units 111 to 114.

  The management unit 106 instructs the other physical machines to measure the communication speed and processing capacity. In addition, the optimal arrangement of the program set is calculated, and the program set arrangement instruction is actually given to another physical machine. When a physical machine becomes unusable, it instructs the program set to switch between active and standby.

  The communication speed measuring units 107 to 110 measure the network communication speed between physical machines. It also measures individual network traffic between program sets.

  The processing capacity measuring units 111 to 114 measure processing capacities of physical machine processors and I / O. Also measure the individual processing load of the program.

  The program sets 401 to 402 include an active program and a standby program. Active and standby run on separate physical machines. If the physical machine on which active was operating becomes unusable, the standby can be switched to active and processing can continue. Even if the physical machine is not disabled, the management unit 106 can issue a switching instruction to the program set and switch between active and standby.

  The number of physical machines, communication speed measuring units, and processing capacity measuring units need not be four, and can be expanded to an arbitrary number. Further, the number of program sets need not be two, and can be expanded to an arbitrary number.

[Operation of this embodiment]
With reference to FIGS. 6 to 10, an operation for measuring in advance the processing capability of the physical machine and the network communication speed between the physical machines will be described in the present embodiment.

  The management unit 106 selects one physical machine and issues a measurement instruction to the processing capacity measurement unit of the physical machine (step 201). The processing capacity measuring unit measures the processing capacity of the physical machine (step 202) and notifies the result (step 203). The management unit 106 stores the result (step 204). The management unit 106 repeats measurement for all physical machines (step 205), and records the processing capabilities of the physical machines in a table as illustrated in FIG.

  The management unit 106 selects one physical machine and issues a measurement instruction to the communication speed measurement unit of the physical machine (step 301). The communication speed measuring unit measures the communication speed with all other physical machines (step 302, step 303) and notifies the result (step 304). The management unit 106 stores the result (step 305). The management unit 106 repeats the measurement for all the physical machines (step 306), and records the communication speed between the physical machines in the table as illustrated in FIG.

  Next, with reference to FIG. 11 thru | or FIG. 14, the operation | movement which measures the network traffic between a program processing load and a program set in this embodiment in advance is demonstrated.

  The management unit 106 selects one program set and places the active program and the standby program on separate physical machines 102 and 103 (step 501). The processing capacity measuring units 107 and 108 measure the processing load of the program (step 502). The communication speed measuring unit measures the communication amount from active to standby and the communication amount from standby to active (step 503). The measurement is repeated for all the program sets 401 to 402 (step 504) and recorded in the table as illustrated in FIGS.

  Next, with reference to FIG. 15 and FIG. 16, the operation | movement which arrange | positions a some program set in a some physical machine in this embodiment is demonstrated.

The management unit 106 selects one pattern in which a plurality of program sets are arranged on a plurality of physical machines (step 601). For the pattern, a processing load distribution rating is calculated (step 602). Specifically, first, the usage rate is obtained for each physical machine by the following equation.
(Physical machine usage rate) = (total processing load on the program) / (physical machine processing capacity)

Further, using the obtained physical machine usage rate, the variance of the processing usage rate is obtained for each physical machine by the following equation.
(Distribution of processing usage) = Distribution of (physical machine usage)

Further, the processing load distribution rating is obtained for each physical machine by the following equation using the obtained distribution of the physical machine processing utilization rate.
(Rating of processing load distribution) = 1 / (Distribution of processing usage rate)

Next, the management unit 106 calculates a network load balancing rating (step 603). Specifically, first, for each of the upper and lower communications between physical machines, the network usage rate is obtained by the following equation.
(Network usage rate) = (Total program traffic) / (Network communication speed)

Further, using the obtained network usage rate, the variance of the network usage rate is obtained by the following equation.
(Distribution of network usage) = Distribution of each (network usage)

Further, the network load distribution rating is obtained by the following equation using the obtained distribution of the network usage rate.
(Rating of network load distribution) = 1 / (Distribution of network usage rate)

Then, the management unit 106 calculates a total rating according to the following equation using the calculated processing load distribution rating, network load distribution rating, and weighting factor (step 604).
(Overall rating) = (Processing load balancing rating) x (Machine processing weighting factor) + (Network load balancing rating) x (Network weighting factor)

  Here, the weighting coefficient for machine processing and the weighting coefficient for network may be determined in advance as constants, or may be specified by the user of the apparatus as variables. FIG. 16 shows an example of the weighting factor.

The management unit 106 repeats the rating calculation for all the arrangement patterns (step 605). When the number of physical machines is m and the number of program sets is n, the number of arrangement patterns is (m × (m−1)) ⁿ . The management unit 106 selects an arrangement pattern having the largest overall rating (step 606), and executes a process of arranging a plurality of program sets on a plurality of physical machines.

  Next, with reference to FIG. 17, an operation for switching between active and standby of a plurality of program sets when a certain physical machine becomes unusable will be described.

  When a certain physical machine becomes unusable, failover is executed for the program set in which the active program is operating on the physical machine, and the standby program operating on another physical machine is switched to the active program. Therefore, the processing load and the network load are biased. The management unit 106 performs the following process to correct this bias.

The management unit 106 selects one pattern for switching between active and standby of a plurality of program sets (step 701). For the pattern, a processing load distribution rating is calculated (step 702), a network load distribution rating is calculated (step 703), and an overall rating is calculated (step 704). A specific method of calculation is based on steps 602 to 604. Rating calculation is repeated for all switching patterns (step 705). When the number of program sets that can be switched is n, the number of switching patterns is 2 ⁿ . A switching pattern having the largest overall rating is selected (step 706), and switching processing of a plurality of program sets is executed.

[Effect of this embodiment]
The first effect is that when a plurality of program sets are arranged in a plurality of physical machines, the load is distributed with respect to processing and network communication, and the overall execution efficiency is improved.

  This is because the processing load of the physical machine, the network communication speed between the physical machines, the processing load of the program set, and the network traffic between the program set are measured in advance, and the processing load of the program set and the network communication load are asymmetric. This is because the load distribution rating is calculated for each arrangement pattern, and the arrangement pattern having the maximum rating is selected.

  The second effect is that even if the performances of a plurality of physical machines in which a plurality of program sets are arranged are different from each other, the load is distributed for processing and network communication, and the overall execution efficiency is increased. . The reason is the same as the reason for the first effect.

  The third effect is to prevent a decrease in overall execution efficiency even when one physical machine becomes unusable and a failover occurs in a state where a plurality of program sets are arranged on a plurality of physical machines. Be able to.

  The reason is to calculate the load distribution rating for each switching pattern, select the switching pattern with the highest rating, deliberately switch the active and standby of one or more program sets that have not failed over, and distribute the load again This is to make it happen.

[Third embodiment]
In this embodiment, when the management unit 106 places a plurality of program sets on a plurality of physical machines, a switching pattern when the physical machine becomes unusable is simulated in advance, and the load distribution after switching is considered. This is different from the second embodiment in that the arrangement pattern is selected.

  FIG. 18 is a flowchart showing an example of a procedure for selecting an arrangement pattern in the present embodiment. Hereinafter, the operation of the present embodiment will be described with reference to FIG.

  The management unit 106 selects one pattern in which a plurality of program sets are arranged on a plurality of physical machines (step 801). A processing load distribution rating is calculated (step 802), a network load distribution rating is calculated (step 803), and a rating before switching is calculated (step 804). A specific method of calculation is based on steps 602 to 604.

  Next, one physical machine is selected (step 805), and it is assumed that the physical machine becomes unusable (step 806). Hereinafter, switching is simulated. The management unit 106 selects one pattern for switching between active and standby of a plurality of program sets (step 807). A processing load distribution rating is calculated (step 808), a network load distribution rating is calculated (step 809), and a post-switching rating is calculated (step 810). Rating calculation is repeated for all switching patterns (step 811). The operation for simulating this switching is in accordance with steps 701 to 705. Then, the maximum value of the rating after switching is calculated (step 812). The operation of simulating the case where all the physical machines become unusable is repeated (step 813), and the average value of the maximum ratings after switching is calculated (step 814).

Next, the total rating is calculated by the following equation (step 815).
(Overall rating) = (Rating before switching) x (Weighting coefficient before switching) + (Average value of rating after switching) x (Weighting coefficient before switching)

  Here, the weighting coefficient before and after switching may be determined in advance as a constant, or may be specified by the user of the apparatus as a variable. FIG. 19 is an example of a weighting coefficient.

  Rating calculation is repeated for all arrangement patterns (step 816). Then, an arrangement pattern having the largest overall rating is selected (step 817), and a process of arranging a plurality of program sets on a plurality of physical machines is executed.

[Effect of this embodiment]
According to this embodiment, the same effects as those of the second embodiment can be obtained. Further, when a plurality of program sets are arranged on a plurality of physical machines, one physical machine becomes unusable and failover occurs. When a problem occurs and the active and standby of a plurality of program sets are switched, it is possible to obtain an effect that it is possible to select in advance an arrangement that facilitates load distribution for processing and network communication.

  The reason is that the switching pattern when the physical machine becomes unusable is simulated in advance, the total rating is calculated from the rating before switching and the rating after switching, and the layout pattern with the maximum rating is selected. is there.

DESCRIPTION OF SYMBOLS 1100 ... 1st computer 1101 ... Communication I / F part 1102 ... Memory | storage part 1103 ... Processor 1104 ... Program 1105 ... Arrangement pattern determination part 1106 ... Program arrangement | positioning part 1107 ... Switching pattern determination part 1108 ... Switching part 1109 ... Program set 1200 ... Second computer 1300... Network

Claims (7)

A program arrangement method executed by a computer system in which a first computer and a plurality of second computers are connected to each other through a network,
The first computer is an arrangement pattern in which a plurality of program sets, each of which is a set of two programs that operate as one active and the other as a standby, are arranged in the plurality of second computers, The two programs are not arranged in the same second computer, and an arrangement pattern is determined in consideration of at least one of the processing load of each of the plurality of second computers and the communication load between the plurality of second computers. And
The first computer arranges the plurality of program sets in the plurality of second computers according to the determined arrangement pattern,
When a failure occurs in any one of the second computers and the first computer is failed over, the active and standby of the two programs of the remaining program sets that are not failed over A switching pattern for switching between the plurality of second computers after the failover is determined in consideration of at least one of a processing load of each of the plurality of second computers and a communication load between the plurality of second computers,
In accordance with the determined switching pattern, the first computer switches the program that is operating as active out of the two programs of the program set that has not been failed over to standby and operates as standby Switch program active <br/> Program placement method. In determining the arrangement pattern,
The first computer calculates, for each of the plurality of candidates for the arrangement pattern, a first index value representing a degree of distribution of at least one of the processing load and the communication load after arrangement,
The program placement method according to claim 1, wherein the first computer determines the placement pattern from the plurality of candidates based on the calculated first index value. In determining the switching pattern,
The first computer calculates, for each of the plurality of candidates for the switching pattern, a second index value representing a degree of distribution of at least one of the processing load and the communication load after switching,
The program placement method according to claim 1 or 2, wherein the first computer determines the switching pattern from the plurality of candidates based on the calculated second index value. In determining the arrangement pattern,
A first index value representing a degree of distribution of at least one of the processing load and the communication load after placement for each of the plurality of placement pattern candidates; The processing expected by switching between active and standby of the two programs of the remaining program set that is not failed over when a failure occurs in the second computer and a failover occurs Calculating a second index value representing a degree of dispersion of at least one of the load and the communication load;
2. The program arrangement method according to claim 1, wherein the first computer determines the arrangement pattern from a plurality of candidates for the arrangement pattern based on the first index value and the second index value. . In determining the arrangement pattern,
The first computer determines the arrangement pattern from a plurality of candidates for the arrangement pattern based on a value obtained by weighting and adding the first index value and the second index value. 5. The program arrangement method according to 4. The first computer in a computer system in which a first computer and a plurality of second computers are connected to be able to communicate with each other through a network,
An arrangement pattern in which a plurality of program sets, each of which is a set of two programs that operate as one active and the other as a standby, are arranged in the plurality of second computers, and the two programs in the same program set are the same An arrangement pattern determining means for determining an arrangement pattern that is not arranged in the second computer and that takes into account at least one of the processing load of each of the plurality of second computers and the communication load between the plurality of second computers;
Program placement means for placing the plurality of program sets on the plurality of second computers according to the determined placement pattern;
When a failure occurs in any of the second computers and a failover is performed, a switching pattern for switching between active and standby of the two programs of the remaining program set that has not been failed over, Switching pattern determining means for determining in consideration of at least one of the processing load of each of the plurality of second computers after the failover and the communication load between the plurality of second computers;
Switching means for switching a program operating as active among the two programs of the program set that has not been failed over to standby and switching a program operating as standby to active according to the determined switching pattern And a computer having. A first computer in a computer system in which a first computer and a plurality of second computers are connected to each other through a network;
An arrangement pattern in which a plurality of program sets, each of which is a set of two programs that operate as one active and the other as a standby, are arranged in the plurality of second computers, and the two programs in the same program set are the same Determining an arrangement pattern in consideration of at least one of a processing load of each of the plurality of second computers and a communication load between the plurality of second computers, which is not arranged in the second computer;
Arranging the plurality of program sets on the plurality of second computers according to the determined arrangement pattern;
When a failure occurs in any of the second computers and a failover is performed, a switching pattern for switching between active and standby of the two programs of the remaining program set that has not been failed over, Determining in consideration of at least one of the processing load of each of the plurality of second computers after the failover and the communication load between the plurality of second computers;
Switching the program operating as active out of the two programs of the program set not being failed over to standby and switching the program operating as standby active according to the determined switching pattern; A program to let you do.

Images