Optimization System, Optimization Device, Optimization Method, and Program for Optimizing Facility Operating Costs

Abstract

By treating the temperature management of individual servers or computers within a data center as an optimization problem using quantum annealing, an optimization system, an optimization device, an optimization method, and a program are provided to prevent thermal damage to very expensive electronic components such as graphics boards while minimizing the operating costs of the data center. 【Solution means】An optimization system having an annealing device, comprising: a problem generation unit that generates an optimization problem related to the operating cost of a facility; a variable definition unit that defines binary variables for executing the optimization problem with the annealing device; a physical parameter acquisition unit that acquires one or more physical parameter values related to the physical characteristics of the facility; a quadratic unconstrained binary optimization conversion unit that converts the optimization problem into a quadratic unconstrained binary optimization problem using the one or more physical parameter values; a quadratic unconstrained binary optimization problem transceiver that transmits and receives the quadratic unconstrained binary optimization problem to and from the annealing device; and a facility operation control unit that controls the operation of the facility based on the solution obtained from the annealing device. An optimization system is provided.

Description

Technical Field

[0001] The present disclosure relates to an optimization system, an optimization device, an optimization method, and a program for minimizing the operating costs of a facility, particularly a data center.

Background Art

[0002] In recent years, with the booming of generative services such as ChatGPT (registered trademark), the demand for data centers composed of servers equipped with high-performance graphics boards has been increasing steadily. On the other hand, these servers generate a large amount of heat during data processing. Furthermore, the failure rate of the graphics board increases when it is heated above a certain temperature. Therefore, data centers require more advanced temperature management, and the power consumption of air conditioners cannot be ignored. There are multiple individually controllable air conditioners in the data center, and the cooling effect on individual servers by turning them on and off depends on the arrangement of servers, air temperature, rack material, etc. in the data center. Therefore, it is generally extremely difficult to uniquely calculate from the current temperature of each server which air conditioner should be operated at what intensity.

[0003] As a prior art related to the temperature management of data centers, for example, there is JP 2023-104385 A (hereinafter referred to as Patent Document 1). Patent Document 1 discloses a technique for achieving power supply balance and cost reduction by lowering the set temperature of the air conditioner in the data center when the power supply by green power or the like becomes excessive.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

[0005] However, the technology disclosed in Patent Document 1 is to lower the set temperature of the entire data center and cannot perform temperature management of individual servers within the data center.

[0006] On the other hand, as a promising method for dealing with the above-described complex optimization problems, optimization by quantum annealing is considered. Quantum annealing is a technique for solving optimization problems by utilizing the fact that a physical Ising model becomes a combination of spin states in which the total energy of the system is minimized for given conditions between spins. However, no prior art is known that treats the temperature management of individual servers within a data center as an optimization problem by quantum annealing.

Summary of the Invention

Problems to be Solved by the Invention

[0007] The present disclosure has been made in view of the above circumstances, and its object is to prevent thermal damage to very expensive electronic components such as graphics boards while minimizing the operating cost of a data center by treating the temperature management of individual servers or computers within the data center as an optimization problem by quantum annealing, and to provide an optimization system, an optimization device, an optimization method, and a program.

Means for Solving the Problems

[0008] In one aspect, the optimization system of the present disclosure is an optimization system having an annealing device, and includes a problem generation unit that generates an optimization problem related to the operation cost of a facility, a physical parameter acquisition unit that acquires one or more physical parameter values related to the physical characteristics of the facility, a quadratic unconstrained binary optimization conversion unit that uses the one or more physical parameter values to convert the optimization problem into a quadratic unconstrained binary optimization problem, and a facility operation control unit that controls the operation of the facility based on the solution of the quadratic unconstrained binary optimization problem obtained from the annealing device. The quadratic unconstrained binary optimization conversion unit converts the optimization problem into a quadratic unconstrained binary optimization problem including at least the total amount of the electricity cost and the equipment consumption amount of the facility. The optimization system is characterized by this.

[0009] In another aspect, the optimization device of the present disclosure is an optimization device that communicates with an external annealing device, and includes a problem generation unit that generates an optimization problem related to the operation cost of a facility, a physical parameter acquisition unit that acquires one or more physical parameter values related to the physical characteristics of the facility, a quadratic unconstrained binary optimization conversion unit that uses the one or more physical parameter values to convert the optimization problem into a quadratic unconstrained binary optimization problem, a processor that functions as a facility operation control unit that controls the operation of the facility based on the solution of the quadratic unconstrained binary optimization problem obtained from the annealing device, and a computer having a memory. The quadratic unconstrained binary optimization conversion unit converts the optimization problem into a quadratic unconstrained binary optimization problem including at least the total amount of the electricity cost and the equipment consumption amount of the facility. The optimization device is characterized by this.

[0010] In another aspect, the optimization method of the present disclosure includes a problem generation step of generating an optimization problem related to the operation cost of a facility, a physical parameter acquisition step of obtaining one or more physical parameter values related to the physical characteristics of the facility, a quadratic unconstrained binary optimization conversion step of using the one or more physical parameter values to convert the optimization problem into a quadratic unconstrained binary optimization problem, and a facility operation control step of controlling the operation of the facility based on the solution of the quadratic unconstrained binary optimization problem obtained from an annealing device. The quadratic unconstrained binary optimization conversion step includes a step of converting the optimization problem into a quadratic unconstrained binary optimization problem including at least the total amount of the electricity cost and the equipment consumption amount of the facility. It is an optimization method executed by a computer.

[0011] In yet another aspect, the program of the present disclosure causes a computer to execute a problem generation procedure for generating an optimization problem related to the operation cost of a facility, a physical parameter acquisition procedure for obtaining one or more physical parameter values related to the physical characteristics of the facility, a quadratic unconstrained binary optimization conversion procedure for using the one or more physical parameter values to convert the optimization problem into a quadratic unconstrained binary optimization problem, and a facility operation control procedure for controlling the operation of the facility based on the solution of the quadratic unconstrained binary optimization problem obtained from an annealing device. The quadratic unconstrained binary optimization conversion procedure includes a procedure of converting the optimization problem into a quadratic unconstrained binary optimization problem including at least the total amount of the electricity cost and the equipment consumption amount of the facility. It is a program.

Advantages of the Invention

[0012] According to the optimization system, optimization device, optimization method, and program according to the present disclosure, the temperature management of individual servers or computers in a data center can be treated as an optimization problem by quantum annealing. Therefore, it is possible to minimize the operation cost of the data center while preventing thermal damage to very expensive electronic components such as graphics boards.

Brief Description of the Drawings

[0013]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

[0014] Hereinafter, embodiments of an optimization system, an optimization apparatus, an optimization method, and a program according to the present disclosure will be described with reference to the drawings. In all the drawings, the same members are denoted by the same reference numerals. Also, the embodiments shown in the drawings are merely examples and do not limit the present disclosure in any sense. Further, the embodiments and their modifications described below can be arbitrarily combined, and such combinations are included in the scope of the present disclosure.

[0015] In the following, a data center will be taken as an example of the facility to be optimized, but the facility is not limited to this. For example, other examples of facilities include agricultural greenhouses, cultivation or aquaculture facilities, seafood storage facilities, bio-product (such as human or animal sperm, eggs, embryos, blood, etc.) storage facilities, pharmaceutical (such as vaccines) storage facilities, zoos, aquariums, office buildings, logistics centers, and the like.

[0016] FIG. 1 is an overall conceptual diagram showing an example in which an optimization system 10 according to an embodiment of the present disclosure is applied to a data center 1. The data center 1 is composed of a plurality of server rooms 2. Each server room 2 includes a plurality of racks 3, a plurality of air conditioners 4, a power supply unit 5, and a task distribution unit 6. The power supply unit 5 may also be referred to as a temperature control unit 5.

[0017] Each rack 3 has a housing and a plurality of shelves provided at equal intervals inside the housing for mounting computers. Each air conditioner 4 may have a cooling capacity of, for example, 40 to 60 kW, and about 3 to 5 units are provided in one server room 2. In this example, for the sake of explanation, one air conditioner 4 is provided for one rack 3, but it is not limited to this, and one air conditioner 4 may be provided for two racks 3.

[0018] The power supply unit 5 has a function of receiving power supply from an external power supply system and distributing power to the air conditioner 4. In addition, the power supply unit 5 controls the cooling intensity, the amount of cooling air flow, and the ON / OFF of the air conditioner 4 so that the set temperature is reached as a temperature control unit. The task distribution unit 6 has a function of receiving a task command from an external network 7 and allocating a task to be executed to a predetermined computer (server) mounted on the rack 3.

[0019] The optimization system 10 is connected to a power supply unit (temperature control unit) 5 and a task distribution unit 6 via an operation instruction unit 8 and a network 7. As will be described later, the operation instruction unit 8 transmits an instruction signal to the power supply unit 5 and the task distribution unit 6 via the network 7 according to the solution obtained by the annealing device. The power supply unit 5 operates the air conditioner 4 according to the instruction signal, and the task distribution unit 6 allocates tasks to any one of the computers (servers) in the rack 3 according to the instruction signal. That is, the operation instruction unit 8, the power supply unit (temperature control unit) 5, and the task distribution unit 6 are configured to optimally operate the server room 2, and thus the data center 1, based on the solution obtained from the annealing device.

[0020] Figure 2 is a diagram schematically showing the inside of the server room 2 of the data center 1. In the following description, the computer may also be a server. In the left rack 1(3), computers 1(31), 2(32), 3(33), and 4(34) are installed, and in the right rack 2(3), computers 5(35) ··· M(36) are installed. An air conditioner 1(4) is installed adjacent to the rack 1(3), and an air conditioner 2(4) is installed adjacent to the rack 2(3). Each rack 3 is installed on a floor panel 30 having a double-floor structure with a large number of small holes formed therein.

[0021] Each air conditioner 4 generates a cold air flow 40 under the double floor in a direction parallel to the front surface of each rack 3 and parallel to the floor surface. The cold air flow 40 rises as cold air 41 above the double floor through the small holes of the floor panel 30. The cold air 41 is sucked to the front surface of each rack 3 to cool the computers 31 to 36 in the rack 3, and is discharged as high-temperature exhaust air 42 from the back surface of the rack 3. The exhaust air 42 becomes a warm air flow 43 in the opposite direction to the cold air flow 40 and returns to the air conditioner 4.

[0022] Computers 31 to 36 are each provided with a temperature sensor (not shown). The temperatures of computers 31 to 36 are measured by the temperature sensors, for example, every 1 minute, every 5 minutes, or every 10 minutes. The time interval for temperature measurement may be arbitrary. The measured temperatures of computers 31 to 36 show scattered values, for example, computer 1 (31) is 48.6 °C, computer 2 (32) is 53.4 °C, computer 3 (33) is 55.0 °C, computer 4 (34) is 50.6 °C, computer 5 (35) is 49.3 °C, and computer M (36) is 65.6 °C. These temperature information are sent to the power supply unit (temperature control unit) 5.

[0023] Figure 3 is a block diagram showing an optimization system, an optimization device, and peripheral devices according to an embodiment of the present disclosure. The optimization device system includes an optimization device 11, which is a classical computer, and an annealing device 19. The optimization device 11 includes a control device 12, a memory 14, a storage device 16, and an interface 18. The optimization device 11 may be connected to peripheral devices such as a server room 2, a machine learning device 20, an input device 22, and a display device 24 via a network 7.

[0024] The control device 12 may be a computer including one or more processors (for example, a CPU). The memory 14 may be a volatile memory such as a DRAM (Dinamic Random Access Memory) or an SRAM (Static Random Access Memory). The storage device 16 may be a non-volatile memory such as a ROM (Read Only Memory), a flash memory such as an SSD (Solid State Drive), or an HDD (Hard Disk Drive).

[0025] Interface 18 includes an input unit, an output unit, and a communication unit. Interface 18 is connected to input device 22, display device 24, and machine learning device 20 via network 7. Network 7 may be a network such as the Internet or WiFi. Input device 22 may be a keyboard, mouse, touch panel, touch pen, etc. Display device 24 may be a liquid crystal display, an organic EL display, etc. Machine learning device 20 may be a machine learning machine using AI (Artificial Intelligence).

[0026] Annealing device 19 may be a quantum annealing type computer, a digital annealer (registered trademark), a CMOS annealing machine, a simulated bifurcation machine, etc. Also, annealing device 19 may be implemented as a cloud service such as Leap (D-Wave Systems), AWS (Amazon Web Services), Braket, Microsoft Azure Quantum, IBM Quantum Experience. That is, annealing device 19 may be installed at a location spatially separated from control device 12. For example, only annealing device 19 may be installed overseas. Note that annealing device 19 is a hardware device capable of executing a quantum annealing algorithm different from optimization device 11. Optimization system 10 may include an optimization solver (not shown) in addition to annealing device 19. The optimization solver may be realized as software implementing the primal-dual interior point method or Newton's method capable of solving optimization problems.

[0027] Control device 12 functions as a problem generation unit that generates an optimization problem, a variable definition unit, a physical parameter acquisition unit, a quadratic unconstrained binary optimization conversion unit, a quadratic unconstrained binary optimization problem transceiver unit, and a facility operation control unit.

[0028] First, the function of the problem generation unit will be described. The problem generation unit generates an optimization problem regarding the operation cost of a facility such as Data Center 1. Here, the operation cost is the total C of the cost based on the power consumption of the facility such as the electricity cost, the cost based on the facility consumption of the facility such as the equipment consumption amount, and the cost based on the productivity of the facility's equipment (in the case of Data Center 1, Computers 31 to 36). total Generally, it is defined as the following (Equation 1). C total = Σ i C i (E, W, -R) (Equation 1) Here, C i is the cost of the i-th server room 2, E is the electricity cost, W is the equipment consumption amount, and R represents the profit. The problem generation unit defines the optimization problem as a solution problem to minimize C total . The problem generation unit stores the defined optimization problem in the optimization problem storage unit in Memory 14.

[0029] Next, the function of the variable definition unit will be described. The variable definition unit defines binary variables for executing the optimization problem on the annealing device 19. In order to solve the optimization problem using the annealing device 19, it is necessary to represent the optimization problem by an Ising model that models the optimization problem as a magnetic body model. The Ising model is a simplified model for dealing with the phase transition of ferromagnetic materials, and is represented by the state of spins with up and down directions representing the properties of the magnetic body, the interaction coefficient representing the force of interaction between two spins, and the external magnetic field coefficient representing the force of an externally applied magnetic field. The Ising model determines the energy by the interaction between spins that take two values, for example, +1 and -1, and the state of the spins is updated so that the energy of the Ising model becomes the minimum. The combination of parameters that minimizes the evaluation index of the optimization problem is obtained as the combination of the states of the spins that minimizes the energy by mapping the optimization problem so as to correspond to the energy of the Ising model and converging the Ising model.

[0030] The evaluation function for the entire spin, represented by a variable that takes two values of 0 and 1 instead of spins of +1 and -1, is in the QUBO (Quadratic Unconstrained Binary Optimizaztion) form. The variable definition section defines variables that take two values of 0 and 1 in order to formulate the optimization problem (Equation 1) defined in the problem generation section in the QUBO form.

[0031] The specific function of the variable definition section in this embodiment will be described. For example, assume that N air conditioners 4 are installed in the data center 1. Each air conditioner 4 can be individually controlled by a power supply unit (temperature control unit) 5, and the power consumption is determined according to the air conditioning intensity. Also assume that M computers 31 to 36 are housed in a plurality of racks 3 in the data center 1. Here, the computer refers to any processing system whose temperature and power consumption can be measured. That is, a plurality of servers may be regarded as one computer, or each individual rack 3 of the server may be regarded as one computer.

[0032] Tasks that are processed every moment come into the data center 1. Numbers are assigned to the expected tasks from one point in time to another point in time for distinction. Each task has a determined reward and the power required for processing. That is, let the reward of the i-th task be (Equation 1) and the power required for processing be (Equation 2).

Equation

Equation

[0033] In this example, the variable definition section defines two variables that can be operated when performing optimization. The first variable is related to the task allocation by the task allocation unit 6 and is defined as a variable (Equation 3) indicating whether the i-th task is executed by the j-th computer.

Equation

Number

Number

Number

[0034] The second variable is related to the power supplied to the air conditioner 4 by the power supply unit (temperature control unit) 5, that is, the power consumption of the air conditioner 4, and is defined as a variable (Equation 7) indicating that the i-th air conditioner 4 is operated at intensity j (for example, j takes values of 0 (intensity 0: off), 1 (intensity 1: weak), 2 (intensity 2: medium), 3 (intensity 3: strong)).

Number

[0035] Here, similar to (Equation 3), (Equation 7) is also a binary variable. When it becomes (Equation 8), the i-th air conditioner 4 is operated at intensity j.

Number

[0036] The variable definition unit stores the two variables (Equation 3) and (Equation 7) defined as above in the variable definition storage unit in the memory 14.

[0037] Next, the function of the physical parameter acquisition unit will be described. The physical parameter acquisition unit acquires one or more physical parameter values related to the physical characteristics of the facility. Here, the physical characteristics of the facility refer to the physical characteristics of the equipment in the data center 1 that change according to the execution of tasks by the computer and / or the operating intensity of the air conditioner 4. Examples of the physical parameter values related to the physical characteristics include the current temperature of the computer, the temperature rise per unit power consumption of the computer, the temperature drop of the computer when the intensity of the air conditioner 4 is set to a certain value, and the thermal fatigue degree of the computer. Among these, the temperature rise per unit power consumption of the computer and the temperature drop of the computer when the intensity of the air conditioner 4 is set to a certain value vary depending on factors such as the actual arrangement of the computers, the material of the rack 3, and the outside air temperature, so they are actually measured.

[0038] Among the above-described physical parameters, the current temperature of the computer is detected by the temperature sensors installed in each computer and sent to the physical parameter acquisition unit via the network 7. The temperature rise per unit power consumption of the computer and the temperature drop of the computer when the intensity of the air conditioner 4 is set to a certain value are actually measured in the server room 2 and sent to the physical parameter acquisition unit via the network 7. This will be described later. Also, the thermal fatigue degree of the computer is calculated from the number of thermal cycles of the computer stored and updated in the storage device 16 and sent to the physical parameter acquisition unit. The physical parameter acquisition unit stores the acquired physical parameter values in the physical parameter storage unit of the memory 14.

[0039] Next, the function of the quadratic unconstrained binary optimization conversion unit will be described. The quadratic unconstrained binary optimization conversion unit uses the above-described one or more physical parameter values to convert the optimization problem (Equation 1) into a quadratic unconstrained binary optimization problem (QUBO format). First, the electricity cost E in (Equation 1) will be formulated. The power consumption when the intensity of the air conditioner 4 is i is represented by (Equation 9).

Equation

Equation

Equation

Equation

Equation

[0040] Next, formulate the profit R of (Equation 1). The profit R is formulated by (Equation 14).

Equation

[0041] Next, formulate the equipment consumption amount W of (Equation 1). This is to define a quantity related to temperature in order to calculate the failure rate of the computer. Among the physical parameter values described above, let the temperature rise amount per unit power consumption in the computer be T unit The heat-resistant temperature of the server is L. L is a constant. Let the current temperature of the i-th server (computer) be represented by (Equation 15). Let the equipment consumption degree per unit temperature be w i be represented by. w iIt is a value representing how much consumption occurs each time the temperature of the computer rises by 1 degree. The amount by which the temperature of computer i drops when air conditioner j is set to intensity k is represented by (Equation 16). The facility consumption amount W is formulated as (Equation 17).

Number

Number

Number

[0042] The quadratic unconstrained binary optimization conversion unit formulates (Equation 1) as an evaluation function in QUBO format as (Equation 18).

Number

[0043] Next, the functions of the quadratic unconstrained binary optimization problem transceiver will be described. The quadratic unconstrained binary optimization problem transceiver outputs the quadratic unconstrained binary optimization problem to the annealing device 19. The quadratic unconstrained binary optimization problem is formulated by converting the optimization problem into the QUBO format by the quadratic unconstrained binary optimization conversion unit, and is represented as (Equation 18) in this example. The quadratic unconstrained binary optimization problem transceiver outputs the evaluation function (Equation 18) stored in the evaluation function storage unit of the memory 14 to the annealing device 19 via a network such as the Internet. The annealing device 19 calculates a set of optimal solutions that minimize the QUBO-formatted evaluation function (Equation 18), that is, an optimal parameter set for task allocation (Equation 3) to the computer and operation control of the air conditioner 4 (Equation 7). In the example where the annealing device 19 is a digital annealer, the digital annealer randomly changes all variables. As a result, the evaluation expression goes up and down, but if it is only changed in the downward direction, the optimal solution cannot be obtained, so changes that initially worsen the evaluation are also probabilistically allowed. Then, the optimal value is found by gradually attenuating the probability of deterioration. By continuously improving the evaluation in this way, the optimal solution can be obtained in a short time. The quadratic unconstrained binary optimization problem transceiver stores the solution obtained from the annealing device 19 via the network in the solution storage unit of the memory 14.

[0044] Next, the functions of the facility operation control unit will be described. The facility operation control unit controls the operation of the facility based on the solution obtained from the annealing device 19 via the network. The operation of the facility refers to the allocation of tasks to the computers in the data center 1 and the operation control of the air conditioner 4. The facility operation control unit sends a control signal to the operation instruction unit 8, and the operation instruction unit 8 transmits an instruction signal to the server room 2, that is, the power supply unit (temperature control unit) 5 and the task allocation unit 6 shown in FIG. 1 via the network 7. Based on the instruction signal, the power supply unit (temperature control unit) 5 controls the ON / OFF and intensity of the air conditioner 4, and the task allocation unit 6 controls the allocation of tasks to the computers.

[0045] FIG. 4 is a flowchart showing an optimization method according to an embodiment. The optimization method according to this embodiment includes a problem generation step (step S1) of generating an optimization problem regarding the operation cost of a facility, a variable definition step (step S2) of defining binary variables in order to execute the optimization problem by the annealing device 19, a physical parameter acquisition step (step S3) of acquiring one or more physical parameter values regarding the physical characteristics of the facility, a quadratic unconstrained binary optimization conversion step (step S4) of converting the optimization problem into a quadratic unconstrained binary optimization problem using the one or more physical parameter values, a quadratic unconstrained binary optimization problem transmission / reception step (steps S5 and S6) of transmitting and receiving the quadratic unconstrained binary optimization problem to and from the annealing device 19, and a facility operation control step (step S7) of controlling the operation of the facility based on the solution obtained from the annealing device 19.

[0046] In step S1, the problem generation unit generates the optimization problem of (Equation 1) and stores it in the optimization problem storage unit of the memory 14.

[0047] In step S2, the variable definition unit defines the first variable (Equation 3) and the second variable (Equation 7), and stores the defined first variable (Equations 3, 4, 5) and the second variable (Equations 7 and 8) in the variable definition storage unit of the memory 14. Also, the conditions (Equation 6) of each variable are stored together.

[0048] In step S3, the physical parameter acquisition unit acquires the current temperature (Equation 15) of the server (computer), the temperature rise amount T per unit power consumption of the computer unit , the temperature decrease amount (Equation 16) of the computer i when the intensity of the air conditioner j is set to k, the consumption degree w per unit temperature of the computer i , and the thermal fatigue degree F of the i-th computer i . Also, parameters other than the physical parameters, for example, the reward (Equation 1) of the i-th task, the power required for the i-th task (Equation 2), the power consumption (Equation 9) when the intensity of the air conditioner 4 is i, and the electricity price P per unit power consumption of the computer or the air conditioner 4 are input from the input device 22 and stored in the storage device 16.

[0049] Among the physical parameters, the current temperature (numeral 15) of the server (computer) is detected by a temperature sensor built into the computer, transmitted to the physical parameter acquisition unit at a predetermined time interval, and overwritten in the physical parameter storage unit of the memory 14.

[0050] The power consumption degree w per unit temperature of the computer i varies depending on the processing speed and purchase time of the computer in the server room 2. It is preferable to create a table of the values of w i with the processing speed on the horizontal axis and the purchase time on the vertical axis, and store it in the storage device 16. The table may be created by inputting from the input device 22 or downloaded via the network 7. Note that the power consumption degree w i per unit temperature of the computer may be a constant.

[0051] The thermal fatigue degree F of the i-th computer i can be determined from the number of thermal cycles of the i-th computer. The number of thermal cycles refers to the number of times the temperature detected by the temperature sensor built into the computer changes at a value equal to or higher than a certain threshold value. The thermal fatigue degree F i can be obtained as F i = α·n. Here, α is a coefficient and n represents the number of thermal cycles. The thermal fatigue degree F of the computer thus obtained i is stored in the storage device 16.

[0052] The temperature rise amount T per unit power consumption of the computer unit and the temperature drop amount (numeral 16) of the i-th computer when the intensity of the air conditioner j is set to k vary depending on factors such as the actual arrangement of the computers, the material of the rack 3, and the outside air temperature as described above, so they are actually measured. The measurement method will be described in detail below.

[0053] FIG. 5 is a schematic diagram showing the correlation between the execution of a task, the operation of the air conditioner 4, and the temperature of the computer over time. In this example, it is assumed that the types of tasks are the same. When a task is assigned to the computer, the computer executes the task. As the task is executed, the temperature of the computer rises, and the air conditioner 4 turns on. In this example, for the sake of explanation, the air conditioner 4 is turned on simultaneously with the start of the task, but the timing of turning on the air conditioner 4 is not limited to this.

[0054] Referring to the upper part of FIG. 5, the air conditioner 4 is operating at intensity 1, i.e., "weak". It can be seen that the temperature of the computer at this time rises at a steep slope angle, reaches the maximum temperature, and then gradually decreases. Next, referring to the middle part of FIG. 4, the air conditioner 4 is operating at intensity 2, i.e., "medium". It can be seen that the temperature of the computer at this time rises at a slightly smaller slope angle than in the case of intensity 1, reaches a slightly lower maximum temperature than in the case of intensity 1, and then decreases at a slightly steeper slope angle than in the case of intensity 1. Next, referring to the lower part of FIG. 4, the air conditioner 4 is operating at intensity 3, i.e., "strong". It can be seen that the temperature of the computer at this time rises at a slightly smaller slope angle than in the case of intensity 2, reaches a slightly lower maximum temperature than in the case of intensity 2, and then decreases at a slightly steeper slope angle than in the case of intensity 2. Also, it can be seen that the maximum temperature difference of the computer is intensity 1 < intensity 2 < intensity 3.

[0055] FIG. 6 is a graph summarizing the results of FIG. 5 in terms of the relationship between the temperature of the computer and time. The temperature rise amount T per unit power consumption of the computer unit can be obtained from the slope of the graph shown in FIG. 6. The temperature of the computer rises at a predetermined speed between t0 and t1, regardless of the intensity of the air conditioner 4. This indicates that it takes time for the computer to be cooled after the air conditioner 4 is turned on. That is, since the influence of the air conditioner 4 can be ignored between t0 and t1, the temperature rise amount T per unit power consumption of the computer unit can be obtained from the slope of the graph and the power consumption of the computer during this period.

[0056] On the other hand, the temperature decrease amount of the computer i (Equation 16) when the intensity of the air conditioner j is set to k can be obtained from the graph shown in FIG. 6 as the temperature decrease ΔT within a predetermined time Δt from the time when the temperature of the computer reaches its maximum. The physical parameter acquisition unit calculates the temperature increase amount T per unit power consumption of the computer obtained in this way unit and the temperature decrease amount of the computer (Equation 16) are stored in the physical parameter storage unit of the memory 14.

[0057] As shown in FIG. 6, the temperature increase amount T per unit power consumption of the computer unit and the temperature decrease amount of the computer (Equation 16) are obtained through actual measurement, but can also be predicted using the machine learning device 20. The arrangement of the computers installed in the rack 3 in the server room 2, the material of the rack 3, the outside air temperature, the amount of tasks, the temperature increase amount T per unit power consumption of the computer unit and the actual measurement sample data of the temperature decrease amount of the computer (Equation 16) are accumulated in the database, and through machine learning by AI, the graph of FIG. 6 can be predicted for any computer. Thus, without actual measurement, from the graph of FIG. 6 predicted using the machine learning device 20, the temperature increase amount T per unit power consumption unit and the temperature decrease amount of the computer (Equation 16) can be obtained. The physical parameter acquisition unit calculates the temperature increase amount T per unit power consumption of the computer obtained using the machine learning device 20 unit and the temperature decrease amount of the computer (Equation 16) may be stored in the physical parameter storage unit of the memory 14 via the network 7.

[0058] In step S4, the quadratic unconstrained binary optimization conversion unit accesses the optimization problem storage unit, variable definition storage unit, and physical parameter storage unit of the memory 14, and the storage device 16 to obtain binary variables (Equation 3), physical parameter values (Equations 15, T unit , Equation 16, w i , F i) and other parameter values (Equation 1, Equation 2, Equation 9, P) are read out, the electricity cost E (Equation 10), the profit R (Equation 14), and the facility consumption amount W (Equation 17) are formulated, and the evaluation function is formulated as in (Equation 18), thereby converting the optimization problem generated by the problem generation unit into the QUBO format. The quadratic unconstrained binary optimization conversion unit stores the evaluation function (Equation 18) in the evaluation function storage unit of the memory 14.

[0059] In step S5, the quadratic unconstrained binary optimization problem transceiver accesses the evaluation function storage unit of the memory 14 to read out the evaluation function (Equation 18), and transmits it to the annealing device 19 via the network. The annealing device 19 calculates a solution that minimizes the evaluation function (Equation 18).

[0060] In step S6, the quadratic unconstrained binary optimization problem transceiver obtains a solution from the annealing device 19 via the network and stores it in the solution storage unit of the memory 14.

[0061] In step S7, the facility operation control unit accesses the solution storage unit of the memory 14 and sends a control signal to the operation instruction unit 8. The operation instruction unit 8 transmits an instruction signal to the server room 2 via the network 7, that is, to the power supply unit (temperature control unit) 5 and the task distribution unit 6 shown in FIG. 1. Based on the instruction signal, the power supply unit (temperature control unit) 5 controls the ON / OFF and intensity of the air conditioner 4, and the task distribution unit 6 controls the task allocation to the computer.

[0062] Next, other embodiments according to the present disclosure will be described with reference to the drawings. The above-described embodiments are applicable to cases where the facility is other than a data center (for example, agricultural greenhouses, cultivation or aquaculture facilities, seafood storage facilities, bioproduct (human or animal sperm, eggs, embryos, blood, etc.) storage facilities, pharmaceutical (vaccine, etc.) storage facilities, zoos, aquariums, office buildings, logistics centers, etc.), but the other embodiments described below are more preferably applicable when the facility is a data center.

[0063] Descriptions of configurations similar to those of the above-described embodiments will be omitted, and only different configurations will be described. In other embodiments, the definitions of variables for causing a computer to execute tasks, the definition of power consumption, the definition of electricity charges, the formula for trouble costs, and the formula for total costs are different from those of the above-described embodiments. Specific descriptions will be given below.

[0064] In this embodiment, a load balancer is used when causing a computer to execute a task. Basically, all tasks to be processed in the data center are processed by computers within the server. However, if tasks are evenly allocated to all computers in the server, computers with a high temperature will have a greater processing load than computers with a low temperature. Therefore, by using a load balancer, fewer tasks are allocated to computers with a high temperature and more tasks are allocated to computers with a low temperature.

[0065] In this embodiment, the strength of the load balancer is represented by the first variable q. The variable q can be set in two levels by using m qubits. The variable q can be represented by a binary variable as in (Equation 19). m It is possible to set two levels. The variable q can be represented by using a binary variable as in (Equation 19).

Equation

[0066] FIG. 7 is a table showing the correspondence between the variable q of the m qubits indicating the strength of the load balancer according to another embodiment and the binary variables (q1, q2, q3, ··· q m )). When q = 0, the strength of the load balancer becomes 0, and processing is allocated equally to all computers in the server. As the strength of the load balancer (that is, the value of q) increases, fewer tasks are allocated to computers with a high temperature and more tasks are allocated to computers with a low temperature.

[0067] The value of the strength q of the load balancer takes discrete values between 0 and 1. Examples of how to determine the specific distribution are shown below. Let the temperature distribution of the servers be T. That is, if the temperature of the i-th server is T i then, the temperature distribution of N servers can be expressed as in (Equation 20).

Equation

Equation

Equation

Equation

Equation

[0068] Also, in this embodiment, a new way of thinking about power consumption is introduced. Focusing on the fact that among the power consumed in the data center, the power that can be optimized is the power of the air conditioner. That is, as described above, all the tasks to be processed in the data center are basically processed by the computers in the server. That is, the power consumption of the server can be excluded from the optimization target, and by narrowing the optimization target to the power consumption of the air conditioner, cost management in line with the actual data center becomes possible.

[0069] Based on the above way of thinking, the power consumption P is re - defined by the formula shown in (Equation 25).

Equation

[0070] Based on the power consumption P defined in (Equation 25), the power cost E for the entire air - conditioning system in the data center is defined by the formula shown in (Equation 26). However, it is assumed that the indoor temperature near the air conditioner is the same for all air conditioners.

Equation

[0071] Also, in this embodiment, a new concept for equipment consumption is introduced. Hereinafter, the term "trouble cost" is used instead of equipment consumption. When a computer continues to operate beyond a certain temperature, the probability of failure increases. Therefore, when considering the long-term operation cost of a data center, the cost caused by this failure (for example, the purchase price of the hardware itself, the equipment replacement work cost, etc.) cannot be ignored. In addition to failures, problems such as a decrease in calculation speed and a decrease in calculation accuracy also occur during operation at high temperatures. The cost consisting of these multiple factors (including factors other than those caused by the computer) during high-temperature operation is collectively referred to as trouble cost. The trouble cost can be calculated as a function W(T) of temperature T by considering the occurrence probability and cost of these factors.

[0072] By the way, the occurrence probability of troubles varies for each factor, but there may be cases where multiple factors are intricately intertwined to form another trouble factor. Also, the specifications and processing capabilities of individual computers are not the same. Therefore, it is generally difficult to accurately calculate the trouble cost function W(T) as a function of temperature T. On the other hand, the trouble cost itself is a parameter for which rough calculation is possible and has rough characteristics. For example, the trouble cost W(T) is high up to a certain temperature T (T high ≥ T) takes a value almost zero, and when it exceeds T high (T high < T), it starts to take a high value. Also, as the temperature increases, the increase rate of the trouble cost becomes larger. Based on these characteristics and the measured trouble cost (if any), the trouble cost function W(T) can be estimated. Note that the trouble cost function W(T) may be obtained by extrapolating the measured values of the trouble cost, or may be predicted by a machine learning device 20 through machine learning by AI.

[0073] Next, a method for obtaining an approximate trouble cost W' by applying a linear approximation to the trouble cost function will be described. FIG. 8 is a graph for explaining the method of obtaining the approximate trouble cost W'. The horizontal axis of the graph shown in FIG. 8 is T high which represents the rising temperature T (°C) from, and the vertical axis represents the trouble cost (for example, the unit is ten thousand yen). In the example shown in FIG. 8, it is assumed that the trouble cost function W(T) is represented as a function of W(T)=T 2 . Note that the trouble cost function W(T) is not limited to this.

[0074] First, divide the temperature T in the domain into several sections. In the example shown in FIG. 8, the domain is divided into three sections (section 1, section 2, section 3) with the values of T = 0, 3, 6, 9. The method of division is not limited to this, and it may be divided at any value of T. Also, the interval between each section may be larger at lower temperatures and smaller at higher temperatures. Next, connect the adjacent values of W(T) corresponding to T = 0, 3, 6, 9 with a straight line, and obtain a linear function formula W(T)=aT + b for each section. Here, the slope a of the linear function formula corresponds to w in (Equation 17), and is a value representing how much consumption occurs each time the temperature of the computer rises by 1 degree. Let the temperature of computer i at the start of calculation be T i , then T i can be measured, so it is known which section of the domain it belongs to. In the example of FIG. 8, T i belongs to section 2. Let the slope a of the linear function formula W(T)=aT + b corresponding to section 2 to which T i belongs be the w of computer i i . As shown in FIG. 8, when the task is executed, the temperature of the computer rises from the measurement start temperature T i to T i ' while being cooled by the air conditioner. The approximate trouble cost W'(T i ') can be obtained from the linear function formula W(T)=aT + b corresponding to section 2 at that temperature T i '. i

[0075] In this way, the approximate formula W' of the overall trouble cost can be formulated as in (Equation 27).

Number

[0076] Therefore, as the evaluation function in QUBO format in the present embodiment, the total cost C can be expressed as in (Equation 28).

Number

[0077] Note that in the present embodiment, the physical parameters to be actually observed are the trouble cost W(T), (Equation 16), T iand X. For W(T), it can be calculated by multiplying the occurrence probability for each trouble factor by the cost, obtained by plotting and extrapolating past trouble data, or obtained from AI machine learning data. For (Equation 16), it can be obtained from Q or available from the specifications of the air conditioner, etc. T i can be obtained from the temperature sensor of the computer. For X, it is possible to predict the task amount for each time period from past data.

[0078] The optimization method according to the above-described embodiment or other embodiments can be executed by a program. The program causes a computer to execute a problem generation procedure for generating an optimization problem related to the operation cost of the facility, a variable definition procedure for defining binary variables for executing the optimization problem by an annealing device, a physical parameter acquisition procedure for acquiring one or more physical parameter values related to the physical characteristics of the facility, a quadratic unconstrained binary optimization conversion procedure for converting the optimization problem into a quadratic unconstrained binary optimization problem using the one or more physical parameter values, a quadratic unconstrained binary optimization problem transmission and reception procedure for transmitting and receiving the quadratic unconstrained binary optimization problem to and from the annealing device, and a facility operation control procedure for controlling the operation of the facility based on the solution obtained from the annealing device.

[0079] The information processing by the program is specifically realized using hardware resources such as the control device 12 (CPU), the memory 14, the storage device 16, and the interface 18. The program may be provided recorded on various computer-readable storage media in an installable format or an executable format file. Also, the program may be stored in a computer connected to a network such as the Internet and downloaded via the network or directly provided via the network.

[0080] The descriptions in the specification and the drawings of the embodiments, other embodiments, and variations described above are merely examples of the present disclosure and do not limit the scope of the present disclosure. Also, various modifications and changes can be made without departing from the spirit and aspects of the present disclosure, and it is known to those skilled in the art that those are also included in the scope of the present disclosure.

Explanation of Signs

[0081] 1 Data center, 2 Server room, 3 Rack, 4 Air conditioner, 5 Power supply unit, Temperature control unit, 6 Task distribution unit, 7 Network, 8 Operation instruction unit, 10 Optimization system, 11 Optimization device, 12 Control device, 14 Memory, 16 Storage device, 18 Interface, 19 Annealing device, 20 Machine learning device, 22 Input device, 24 Display device, 30 Floor panel, 31 - 36 Computers, 40 Flow of cold air, 41 Cold air, 42 Exhaust, 43 Flow of warm air

Claims

1. An optimization system having an annealing device, comprising: a problem generation unit that generates an optimization problem related to the operation cost of a facility; a physical parameter acquisition unit that acquires one or more physical parameter values related to the physical characteristics of the facility; a quadratic unconstrained binary optimization conversion unit that uses the one or more physical parameter values to convert the optimization problem into a quadratic unconstrained binary optimization problem; a facility operation control unit that controls the operation of the facility based on the solution of the quadratic unconstrained binary optimization problem obtained from the annealing device ; the quadratic unconstrained binary optimization conversion unit converts the optimization problem into a quadratic unconstrained binary optimization problem including at least the total amount of electricity cost and equipment consumption of the facility; An optimization system characterized by the above.

2. An optimization device that communicates with an external annealing device, comprising: a problem generation unit that generates an optimization problem related to the operation cost of a facility; a physical parameter acquisition unit that acquires one or more physical parameter values related to the physical characteristics of the facility; a quadratic unconstrained binary optimization conversion unit that uses the one or more physical parameter values to convert the optimization problem into a quadratic unconstrained binary optimization problem; a facility operation control unit that controls the operation of the facility based on the solution of the quadratic unconstrained binary optimization problem obtained from the annealing device ; a processor that functions as such; a memory ; and a computer having the same; the quadratic unconstrained binary optimization conversion unit converts the optimization problem into a quadratic unconstrained binary optimization problem including at least the total amount of electricity cost and equipment consumption of the facility; An optimization device characterized by the above.

3. The optimization system according to claim 1, wherein the facility includes a load and a cooling device for cooling the load.

4. The optimization device according to claim 2, wherein the facility includes a load and a cooling device for cooling the load.

5. The one or more physical parameter values include at least one of the current temperature of the load, the temperature decrease amount of the load by the cooling device, the temperature increase amount per unit power consumption of the load, the consumption degree per unit temperature of the load, the actual trouble cost, and the predicted task amount per unit time. The optimization system according to claim 3, characterized by the above.

6. The one or more physical parameter values include at least one of the current temperature of the load, the amount of temperature decrease of the load by the cooling device, the amount of temperature increase per unit power consumption of the load, the degree of consumption per unit temperature of the load, the actual trouble cost, and the predicted amount of tasks per unit time. The optimization device according to claim 4 is characterized by this.

7. A problem generation step of generating an optimization problem related to the operation cost of a facility, A physical parameter acquisition step of acquiring one or more physical parameter values related to the physical characteristics of the facility, A quadratic unconstrained binary optimization conversion step of converting the optimization problem into a quadratic unconstrained binary optimization problem using the one or more physical parameter values, A facility operation control step of controlling the operation of the facility based on the solution of the quadratic unconstrained binary optimization problem obtained from the annealing device and The quadratic unconstrained binary optimization conversion step includes a step of converting the optimization problem into a quadratic unconstrained binary optimization problem including at least the total amount of electricity cost and equipment consumption of the facility. An optimization method executed by a computer.

8. Cause a computer to Execute a problem generation procedure for generating an optimization problem related to the operation cost of a facility, Execute a physical parameter acquisition procedure for acquiring one or more physical parameter values related to the physical characteristics of the facility, Execute a quadratic unconstrained binary optimization conversion procedure for converting the optimization problem into a quadratic unconstrained binary optimization problem using the one or more physical parameter values, Execute a facility operation control procedure for controlling the operation of the facility based on the solution of the quadratic unconstrained binary optimization problem obtained from the annealing device and The quadratic unconstrained binary optimization conversion procedure includes a procedure of converting the optimization problem into a quadratic unconstrained binary optimization problem including at least the total amount of electricity cost and equipment consumption of the facility. A program.

Images