Dynamic frequency optimization method, device, system and computer readable storage medium based on time-of-use electricity price

Abstract

The application relates to a dynamic frequency adjustment optimization method, device, system and computer readable storage medium based on time-of-use electricity price. The dynamic frequency adjustment optimization method based on time-of-use electricity price establishes a dynamic frequency adjustment optimization model based on time-of-use electricity price, decides the running frequency of a program in each time period, adjusts the processor running frequency in different time periods by considering the difference of electricity prices in different time periods, minimizes the total electricity fee expenditure, and does not cause great change of job running time.

Description

Technical Field

[0001] This application relates to dynamic frequency regulation technology for supercomputer systems, and more specifically, to a dynamic frequency regulation optimization method, system, and computer-readable storage medium based on time-of-use pricing. Background Technology

[0002] Supercomputers, with their outstanding characteristics of high computing speed, high communication bandwidth, and large storage capacity, support large-scale computing and problem-solving in the country's high-tech fields and cutting-edge technologies. They provide strong support for traditional fields such as scientific computing, energy, meteorology, and engineering simulation, as well as artificial intelligence applications, and are of paramount importance to national security, economic and social development. Currently, the primary challenge facing the development of large-scale supercomputing is energy consumption. For example, the power consumption of an "exascale" supercomputer can reach up to 50MW, with annual operating electricity costs reaching hundreds of millions. Therefore, given the difficulties in reducing the energy consumption of supercomputing systems in chip design and system architecture, research on energy-saving management to reduce operating expenses is particularly important.

[0003] Dynamic frequency scaling (VFS) is essentially a low-power technology that aims to set the operating voltage and clock frequency according to the chip's actual power consumption needs, ensuring that the provided power is sufficient without overkill, thereby reducing power consumption. VFS optimization technology utilizes the chip's characteristic that its energy consumption is proportional to the square of the voltage and the clock frequency; therefore, reducing the clock frequency can lower processor power consumption. However, simply reducing the clock frequency does not save energy, as reduced performance increases job execution time. Most VFS research focuses on frequency scaling for parallel tasks within a single job, adjusting the running frequencies of different tasks to improve their coordination and complete the job in a shorter time, thus saving energy. These VFS optimization methods do not consider the differences in electricity prices at different times of day. The job scheduling method proposed by Yang et al. (see: Yang X, Zhou Z, Wallace S, et al. Integrating dynamic pricing of electricity into energy aware scheduling for HPC systems[C]. IEEE HighPerformance Computing, Networking, Storage & Analysis, 2016, 58-64.) takes into account the differences in electricity prices at different times, but reduces resource utilization when electricity prices are high and increases resource utilization when electricity prices are low, thereby achieving the goal of reducing total electricity costs. Summary of the Invention

[0004] The technical problem to be solved by this application is to provide a dynamic frequency adjustment optimization method, system and computer-readable storage medium based on time-of-use pricing, which reduces total electricity costs by adjusting the operating frequency of the processor during different time periods, in order to address the above-mentioned deficiencies of the prior art.

[0005] To address its technical problem, this application proposes a dynamic frequency regulation optimization method based on time-of-use pricing in its first aspect, the method comprising the following steps:

[0006] S1. Divide a day into S time periods at a time interval τ, and define the decision variable f. s Let w represent the frequency at which the program runs during time period s, and let w be the electricity price during time period s. s The following dynamic frequency regulation optimization model based on time-of-use pricing is established:

[0007]

[0008]

[0009]

[0010] in:

[0011] Equation (1) represents the objective function for minimizing total electricity costs, where π is the gate flip probability of the processor core, C is the capacitance of the processor core, η is the system parameter, and n s This represents the number of times the program occupies the time period s;

[0012] Equation (2) indicates that the total running cycle of the program needs to be completed within the specified number of time periods, where r represents the proportion of time used for calculation in each time period, and c represents the total running cycle of the program executed by the processor core;

[0013] Equation (3) represents the decision variable f s The range of values, where f min f represents the minimum operating frequency. max Indicates the maximum operating frequency;

[0014] S2. Solve the dynamic frequency regulation optimization model based on time-of-use pricing;

[0015] S3. Adjust the processor core to run the program at the operating frequency obtained in step S2.

[0016] According to one embodiment of the first aspect of this application, step S1 further includes:

[0017] Define the set of selectable operating frequencies D for the program, and define f. d Let d be the operating frequency in set D, and let:

[0018]

[0019] I s,d =rτn s f d

[0020] Define decision variable x s,d The value d indicates whether the operating frequency is selected in time period s; it is 1 if yes and 0 otherwise. The following dynamic frequency regulation optimization model based on time-of-use pricing is established:

[0021]

[0022]

[0023]

[0024]

[0025] in:

[0026] Equation (4) represents the objective function for minimizing total electricity expenses;

[0027] Equation (5) indicates the total running cycle of the program that needs to be completed within the specified number of time periods;

[0028] Equation (6) indicates that one and only one operating frequency is selected for each time period;

[0029] Equation (7) represents the decision variable x s,d The range of values ​​for .

[0030] According to one embodiment of the first aspect of this application, in the dynamic frequency regulation optimization model based on time-of-use pricing, the following relationship between operating time and frequency is established:

[0031] t = t F +t H

[0032]

[0033]

[0034] Where t is the time it takes for the program to run at a frequency of f0 across all processor cores on n nodes. F For the frequency-dependent portion of the runtime, t H denoted as frequency-independent runtime, c represents the total runtime of the program executed by each processor core, f represents the system operating frequency, and r represents the percentage of time spent on computation in each time period.

[0035] According to one embodiment of the first aspect of this application, in the dynamic frequency regulation optimization model based on time-of-use pricing, the following relationship between operating energy consumption and frequency is established:

[0036] f = ηV

[0037] P D =CV 2 f=πCη -2 f 3

[0038] Where f is the system operating frequency, V is the supply voltage, and P is the power supply voltage. D η represents dynamic power consumption, π represents the gate toggle probability of the processor core, C represents the capacitance of the processor core, and η represents system parameters.

[0039] According to one embodiment of the first aspect of this application, in step S2, a commercial or open-source solver is used to solve the dynamic frequency regulation optimization model based on time-of-use pricing.

[0040] To address its technical problem, this application proposes a dynamic frequency regulation optimization device based on time-of-use pricing in a second aspect, comprising:

[0041] The modeling module is used to divide a day into S time periods at a time interval τ, and define the decision variable f. s Let w represent the frequency at which the program runs during time period s, and let w be the electricity price during time period s. s The following dynamic frequency regulation optimization model based on time-of-use pricing is established:

[0042]

[0043]

[0044]

[0045] in:

[0046] Equation (1) represents the objective function for minimizing total electricity costs, where π is the gate flip probability of the processor core, C is the capacitance of the processor core, η is the system parameter, and n s This represents the number of times the program occupies the time period s;

[0047] Equation (2) indicates that the total running cycle of the program needs to be completed within the specified number of time periods, where r represents the proportion of time used for calculation in each time period, and c represents the total running cycle of the program executed by the processor core;

[0048] Equation (3) represents the decision variable f s The range of values, where f min f represents the minimum operating frequency. max Indicates the maximum operating frequency;

[0049] The solution module is used to solve the dynamic frequency regulation optimization model based on time-of-use pricing;

[0050] The frequency modulation module is used to adjust the processor core to execute the program at the operating frequency obtained by the solution module.

[0051] According to one embodiment of the second aspect of this application, the modeling module is further configured to:

[0052] Define the set of selectable operating frequencies D for the program, and define f. d Let d be the operating frequency in set D, and let:

[0053]

[0054] I s,d =τn s f d

[0055] Define decision variable x s,d The value d indicates whether the operating frequency is selected in time period s; it is 1 if yes and 0 otherwise. The following dynamic frequency regulation optimization model based on time-of-use pricing is established:

[0056]

[0057]

[0058]

[0059]

[0060] in:

[0061] Equation (4) represents the objective function for minimizing total electricity expenses;

[0062] Equation (5) indicates the total running cycle of the program that needs to be completed within the specified number of time periods;

[0063] Equation (6) indicates that one and only one operating frequency is selected for each time period;

[0064] Equation (7) represents the decision variable x s,d The range of values ​​for .

[0065] According to one embodiment of the second aspect of this application, the solving module uses a commercial or open-source solver to solve the dynamic frequency regulation optimization model based on time-of-use pricing.

[0066] To address its technical problem, this application proposes a dynamic frequency regulation optimization system based on time-of-use pricing in a third aspect, comprising a processor and a memory, wherein the memory stores a computer program, and when the computer program is executed by the processor, it implements the steps of the dynamic frequency regulation optimization method based on time-of-use pricing as described above.

[0067] In order to solve its technical problem, this application proposes a computer-readable storage medium in a fourth aspect, which stores a computer program that, when executed by a processor, implements the steps of the dynamic frequency regulation optimization method based on time-of-use pricing as described above.

[0068] The implementation of the time-of-use pricing-based dynamic frequency regulation optimization method, apparatus, system, and computer-readable storage medium of this application has the following beneficial effects: The time-of-use pricing-based dynamic frequency regulation optimization method according to the embodiments of this application establishes a time-of-use pricing-based dynamic frequency regulation optimization model, determines the program's running frequency in each time period, and adjusts the processor's running frequency in different time periods by considering the differences in electricity prices in different time periods, so as to minimize the total electricity cost and keep the job running time from changing too much. Attached Figure Description

[0069] The present application will be further described below with reference to the accompanying drawings and embodiments. In the accompanying drawings:

[0070] Figure 1 This is a flowchart of a dynamic frequency regulation optimization method based on time-of-use pricing according to an embodiment of this application;

[0071] Figure 2 This is a diagram showing the operating frequency of each operation at different times.

[0072] Figure 3 This is a diagram comparing electricity costs under normal frequency and frequency regulation operation.

[0073] Figure 4 This is a logic block diagram of a dynamic frequency regulation optimization device based on time-of-use pricing according to an embodiment of this application;

[0074] Figure 5 This is a logic block diagram of a dynamic frequency regulation optimization system based on time-of-use pricing according to an embodiment of this application. Detailed Implementation

[0075] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application. Furthermore, the embodiments and features described herein can be combined with each other unless otherwise specified.

[0076] Supercomputing centers often have long-running jobs, some lasting several days. Users are generally not very sensitive to fluctuations in the runtime of these jobs within a small range. Electricity prices vary throughout the day, with higher prices during peak daytime hours and lower prices during off-peak nighttime hours. Therefore, the frequency of these long-running jobs can be reduced when electricity prices are high and increased when prices are low, thereby reducing overall operating electricity costs.

[0077] Based on this, this application proposes a dynamic frequency regulation optimization method based on time-of-use pricing, which determines the running frequency of the program in each time period, so as to minimize the total operating electricity cost of the supercomputing center and keep the operation time relatively stable.

[0078] Figure 1 A flowchart of a dynamic frequency regulation optimization method 100 based on time-of-use pricing according to an embodiment of this application is shown. Figure 1 As shown, the dynamic frequency regulation optimization method 100 based on time-of-use pricing includes the following steps:

[0079] In step S110, a day is divided into S time periods by a time interval τ, and the decision variable f is defined. s Let w represent the frequency at which the program runs during time period s, and let w be the electricity price during time period s. s The following dynamic frequency regulation optimization model based on time-of-use pricing is established:

[0080]

[0081]

[0082]

[0083] in:

[0084] Equation (1) represents the objective function for minimizing total electricity costs, where π is the gate flip probability of the processor core, C is the capacitance of the processor core, η is the system parameter, and n s This represents the number of times the program occupies the time period s;

[0085] Equation (2) indicates that the total running cycle of the program needs to be completed within the specified number of time periods, where r represents the proportion of time used for calculation in each time period, and c represents the total running cycle of the program executed by the processor core;

[0086] Equation (3) represents the decision variable f s The range of values, where f min f represents the minimum operating frequency. max This indicates the maximum operating frequency.

[0087] In step S110, this application establishes the relationship between running time and running frequency as follows:

[0088] Suppose that the program takes t minutes to run on all processor cores across n nodes at a frequency of f0, and the frequency-dependent portion of the program takes t minutes to run. F The frequency-independent portion of the runtime is t. H Then the following equation is satisfied:

[0089] t = t F +t H

[0090] The percentage of time allocated for calculation in each time period is as follows:

[0091]

[0092] Assuming the program load is balanced, each processor core executes the program for the same total number of cycles, c, and the system operating frequency is f, then:

[0093]

[0094] Divide a day into S time periods at regular intervals τ (e.g., 1 hour). Assume that the electricity price is the same for the same time period on different days, and that the operating frequency is the same. Let w be the electricity price for time period s. s Given the duration of the program's execution, we can deduce the number of times it occupies each time slot, let's call this n. s For example, if a program is submitted at midnight and has a runtime of 1 day, then the number of times it is used in each time period is 1.

[0095] This application further establishes the relationship between operating energy consumption and operating frequency in step S110 as follows:

[0096] Regarding energy consumption, let E be the total energy consumption, P be the power consumption, and t be the running time. The relationship between total energy consumption and power consumption is as follows:

[0097] E = Pt

[0098] Power consumption P is determined by dynamic power consumption P D and static power consumption P S It consists of two parts, and the relationship between each part and voltage and frequency is shown in the following formula:

[0099] P = P D +P S =πCV 2 f+VI leak

[0100] Where π is the gate flip probability of the processor core (i.e., the probability of the circuit transitioning from 0 to 1), C is the capacitance of the processor core, V is the supply voltage, f is the system operating frequency, and I...leak This is the leakage current. f and V need to satisfy the following relationship:

[0101]

[0102] Among them, V th Let V be the threshold voltage, and η and α be system parameters. Generally, V... th Since f is much smaller than V, and α is taken as 2, the relationship between f and V is approximately linear as follows:

[0103] f = ηV

[0104] In summary:

[0105] P D =πCV 2 f=πCη -2 f 3

[0106] Define decision variable f s If we represent the program's operating frequency during time period s, then we can suggest the dynamic frequency regulation optimization model based on time-of-use pricing, as shown in equations (1), (2), and (3) above. Equation (2) ensures that the total program runtime does not fluctuate too much.

[0107] The above model contains a cubic term in the objective function, classifying it as a nonlinear programming model, which is difficult to solve. Since the program's running frequency can generally be selected within a certain set, it facilitates frequency control. For example, the running frequency might fall within the range [1.5, 1.6, ..., 2.5, 2.6]. Based on this, we define D as the set of selectable running frequencies for the program, f d Let be the d-th operating frequency in set D. Also, let:

[0108]

[0109] I s,d =τn s f d

[0110] Define decision variable x s,d This indicates whether the d-th operating frequency is selected in time period s; the value is 1 if yes and 0 otherwise. The 0-1 linear programming model is as follows:

[0111]

[0112]

[0113]

[0114]

[0115] in:

[0116] Equation (4) represents the objective function for minimizing total electricity expenses;

[0117] Equation (5) indicates the total running cycle of the program that needs to be completed within the specified number of time periods;

[0118] Equation (6) indicates that one and only one operating frequency is selected for each time period;

[0119] Equation (7) represents the decision variable x s,d The range of values ​​for .

[0120] Then, in step S120, the dynamic frequency regulation optimization model based on time-of-use pricing is solved. In practice, step S120 can use existing commercial solvers (such as CPLEX, GUROBI) or open-source solvers (such as SCIP) to solve the problem.

[0121] Then in step S130, the processor core is adjusted to execute the program at the operating frequency obtained in step S120.

[0122] According to the above embodiments of this application, the dynamic frequency adjustment optimization method 100 based on time-of-use pricing establishes a dynamic frequency adjustment optimization model based on time-of-use pricing, determines the running frequency of the program in each time period, and adjusts the processor running frequency in different time periods by considering the differences in electricity prices in different time periods, so as to minimize the total electricity cost and keep the job running time from changing too much.

[0123] To verify the effectiveness of the dynamic frequency regulation optimization method based on time-of-use pricing described in this application, a total of 12 operations were designed as shown in Table 1. In Table 1, the start and end times are on different days, and the total runtime is (24 - start time) + end time + 24 * number of intermediate days. For example, the first operation is 23 + 24 + 24 = 71. Other parameter settings are shown in Table 2.

[0124] Table 1 - Relevant Information for the Example Exercises

[0125]

[0126] Table 2 - Parameter Settings

[0127]

[0128] Figure 2 The operating frequency of each task is given for each time period. It can be seen that the operating frequency is lower during peak electricity consumption periods [9h, 23h], and higher during off-peak electricity consumption periods. Figure 3 A cost comparison between normal frequency and frequency modulation operation is given, with the average cost per core decreasing by approximately 0.005 yuan per hour.

[0129] Therefore, the dynamic frequency regulation optimization method based on time-of-use pricing according to the foregoing embodiments of this application has the following advantages:

[0130] 1. Effectively saves on supercomputing electricity operating costs without affecting user experience.

[0131] 2. It is highly practical and has a wider range of applications. It is suitable for both serial and parallel processing of tasks that run for a longer period of time.

[0132] 3. The constructed optimization model is a 0-1 integer programming model. The model size is relatively small, and it can be solved using the general open-source solver SCIP, which will meet the efficiency requirements.

[0133] Based on the dynamic frequency regulation optimization method based on time-of-use pricing in the foregoing embodiments of this application, this application also proposes a dynamic frequency regulation optimization device based on time-of-use pricing. Figure 4 A logic block diagram of a dynamic frequency regulation optimization device 200 based on time-of-use pricing according to one embodiment of this application is shown. Figure 4 As shown, the dynamic frequency regulation optimization device 200 based on time-of-use pricing includes a modeling module 210, a solution module 220, and a frequency regulation module 230. The modeling module 210 is used to divide a day into S time periods at a time interval τ, and define the decision variable f. s Let w represent the frequency at which the program runs during time period s, and let w be the electricity price during time period s. s The following dynamic frequency regulation optimization model based on time-of-use pricing is established:

[0134]

[0135]

[0136]

[0137] Where: Equation (1) represents the objective function for minimizing total electricity expenditure, where π is the gate flip probability of the processor core, C is the capacitance of the processor core, η is the system parameter, and n s Let be the number of times the program occupies time period s; Equation (2) indicates that the total running cycle of the program needs to be completed within the specified number of time periods, where r represents the proportion of time used for calculation in each time period, and c represents the total running cycle of the processor core executing the program; Equation (3) represents the decision variable f s The range of values, where f min f represents the minimum operating frequency. maxThis represents the maximum operating frequency. The solver module 220 is used to solve the dynamic frequency regulation optimization model based on time-of-use pricing. Specifically, the solver module 220 can use existing commercial solvers (such as CPLEX, GUROBI) or open-source solvers (such as SCIP). The frequency regulation module 230 is used to adjust the processor core to execute the program at the operating frequency obtained by the solver module 220.

[0138] In a further optimized embodiment, the modeling module 210 is also used to set the set D of selectable running frequencies for the program, and define f d Let d be the operating frequency in set D, and let:

[0139]

[0140] I s,d =τn s f d

[0141] Define decision variable x s,d The value d indicates whether the operating frequency is selected in time period s; it is 1 if yes and 0 otherwise. The following dynamic frequency regulation optimization model based on time-of-use pricing is established:

[0142]

[0143]

[0144]

[0145]

[0146] Wherein: Equation (4) represents the objective function of minimizing total electricity expenditure; Equation (5) represents the total running cycle of the program that needs to be completed within the specified number of time periods; Equation (6) represents that each time period has one and only one operating frequency; Equation (7) represents the decision variable x s,d The range of values ​​for .

[0147] The time-of-use pricing-based dynamic frequency regulation optimization device 200 according to the above embodiments of this application is used to implement the aforementioned time-of-use pricing-based dynamic frequency regulation optimization method 100. For further specific implementation of each module of the time-of-use pricing-based dynamic frequency regulation optimization device 200, please refer to the foregoing detailed description of each step of the time-of-use pricing-based dynamic frequency regulation optimization method 100.

[0148] Based on the dynamic frequency regulation optimization method based on time-of-use pricing in the foregoing embodiments of this application, this application also proposes a dynamic frequency regulation optimization system 300 based on time-of-use pricing. See also Figure 5As shown, the dynamic frequency regulation optimization system 300 based on time-of-use pricing includes a processor 310 and a memory 320, which are communicatively connected. The memory 320 stores a computer program, which, when executed by the processor 310, causes the processor 310 to implement the steps of the dynamic frequency regulation optimization method based on time-of-use pricing according to the foregoing embodiments of this application.

[0149] Processor 310 can be a Central Processing Unit (CPU). Processor 310 can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, or combinations of the above types of chips. The general-purpose processor can be a microprocessor or any conventional processor.

[0150] The memory 320, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs, non-transitory computer-executable programs, and modules, such as the program instructions / modules corresponding to the dynamic frequency regulation optimization method based on time-of-use pricing in the embodiments of the present invention. The processor 310 executes various functional applications and data processing of the server by running the non-transitory software programs, instructions, and modules stored in the memory 320, thereby implementing the dynamic frequency regulation optimization method based on time-of-use pricing in the aforementioned embodiments.

[0151] The memory 320 may include a program storage area and a data storage area. The program storage area may store the operating system and application programs required for at least one function. The data storage area may store data created based on the use of the time-of-use pricing-based dynamic frequency regulation optimization method. Furthermore, the memory 320 may include high-speed random access memory and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, the memory 320 may optionally include memory remotely located relative to the processor 310, and these remote memories may be connected to the processor 310 via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0152] This application also proposes a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the dynamic frequency regulation optimization allocation method based on time-of-use pricing as described in the foregoing embodiments of this application. The computer-readable storage medium may be a magnetic disk, optical disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk drive (HDD), or solid-state drive (SSD), etc. The computer-readable storage medium may also include combinations of the above-mentioned types of memory.

[0153] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A dynamic frequency regulation optimization method based on time-of-use pricing, characterized in that, The method includes the following steps: S1, dividing a day into time intervals Divided into S Define decision variables for each time period. Indicates the program is in a time period The operating frequency makes the time period The electricity price is The following dynamic frequency regulation optimization model based on time-of-use pricing is established: in: Equation (1) represents the objective function for minimizing total electricity expenditure, where, This represents the gate flip probability of the processor core. For the processor core capacitors, For system parameters, Program usage time period The number of times; Equation (2) represents the total running cycle of the program that needs to be completed within a specified number of time periods, where, This indicates the percentage of time used for calculation in each time period. This indicates the total execution cycle of the program executed by the processor core; Equation (3) represents the decision variables. The range of values, where, Indicates the minimum operating frequency. Indicates the maximum operating frequency; S2. Solve the dynamic frequency regulation optimization model based on time-of-use pricing; S3. Adjust the processor core to execute the program at the operating frequency obtained in step S2; In the dynamic frequency regulation optimization model based on time-of-use pricing, the relationship between operating time and frequency is established as follows: in, For the program, all processor cores on n nodes are configured at various frequencies. Runtime For the frequency-dependent portion of the runtime, The portion of runtime that is independent of frequency. The total execution cycle of a program for each processor core. For system operating frequency, The percentage of time used for calculation in each time period.

2. The dynamic frequency regulation optimization method based on time-of-use pricing according to claim 1, characterized in that, Step S1 further includes: The program can select a set of operating frequencies. ,definition For set Middle One operating frequency, and simultaneously set: Define decision variables Indicates time period Do you want to choose the first one? The operating frequency is set to 1 if it is set to 1 otherwise, and a dynamic frequency regulation optimization model based on time-of-use pricing is established as follows: in: Equation (4) represents the objective function for minimizing total electricity expenditure; Equation (5) indicates the total running cycle of the program that needs to be completed within the specified number of time periods; Equation (6) indicates that one and only one operating frequency is selected for each time period; Equation (7) represents the decision variables. The range of values ​​for .

3. The dynamic frequency regulation optimization method based on time-of-use pricing according to claim 1, characterized in that, In the dynamic frequency regulation optimization model based on time-of-use pricing, the following relationship between operating energy consumption and frequency is established: in, For system operating frequency, The power supply voltage, For dynamic power consumption, This represents the gate flip probability of the processor core. For the processor core capacitors, These are system parameters.

4. The dynamic frequency regulation optimization method based on time-of-use pricing according to claim 1, characterized in that, In step S2, a commercial or open-source solver is used to solve the dynamic frequency regulation optimization model based on time-of-use pricing.

5. A dynamic frequency regulation optimization device based on time-of-use pricing, characterized in that, include: The modeling module is used to divide a day into time intervals. Divided into S Define decision variables for each time period. Indicates the program is in a time period The operating frequency makes the time period The electricity price is The following dynamic frequency regulation optimization model based on time-of-use pricing is established: in: Equation (1) represents the objective function for minimizing total electricity expenditure, where, This represents the gate flip probability of the processor core. For the processor core capacitors, For system parameters, Program usage time period The number of times; Equation (2) represents the total running cycle of the program that needs to be completed within a specified number of time periods, where, This indicates the percentage of time used for calculation in each time period. This indicates the total execution cycle of the program executed by the processor core; Equation (3) represents the decision variables. The range of values, where, Indicates the minimum operating frequency. Indicates the maximum operating frequency; The solution module is used to solve the dynamic frequency regulation optimization model based on time-of-use pricing; The frequency modulation module is used to adjust the processor core to execute the program at the operating frequency obtained by the solution module; In the dynamic frequency regulation optimization model based on time-of-use pricing, the relationship between operating time and frequency is established as follows: in, For the program, all processor cores on n nodes are configured at various frequencies. Runtime For the frequency-dependent portion of the runtime, The portion of runtime that is independent of frequency. The total execution cycle of a program for each processor core. For system operating frequency, The percentage of time used for calculation in each time period.

6. The dynamic frequency regulation optimization device based on time-of-use pricing according to claim 5, characterized in that, The modeling module is further used for: The program can select a set of operating frequencies. ,definition For set Middle One operating frequency, and simultaneously set: Define decision variables Indicates time period Do you want to choose the first one? The operating frequency is set to 1 if it is set to 1 otherwise, and a dynamic frequency regulation optimization model based on time-of-use pricing is established as follows: in: Equation (4) represents the objective function for minimizing total electricity expenditure; Equation (5) indicates the total running cycle of the program that needs to be completed within the specified number of time periods; Equation (6) indicates that one and only one operating frequency is selected for each time period; Equation (7) represents the decision variables. The range of values ​​for .

7. The dynamic frequency regulation optimization device based on time-of-use pricing according to claim 5, characterized in that, The solution module uses a commercial or open-source solver to solve the dynamic frequency regulation optimization model based on time-of-use pricing.

8. A dynamic frequency regulation optimization system based on time-of-use pricing, characterized in that, It includes a processor and a memory, the memory storing a computer program that, when executed by the processor, implements the steps of the dynamic frequency regulation optimization method based on time-of-use pricing as described in any one of claims 1-4.

9. A computer-readable storage medium, characterized in that, The device contains a computer program that, when executed by a processor, implements the steps of the dynamic frequency regulation optimization method based on time-of-use pricing as described in any one of claims 1-4.

Images