Method and device for intelligent scheduling of computing and electricity coordination in multi-interconnected data center park

Abstract

The application provides a kind of multi-interconnected data center park's algorithm electricity collaborative intelligent scheduling method and device, comprising: each data center park calculates respective algorithm electricity bearing capacity index;Each data center park calculates the migratability of local task, and classifies local task;Each data center park sends report information to collaborative scheduling platform;Report information includes: algorithm electricity bearing capacity index, migratability, the classification result of local task and comprehensive credit score;Collaborative scheduling platform is based on report information and carries out the many-to-many dynamic matching of each data center park;Each data center park is based on the matching result of many-to-many dynamic matching and executes computing power migration and operation task, and updates the comprehensive credit score of undertaker based on execution result.The algorithm electricity bearing capacity index and migratability index are proposed, the quantitative representation of park state and task characteristics is realized, which not only protects the commercial privacy of park, but also provides a unified decision basis for cross-park collaboration.

Description

Technical Field

[0001] This invention relates to the field of energy dispatching technology, and in particular to a computing and power collaborative intelligent dispatching method and apparatus for multi-interconnected data center parks. Background Technology

[0002] In the context of the deep integration of the digital economy, computing-power collaboration has become a core direction for the coordinated development of computing infrastructure and energy systems. The basic idea of ​​computing-power collaboration is to break down the geographical boundaries between computing power and energy, treating computing power as a mobile resource. Based on the energy supply, electricity price levels, and computing power supply and demand in different regions, computing tasks are dynamically scheduled across different parks to achieve optimal matching of computing power resources and energy resources, thereby improving overall operational efficiency and reducing carbon emissions and operating costs.

[0003] Multi-interconnected data center parks, serving as the core carriers of computing-power collaboration, consist of multiple geographically distributed, network-connected independent data center parks. Each park integrates photovoltaic power generation systems, electrochemical energy storage systems, and cooling systems to provide stable and reliable energy for IT (Information Technology) computing load equipment. Compared to the operation and scheduling of single data center parks, multi-interconnected data center parks are characterized by the flexible migration of computing loads between parks, deep coupling between energy systems and computing systems, and a much larger global optimization space than a single park. Research on computing-power collaborative scheduling strategies for multi-interconnected data center parks has significant engineering application value for achieving cross-regional green energy consumption, reducing overall operating costs, and improving the reliability of computing services.

[0004] However, current research generally employs centralized optimization frameworks such as model predictive control and reinforcement learning, which require collecting detailed operational data from all parks, including commercially sensitive information such as computing power costs, energy reserves, and equipment parameters. This not only results in an explosive increase in computational load and high communication latency, but also often makes engineering implementation difficult due to park operators' reluctance to disclose core commercial data. On the other hand, most existing methods use fixed rules or offline-trained models for task matching, which cannot adapt to dynamic scenarios such as fluctuations in photovoltaic output, real-time changes in electricity prices, and sudden arrival of tasks, easily leading to local optima or even system imbalance. Summary of the Invention

[0005] In view of this, the purpose of this invention is to provide a computing and power collaborative intelligent scheduling method and device for multi-interconnected data center parks, so as to realize the quantitative characterization of park status and task characteristics, realize global optimal matching in distributed scenarios, and also correct the resource reservation parameters of key unbalanced parks.

[0006] In a first aspect, embodiments of the present invention provide a computing power collaborative intelligent scheduling method for multi-interconnected data center parks. The method includes: each data center park calculating its own computing power carrying capacity index; each data center park calculating the portability of local tasks and classifying local tasks based on the portability; each data center park sending reporting information to a collaborative scheduling platform; wherein the reporting information includes: computing power carrying capacity index, portability, classification results of local tasks, and comprehensive credit score; the collaborative scheduling platform performing many-to-many dynamic matching of each data center park based on the reporting information; each data center park executing computing power migration and computation tasks based on the matching results of the many-to-many dynamic matching, and updating the comprehensive credit score of the receiving party based on the execution results.

[0007] In an optional embodiment of this application, the step of calculating the computing power carrying capacity index of each of the above-mentioned data center parks includes: each data center park calculates its own computing power carrying capacity index based on its own computing power surplus index, energy surplus index and cross-domain transmission cost index.

[0008] In optional embodiments of this application, the above method further includes: each data center park calculating its own remaining computing power index based on the park's total computing power rating, the computing power reserved for latency-sensitive tasks, and the local running computing power; each data center park calculating its own energy surplus index based on the electricity price at the target time, the high value of the electricity price, the total rated power of the data center park's energy system, the predicted value of photovoltaic system power generation, the adjustment margin of the energy storage system at the target time, and the adjustment margin of the cooling system at the target time; and each data center park calculating its own cross-domain transmission cost index based on the maximum allowable network latency, the baseline bandwidth cost, the baseline data migration energy consumption, the real-time network latency between different parks, the real-time bandwidth unit price between different parks, and the migration energy consumption per unit of data volume.

[0009] In optional embodiments of this application, the step of calculating the portability of local tasks for each data center campus includes: each data center campus calculating the portability of local tasks based on the remaining deadline, minimum allowed deadline, computational load, baseline computational load, data volume, and baseline data volume of the computing task; the step of classifying local tasks based on portability includes: if the portability is greater than or equal to a preset first threshold, classifying the local task as a highly portable task; if the portability is less than the first threshold but greater than or equal to a preset second threshold, classifying the local task as a medium-portable task; if the portability is less than the second threshold, classifying the local task as a low-portable task.

[0010] In an optional embodiment of this application, the above-mentioned reported information includes: the task provider's quotation and the recipient's quotation; the task provider's quotation includes: the unit computing power quotation, computing volume, portability, and cross-domain transmission cost index of each highly portable task; the recipient's quotation includes: the unit computing power acceptance quotation of the park, the remaining computing power, the computing power carrying capacity index, the current period's committed green electricity ratio, and the comprehensive credit score.

[0011] In an optional embodiment of this application, the step of the above-mentioned collaborative scheduling platform performing many-to-many dynamic matching of various data center parks based on reported information includes: the collaborative scheduling platform initializing a matching pool based on the reported information; wherein, the matching pool includes: a pool of valid recipients and a pool of valid task providers; the collaborative scheduling platform determining multiple combinations of task providers and recipients based on the matching pool, and calculating a comprehensive priority score for each combination based on the reported information; the collaborative scheduling platform sorting multiple combinations in descending order of comprehensive priority score, and matching tasks to each combination sequentially based on the remaining computing power of the task provider and the remaining computing power of the recipient; the collaborative scheduling platform sending the matching results to the corresponding task provider and recipient.

[0012] In an optional embodiment of this application, the step of updating the comprehensive credit score of the contractor based on the execution result includes: if the execution result indicates that the contractor has not completed the task on time, updating the comprehensive credit score of the contractor based on the actual delay time of the task; if the execution result indicates that the contractor's actual green electricity ratio is higher than the promised green electricity ratio, updating the comprehensive credit score of the contractor based on the actual green electricity ratio and the promised green electricity ratio.

[0013] In optional embodiments of this application, the above method further includes: the collaborative scheduling platform calculates the global coordination degree based on energy efficiency score, economic score, carbon emission score, and service quality score, and performs adaptive resource reservation adjustment based on the global coordination degree.

[0014] In an optional embodiment of this application, the above-mentioned step of adaptive resource reservation adjustment based on global coordination includes: if the global coordination is greater than a preset threshold, calculating the historical load volatility of each park; calculating the basic reservation ratio of each park based on the historical load volatility of each park; updating the reserved computing power for latency-sensitive tasks based on the nature of each park and the basic reservation ratio of each park; wherein, the nature of the park includes: parks with surplus energy and computing power, and parks with tight energy and computing power.

[0015] Secondly, embodiments of the present invention also provide a computing power collaborative intelligent scheduling device for multiple interconnected data center parks. The device includes: a computing power carrying capacity index calculation module, used for each data center park to calculate its own computing power carrying capacity index; a portability calculation module, used for each data center park to calculate the portability of local tasks and classify local tasks based on portability; a reporting information sending module, used for each data center park to send reporting information to the collaborative scheduling platform; wherein the reporting information includes: computing power carrying capacity index, portability, classification results of local tasks, and comprehensive credit score; a many-to-many dynamic matching module, used for the collaborative scheduling platform to perform many-to-many dynamic matching of each data center park based on the reporting information; and a computing power migration and computing task execution module, used for each data center park to execute computing power migration and computing tasks based on the matching results of the many-to-many dynamic matching, and update the comprehensive credit score of the receiving party based on the execution results.

[0016] The embodiments of the present invention bring the following beneficial effects: This invention provides a computing and power collaborative intelligent scheduling method and device for multi-interconnected data center parks. It proposes a standardized computing and power carrying capacity index and a continuous task portability index to achieve quantitative characterization of park status and task characteristics. This not only protects the commercial privacy of the park, but also provides a unified decision-making basis for cross-park collaboration.

[0017] Other features and advantages of this disclosure will be set forth in the following description, or some features and advantages may be inferred from the description or determined without doubt, or may be learned by practicing the techniques described above.

[0018] To make the above-mentioned objects, features and advantages of this disclosure more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0019] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0020] Figure 1 A flowchart of a computing and power collaborative intelligent scheduling method for a multi-interconnected data center campus provided in an embodiment of the present invention; Figure 2 A schematic diagram of a many-to-many dynamic matching algorithm provided in an embodiment of the present invention; Figure 3A flowchart of another intelligent scheduling method for computing and power coordination in a multi-interconnected data center campus provided by an embodiment of the present invention; Figure 4 This is a schematic diagram of the structure of a computing and power collaborative intelligent scheduling device for a multi-interconnected data center park, provided in an embodiment of the present invention. Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Currently, existing research generally employs centralized optimization frameworks such as model predictive control and reinforcement learning. These frameworks require the collection of detailed operational data from all parks, including commercially sensitive information such as computing power costs, energy reserves, and equipment parameters. This not only results in an explosive increase in computational load and high communication latency, but also often hinders practical implementation due to park operators' reluctance to disclose core commercial data. On the other hand, most existing methods use fixed rules or offline-trained models for task matching, which cannot adapt to dynamic scenarios such as fluctuations in photovoltaic output, real-time changes in electricity prices, and sudden arrival of tasks, making them prone to local optima or even system imbalance.

[0023] Based on this, the present invention provides a computing and power collaborative intelligent scheduling method and device for multi-interconnected data center parks, which can be applied to computing and power collaborative scheduling in data center parks.

[0024] To facilitate understanding of this embodiment, a detailed description of a computing and power collaborative intelligent scheduling method for a multi-interconnected data center campus disclosed in this embodiment of the invention will be provided first.

[0025] Example 1: This invention provides a computing and power collaborative intelligent scheduling method for multi-interconnected data center campuses. See [link to relevant documentation]. Figure 1 The flowchart shown illustrates a computing power collaborative intelligent scheduling method for a multi-interconnected data center campus. This method includes the following steps: Step S102: Each data center campus calculates its own computing power carrying capacity index.

[0026] In this embodiment, each data center campus independently calculates its own computing power capacity index. .

[0027] In some embodiments, each data center campus calculates its own computing power carrying capacity index based on its own computing power surplus index, energy surplus index, and cross-domain transmission cost index.

[0028] In this embodiment, the computing power carrying capacity index of each data center campus can be calculated using the following formula. : ; In the formula, This is the remaining computing power index. As an energy surplus index, This is a cross-domain transmission cost index. , and These are the corresponding weighting factors, and the sum of the three is 1.

[0029] In engineering, a set of options , and The weighting coefficients are 0.35, 0.45, and 0.2.

[0030] Traditional multi-campus scheduling methods lack unified cross-campus collaborative decision-making indicators: they either only assess computing power load rate, ignoring the energy system's capacity, or only assess energy status, ignoring computing resource availability; furthermore, they fail to consider the impact of cross-campus transmission costs on task migration feasibility. Simultaneously, traditional centralized methods require exposing all detailed operational data for global evaluation, posing a serious risk of privacy breaches. This embodiment proposes, for the first time, a standardized three-dimensional comprehensive index by calculating the power carrying capacity index. This index quantifies the three core dimensions determining task migration feasibility—remaining computing power, dynamic energy carrying capacity, and cross-domain transmission costs—into a unified comprehensive indicator. Each campus only needs to calculate the index locally and then publish the value, fundamentally protecting the campus's commercial privacy and providing a unified and fair decision-making basis for cross-campus task matching.

[0031] In some embodiments, each data center campus calculates its own remaining computing power index based on the campus's total rated computing power, reserved computing power for latency-sensitive tasks, and locally running computing power; each data center campus calculates its own energy surplus index based on the target time's electricity price, the high value of the electricity price, the total rated power of the data center campus's energy system, the predicted value of photovoltaic system power generation, the adjustment margin of the energy storage system at the target time, and the adjustment margin of the cooling system at the target time; each data center campus calculates its own cross-domain transmission cost index based on the maximum allowable network latency, the baseline bandwidth cost, the baseline data migration energy consumption, the real-time network latency between different campuses, the real-time bandwidth unit price between different campuses, and the migration energy consumption per unit of data volume.

[0032] 1. The remaining computing power index in this embodiment The calculation is as follows: ; In the formula, This is the rated value of the total computing power of the park. Reserve computing power for latency-sensitive tasks. This refers to the local computing power already in operation, including latency-sensitive tasks and allocated latency-tolerant tasks. In engineering terms, The initial value can be set to .

[0033] This indicator reflects the real-time remaining computing power capacity within the park, with a value range of [0, 1]. The larger the value, the more computing power can be used to undertake cross-park tasks.

[0034] 2. Energy surplus index in this embodiment The calculation is as follows: ; In the formula, for Electricity price at any time For high electricity prices, the engineering practice can use 1.2-2 times the peak electricity price. The total rated power of the data center campus energy system, for Predicted power generation of photovoltaic systems at any given time, in engineering Take 5 minutes. and They are respectively Adjustment margin of energy storage system and cooling system at all times.

[0035] This indicator reflects the current energy dispatch capacity of the park available for computing power migration: the first item is the normalized local electricity price, which is higher the lower the price; the second item is the normalized dispatchable energy volume of the park, which is higher the greater the dispatchable energy volume of photovoltaic, energy storage, and cooling systems. In summary, the energy surplus index ranges from [0, 1]. The higher the value, the greater the dispatchable energy volume and the lower the electricity price of the park, and the greater its capacity to undertake computing power tasks.

[0036] Specifically, the margin index is calculated as follows: ; ; In the formula, The rated capacity of the energy storage system, for The energy storage status of the energy storage system at all times. This represents the minimum value of the energy storage state. The rated power of the cooling system. for The actual power of the cooling system at any given time. The margin of the energy storage system is determined by its current State of Charge (SOC). The larger the SOC, the more electrical energy it stores, and the stronger its discharge capacity and load-bearing capacity. The margin of the cooling system is determined by its current power. The lower the current cooling power, the more the cooling system can increase its power to handle the increased cooling load brought about by the migration of computing tasks, that is, the stronger its load-bearing capacity.

[0037] 3. Cross-domain transmission cost index in this embodiment The calculation is as follows: ; In the formula, For the maximum allowable network latency, Based on the baseline bandwidth cost, Energy consumption for data baseline migration For the park and Real-time network latency between them For the park and Real-time bandwidth unit price between Energy consumption per unit of data migration , and The corresponding weighting coefficient is denoted as . This indicator quantifies the comprehensive cost of transmitting computing power tasks between parks, including network latency, bandwidth cost, and migration energy consumption. The value range is [0, 1]. The larger the value, the higher the transmission cost, and the corresponding decrease in computing power capacity.

[0038] Step S104: Each data center campus calculates the portability of local tasks and classifies local tasks based on the portability.

[0039] In this embodiment, each data center campus performs a portability assessment of local tasks. calculate.

[0040] In some embodiments, each data center campus calculates the portability of local tasks based on the remaining deadline, minimum allowed deadline, computational load, baseline computational load, data volume, and baseline data volume of the computing task.

[0041] This embodiment calculates the portability. The formula is as follows: ; In the formula, For the Sigmoid function, This represents the remaining deadline for the computing task. The minimum allowable deadline. The computational load of the computing task. As the baseline calculation quantity, The amount of data for computing power tasks, As the baseline data volume, , , For coefficient factors.

[0042] In this embodiment, the physical meaning of the portability index is that the longer the deadline of the computing task and the larger the amount of computation and data, the higher the expected benefit of cross-campus migration, meaning the computing task is considered "higher quality." In engineering terms, , , It can be set to 5, 3, or 2.

[0043] In some embodiments, if the portability is greater than or equal to a preset first threshold, the local task is classified as a highly portable task; if the portability is less than the first threshold and the portability is greater than or equal to a preset second threshold, the local task is classified as a medium portable task; if the portability is less than the second threshold, the local task is classified as a low portable task.

[0044] In calculating portability Then, grading can be performed. In this embodiment, a first threshold can be set to 0.7, and a second threshold to 0.3. For example: if... If it is classified as a highly transferable task, it will be given priority in participating in subsequent cross-park auctions; if The task is classified as a medium-portable task, and participation in cross-park auctions can be selected based on the needs of park operations and maintenance personnel; if If so, it is classified as a low-portability task and forced to be processed locally.

[0045] The setting of the first threshold and the second threshold in this embodiment is not unique. The values ​​of 0.7 and 0.3 are only preferred options, and this embodiment does not limit them.

[0046] Traditional methods simply categorize tasks into "latency-sensitive" and "latency-tolerant" types, employing a binary scheduling strategy that fails to achieve fine-grained scheduling. This embodiment designs a continuous transferability index that comprehensively considers three core characteristics of a task: deadline, computational load, and data volume. By quantifying the transferability of a task with continuous values, it can formulate optimal scheduling strategies based on the characteristics of different tasks.

[0047] In step S106, each data center campus sends reporting information to the collaborative scheduling platform; the reporting information includes: computing power carrying capacity index, portability, local task classification results, and comprehensive credit score.

[0048] In this embodiment, each data center campus can send reporting information to the collaborative scheduling platform. The collaborative scheduling platform in this embodiment can be a scheduling center or a transaction center jointly recognized by the interconnected campuses.

[0049] In some embodiments, the reported information includes: the task provider's quote and the service provider's quote; the task provider's quote includes: the unit computing power quote for each highly portable task. Calculation workload Migration Cross-domain transmission cost index with each valid recipient The contractor's quote includes: the unit computing power quote for the park. Remaining computing power Power carrying capacity index The current period's commitment to green electricity ratio and comprehensive credit score .

[0050] In this embodiment, each park does not need to upload any critical business data (such as the overall cost of local processing unit computing power within the park). This implementation only requires transmitting the previously calculated index and final price, thus protecting privacy. Simultaneously, this embodiment converts default penalties and green electricity rewards into credit scores and integrates them into the pricing system. Higher credit scores or a higher proportion of green electricity result in lower prices and easier matching; conversely, lower credit scores or a lower proportion of green electricity result in higher prices and more difficult matching.

[0051] The task-side unit computing power pricing in this embodiment The calculation is as follows: ; In the formula, The overall cost of computing power for local processing units in the park. For task urgency premium, it can be based on The specific size is set, preferably when... The value can be set to 0.3 if the condition is met, and 0.1 otherwise. Additionally, the task provider can set a price cap based on their own circumstances to avoid excessive migration costs.

[0052] The unit computing power quotation of the contractor in this embodiment The calculation is as follows: ; In the formula, For green electricity discounts, the engineering standard can be set at 0.3, meaning that for every 10% increase in the proportion of green electricity, the price will decrease by 3%.

[0053] In step S108, the collaborative scheduling platform performs multi-to-multi dynamic matching of various data center parks based on the reported information.

[0054] In this embodiment, the collaborative scheduling platform can execute a many-to-many dynamic matching algorithm based on the information reported by each data center campus.

[0055] See Figure 2 The diagram shown illustrates a many-to-many dynamic matching algorithm, which mainly includes the following steps 1-4: Step 1: The collaborative scheduling platform initializes the matching pool based on the reported information; the matching pool includes: a pool of valid acceptors and a pool of valid task providers.

[0056] In this embodiment, the matching pool can be initialized. The computational power carrying capacity index is filtered out from the recipients. Less than the lower limit (0.3 can be used in engineering) and comprehensive credit score Parks with a migrationability score below the lower limit (0.6 is acceptable in engineering) form an effective pool of contractors; those with a migrationability score are filtered out from the contractors. Park tasks with a value less than 0.3 form an effective task pool.

[0057] Step 2: The collaborative scheduling platform determines the combination of multiple task parties and recipients based on the matching pool, and calculates the comprehensive priority score of each combination based on the reported information.

[0058] This embodiment can be applied to each task-responsible combination. (Task Belongs to the park ), calculate the comprehensive priority score : ; Then, filter out The combination of .

[0059] Step 3: The collaborative scheduling platform sorts the multiple combinations in descending order of comprehensive priority score, and matches tasks to each combination in turn based on the remaining computing power of the task party and the remaining computing power of the receiving party.

[0060] This embodiment can sort all valid combinations from highest to lowest score and process each combination sequentially: If the task has remaining computation time And the remaining computing power of the receiving party The allocation amount is calculated according to the following formula. : ; Generate a new task matching record, i.e., the park Task Assigned to the park The allocation amount is After that, the remaining computational load of the task-taking party and the remaining computing power of the receiving party will be deducted from the allocated amount respectively.

[0061] If the task has remaining computation time If the task is not found, it will be removed from the available task pool; if the receiving party has remaining computing power... If the park is removed from the valid acceptance pool, then the park will be removed from the pool.

[0062] Repeat the above process until the valid combination traversal is completed or the task pool is empty.

[0063] Step 4: The collaborative scheduling platform sends the matching results to the corresponding task party and the recipient.

[0064] In this embodiment, the matching results can be recorded in the task matching record table, and the matching results can be sent to the corresponding task party and the recipient according to the task matching record table.

[0065] Traditional centralized optimization methods require global search, and the computational cost increases exponentially with the number of parks. This embodiment proposes a bidirectional auction matching algorithm suitable for distributed environments. Based on greedy matching and priority sorting, this algorithm has low computational cost, fast convergence, and can meet the requirements of real-time scheduling. Furthermore, the bidirectional auction mechanism guarantees the interests of both parties, incentivizing parks to actively participate in cross-park collaboration.

[0066] The core basis of the two-way auction matching algorithm in this embodiment is a comprehensive priority score, which is determined by the price difference. Green electricity ratio and transmission costs The scores are determined jointly by: the greater the price difference and the higher the benefits for both parties; the higher the proportion of green electricity and the lower the carbon emissions; and the lower the transmission costs and the smaller the migration costs.

[0067] In step S110, each data center park performs computing power migration and computation tasks based on the matching results of many-to-many dynamic matching, and updates the comprehensive credit score of the receiving party based on the execution results.

[0068] In this embodiment, each data center campus executes computing power migration and computation tasks based on the matching results. After the tasks are completed, the comprehensive credit score of each receiving party is updated based on the execution results.

[0069] In some embodiments, if the execution result indicates that the contractor has not completed the task on time, the contractor's comprehensive credit score is updated based on the actual delay time of the task; if the execution result indicates that the contractor's actual green electricity ratio is higher than the promised green electricity ratio, the contractor's comprehensive credit score is updated based on the actual green electricity ratio and the promised green electricity ratio.

[0070] If the contractor fails to complete the task on time, the overall credit score will be updated as follows: ; In the formula, This refers to the actual delay time of the task. This is the delay factor, which can be set to 0.2 in engineering practice.

[0071] If the actual percentage of green electricity used by the contractor is higher than the committed value, the overall credit score will be updated as follows: ; In the formula, and These represent the actual and committed percentages of green electricity, respectively. This is the green electricity ratio coefficient, which can be set to 0.1 in engineering projects.

[0072] Generally, the comprehensive credit score in this embodiment can be set to a range of 0.5-1.5, and can be reset periodically according to the assessment cycle (such as monthly, quarterly, annual, etc.).

[0073] The comprehensive credit score update mechanism proposed in this embodiment takes into account both penalties for breach of contract and rewards for green electricity. It can be dynamically adjusted according to the actual performance of the receiving park, thus both punishing breach of contract and rewarding contributions to green electricity, thereby ensuring the long-term stable operation of the system.

[0074] This invention provides a computing and power collaborative intelligent scheduling method and device for multi-interconnected data center parks. It proposes a standardized computing and power carrying capacity index and a continuous task portability index to achieve quantitative characterization of park status and task characteristics. This not only protects the commercial privacy of the park, but also provides a unified decision-making basis for cross-park collaboration.

[0075] The method provided in this embodiment of the invention also developed a bidirectional auction matching algorithm for computing power and energy, which transforms default penalties and green electricity incentives into credit scores and directly integrates them into the bidding system, achieving global optimal matching in a distributed scenario without the need for a central node to collect sensitive data.

[0076] Example 2: This invention provides another intelligent scheduling method for computing and power coordination in multi-interconnected data center campuses, implemented based on the aforementioned embodiments. The focus is on describing the specific methods of the adaptive resource reservation and adjustment strategy. (See [link to relevant documentation]). Figure 3 The flowchart shown represents another intelligent scheduling method for computing and power coordination in a multi-interconnected data center campus. This method includes the following steps: Step S302: Each data center campus calculates its own computing power carrying capacity index.

[0077] Step S304: Each data center campus calculates the portability of local tasks and classifies local tasks based on the portability.

[0078] In step S306, each data center campus sends reporting information to the collaborative scheduling platform; the reporting information includes: computing power carrying capacity index, portability, local task classification results, and comprehensive credit score.

[0079] Step S308: The collaborative scheduling platform performs multi-to-multi dynamic matching of various data center parks based on the reported information.

[0080] In step S310, each data center park performs computing power migration and computation tasks based on the matching results of many-to-many dynamic matching, and updates the comprehensive credit score of the receiving party based on the execution results.

[0081] In step S312, the collaborative scheduling platform calculates the global coordination degree based on energy efficiency score, economic score, carbon emission score, and service quality score, and performs adaptive resource reservation adjustment based on the global coordination degree.

[0082] This embodiment also establishes an adaptive resource reservation adjustment strategy to correct the resource reservation parameters of key unbalanced parks, overcoming the local optimum problem under distributed scheduling.

[0083] This embodiment can calculate the global coordination degree at regular intervals and execute an adaptive resource reservation adjustment strategy. Global Coordination Degree The calculation is as follows: ; In the formula, β is a weighting factor, which can be set to 0.3, 0.3, 0.2, and 0.2 in engineering practice. The following values ​​represent the energy efficiency score, economic performance score, carbon emission score, and service quality score, respectively. 1. Energy efficiency score ,in The current average PUE of the interconnected park, Baseline PUE; 2. Economic score ,in The current average operating cost of the interconnected park, Baseline operating costs; 3. Carbon emission score ,in This represents the current average carbon emissions of the interconnected park. As a benchmark carbon emission level; 4. Service quality score ,in The latency of the computing task. This is an overview of computing power tasks.

[0084] In some embodiments, if the global coordination degree is greater than a preset threshold, the historical load volatility of each park is calculated; the basic reservation ratio of each park is calculated based on the historical load volatility of each park; the reserved computing power for latency-sensitive tasks is updated based on the nature of each park and the basic reservation ratio of each park; wherein, the nature of the park includes: parks with surplus energy and computing power, and parks with tight energy and computing power.

[0085] In this embodiment, the threshold can be set to 0.5. If If no correction is performed, then an adaptive resource reservation adjustment strategy will be implemented. First, calculate the historical load volatility of each park. : ; In the formula, The historical average computing power utilization rate of the Internet Park For historical sample size, for The computing power utilization rate of the park.

[0086] Then, calculate the basic reserve ratio. ; Finally, update the reserved computing power for delay-sensitive tasks according to the nature of the park. If it is a park with surplus energy / computing power, If it is a park with energy / computing power shortages, .

[0087] Among them, energy / computing power surplus parks refer to parks with a large amount of low-priced green electricity or severely limited computing power. The criteria for judgment can be: computing power carrying capacity index greater than 1.2 and computing power utilization rate less than 25%; energy / computing power shortage parks refer to parks with insufficient energy or insufficient computing power. The criteria for judgment can be: computing power carrying capacity index less than 0.6 and computing power utilization rate greater than 85%.

[0088] The values ​​involved in the adaptive resource reservation adjustment strategy in this embodiment (including 0.15, 0.1, 0.8, and 1.2) are not fixed and can be adjusted according to the actual system characteristics. This embodiment does not limit this.

[0089] This embodiment proposes a global synergy index, which can be used to determine whether the current distributed strategy can achieve global optimality, avoiding individual distributed decisions from focusing excessively on local optima and thus deviating from global optimality. This index quantifies the overall operational performance of the interconnected park from four aspects: energy efficiency, economics, carbon emissions, and service quality, and normalizes it to a range of 0-1. This embodiment further proposes an adaptive resource reservation adjustment strategy, which is essentially used to correct the reserved computing power for latency-sensitive tasks. This reflects the system computing power reserved by the park itself for subsequent highly random computing tasks. This strategy first updates based on the historical load volatility of each park. Higher volatility indicates a greater likelihood of sudden tasks occurring in the park, thus requiring a corresponding increase in the basic reserved ratio to enhance reserved computing power. Conversely, lower volatility indicates a more stable park, allowing for a reduction in reserved computing power to fully release computing capacity. Furthermore, this strategy adjusts the reserved computing power based on the park's characteristics. For parks with surplus energy / computing power, high reliability and redundancy, the reserved computing power can be appropriately reduced; for parks with strained energy / computing power, it needs to be increased to ensure sufficient reserved computing power.

[0090] In summary, this invention provides a computing and power collaborative intelligent scheduling method for multi-interconnected data center parks, mainly providing the following: 1. A standardized computing power carrying capacity index and a continuous task portability index were proposed to achieve quantitative characterization of park status and task characteristics, which not only protects the park's business privacy, but also provides a unified decision-making basis for cross-park collaboration.

[0091] 2. A bidirectional auction matching algorithm for "computing power-energy" was developed, which transforms default penalties and green electricity incentives into credit scores and directly integrates them into the bidding system, achieving global optimal matching in a distributed scenario without the need for a central node to collect sensitive data.

[0092] 3. An adaptive resource reservation adjustment strategy was established to correct the resource reservation parameters of key unbalanced parks, overcoming the local optimum problem under distributed scheduling.

[0093] Example 3: Corresponding to the above method embodiments, this invention provides a computing and power collaborative intelligent scheduling device for multi-interconnected data center parks. See [link to relevant documentation]. Figure 4 The diagram shows a structural schematic of a computing power collaborative intelligent scheduling device for a multi-interconnected data center campus. This device includes: The computing power carrying capacity index calculation module 41 is used to calculate the computing power carrying capacity index of each data center campus. The portability calculation module 42 is used to calculate the portability of local tasks in each data center campus and classify local tasks based on the portability. The reporting information sending module 43 is used by each data center campus to send reporting information to the collaborative scheduling platform; the reporting information includes: computing power carrying capacity index, portability, local task classification results and comprehensive credit score; The many-to-many dynamic matching module 44 is used by the collaborative scheduling platform to perform many-to-many dynamic matching of various data center parks based on the reported information. The computing power migration and operation task execution module 45 is used to execute computing power migration and operation tasks based on the matching results of many-to-many dynamic matching in various data center parks, and update the comprehensive credit score of the undertaking party based on the execution results.

[0094] This invention provides a computing power collaborative intelligent scheduling device for multi-interconnected data center parks. It proposes a standardized computing power carrying capacity index and a continuous task portability index to achieve quantitative characterization of park status and task characteristics. This not only protects the commercial privacy of the park, but also provides a unified decision-making basis for cross-park collaboration.

[0095] The aforementioned computing power carrying capacity index calculation module is used by each data center campus to calculate its own computing power carrying capacity index based on its own computing power surplus index, energy surplus index, and cross-domain transmission cost index.

[0096] The aforementioned computing power carrying capacity index calculation module is also used to calculate the remaining computing power index of each data center park based on the park's total rated computing power, the reserved computing power for latency-sensitive tasks, and the local running computing power; to calculate the energy surplus index of each data center park based on the electricity price at the target time, the high value of the electricity price, the total rated power of the data center park's energy system, the predicted value of photovoltaic system power generation, the adjustment margin of the energy storage system at the target time, and the adjustment margin of the cooling system at the target time; and to calculate the cross-domain transmission cost index of each data center park based on the maximum allowable network latency, the baseline bandwidth cost, the baseline data migration energy consumption, the real-time network latency between different parks, the real-time bandwidth unit price between different parks, and the migration energy consumption per unit of data volume.

[0097] The aforementioned portability calculation module is used by each data center campus to calculate the portability of local tasks based on the remaining deadline, minimum allowed deadline, computational load, baseline computational load, data volume, and baseline data volume of the computing task. This portability calculation module is used to classify local tasks as highly portable if the portability is greater than or equal to a preset first threshold; classify local tasks as medium portable if the portability is less than the first threshold but greater than or equal to a preset second threshold; and classify local tasks as low portable if the portability is less than the second threshold.

[0098] The above-mentioned reported information includes: the task provider's quotation and the contractor's quotation; the task provider's quotation includes: the unit computing power quotation, computing volume, portability, and cross-domain transmission cost index of each highly portable task; the contractor's quotation includes: the unit computing power contracting quotation of the park, remaining computing power, computing power carrying capacity index, the current period's committed green electricity ratio, and comprehensive credit score.

[0099] The aforementioned many-to-many dynamic matching module is used by the collaborative scheduling platform to initialize the matching pool based on the reported information. The matching pool includes a pool of valid task providers and a pool of valid task providers. The collaborative scheduling platform determines multiple combinations of task providers and recipients based on the matching pool, and calculates the comprehensive priority score for each combination based on the reported information. The collaborative scheduling platform sorts the multiple combinations in descending order of their comprehensive priority scores, and matches tasks to each combination sequentially based on the remaining computational load of the task provider and the remaining computing power of the recipient. The collaborative scheduling platform then sends the matching results to the corresponding task provider and recipient.

[0100] The aforementioned computing power migration and computation task execution module is used to update the comprehensive credit score of the service provider based on the actual delay time if the execution result indicates that the service provider has not completed the task on time; and to update the comprehensive credit score of the service provider based on the actual green electricity ratio and the promised green electricity ratio if the execution result indicates that the service provider's actual green electricity ratio is higher than the promised green electricity ratio.

[0101] The aforementioned device also includes: an adaptive resource reservation adjustment module, used by the collaborative scheduling platform to calculate the global coordination degree based on energy efficiency score, economic score, carbon emission score, and service quality score, and to perform adaptive resource reservation adjustment based on the global coordination degree.

[0102] The aforementioned adaptive resource reservation adjustment module is used to calculate the historical load volatility of each park if the global coordination degree is greater than a preset threshold; calculate the basic reservation ratio of each park based on the historical load volatility of each park; and update the reserved computing power for latency-sensitive tasks based on the nature of each park and the basic reservation ratio of each park. The nature of the parks includes: parks with surplus energy and computing power, and parks with tight energy and computing power.

[0103] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the computing and power collaborative intelligent scheduling device for multi-interconnected data center parks described above can be referred to the corresponding process in the embodiments of the aforementioned computing and power collaborative intelligent scheduling method for multi-interconnected data center parks, and will not be repeated here.

[0104] Example 4: This invention also provides an electronic device for running the aforementioned intelligent scheduling method for computing and power coordination in multi-interconnected data center campuses; see also Figure 5The diagram shows the structure of an electronic device, which includes a memory 100 and a processor 101. The memory 100 is used to store one or more computer instructions, which are executed by the processor 101 to realize the above-mentioned intelligent scheduling method for computing and power coordination in a multi-interconnected data center park.

[0105] Furthermore, Figure 5 The electronic device shown also includes a bus 102 and a communication interface 103, with the processor 101, the communication interface 103 and the memory 100 connected via the bus 102.

[0106] The memory 100 may include high-speed random access memory (RAM) and may also include non-volatile memory, such as at least one disk storage device. Communication between this system network element and at least one other network element is achieved through at least one communication interface 103 (which can be wired or wireless), such as the Internet, wide area network, local area network, metropolitan area network, etc. The bus 102 may be an ISA bus, PCI bus, or EISA bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 5 The symbol is represented by a single double-headed arrow, but this does not mean that there is only one bus or one type of bus.

[0107] Processor 101 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuitry in the hardware of processor 101 or by instructions in software form. Processor 101 can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this invention. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this invention can be directly manifested as execution by a hardware decoding processor, or execution by a combination of hardware and software modules in the decoding processor. The software module can reside in a readily available storage medium in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory 100, and processor 101 reads information from memory 100 and, in conjunction with its hardware, completes the steps of the method described in the foregoing embodiments.

[0108] This invention also provides a computer-readable storage medium storing computer-executable instructions. When these computer-executable instructions are invoked and executed by a processor, they cause the processor to implement the aforementioned intelligent scheduling method for computing and power coordination in a multi-interconnected data center campus. For specific implementation details, please refer to the method embodiments, which will not be repeated here.

[0109] The computer program product of the computing and power collaborative intelligent scheduling method and device for multi-interconnected data center parks provided in this embodiment of the invention includes a computer-readable storage medium storing program code. The instructions included in the program code can be used to execute the methods in the preceding method embodiments. For specific implementation, please refer to the method embodiments, which will not be repeated here.

[0110] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the system and / or device described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0111] Furthermore, in the description of the embodiments of the present invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in the present invention based on the specific circumstances.

[0112] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0113] In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0114] Finally, it should be noted that the above-described embodiments are merely specific implementations of the present invention, used to illustrate the technical solutions of the present invention, and not to limit it. The scope of protection of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments within the technical scope disclosed in the present invention, or make equivalent substitutions for some of the technical features; and these modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A computing and power collaborative intelligent scheduling method for multi-interconnected data center parks, characterized in that, The method includes: Each data center campus calculates its own computing power carrying capacity index; Each of the aforementioned data center campuses calculates the portability of local tasks and classifies the local tasks based on the portability. Each of the aforementioned data center parks sends reporting information to the collaborative scheduling platform; wherein, the reporting information includes: the computing power carrying capacity index, the portability, the classification result of the local task, and the comprehensive credit score; The collaborative scheduling platform performs multi-to-many dynamic matching of each of the data center parks based on the reported information; Each of the aforementioned data center parks performs computing power migration and computation tasks based on the matching results of many-to-many dynamic matching, and updates the comprehensive credit score of the receiving party based on the execution results.

2. The method according to claim 1, characterized in that, The steps for each data center campus to calculate its own computing power capacity index include: Each data center campus calculates its own computing power carrying capacity index based on its own computing power surplus index, energy surplus index, and cross-domain transmission cost index.

3. The method according to claim 2, characterized in that, The method further includes: Each of the aforementioned data center parks calculates its own remaining computing power index based on the park's total computing power rating, the computing power reserved for latency-sensitive tasks, and the local computing power already in operation. Each of the aforementioned data center parks calculates its own energy surplus index based on the target time's electricity price, the high value of the electricity price, the total rated power of the data center park's energy system, the predicted value of photovoltaic system power generation, the adjustment margin of the energy storage system at the target time, and the adjustment margin of the cooling system at the target time. Each of the aforementioned data center campuses calculates its own cross-domain transmission cost index based on the maximum allowable network latency, baseline bandwidth cost, baseline data migration energy consumption, real-time network latency between different campuses, real-time bandwidth unit price between different campuses, and migration energy consumption per unit of data volume.

4. The method according to claim 1, characterized in that, The steps for each of the data center campuses to calculate the portability of local tasks include: each of the data center campuses calculating the portability of local tasks based on the remaining deadline of the computing power task, the minimum allowed deadline, the computing power task's computing load, the baseline computing load, the computing power task's data volume, and the baseline data volume. The step of classifying the local task based on the portability includes: classifying the local task as a high-portable task if the portability is greater than or equal to a preset first threshold; classifying the local task as a medium-portable task if the portability is less than the first threshold and the portability is greater than or equal to a preset second threshold; and classifying the local task as a low-portable task if the portability is less than the second threshold.

5. The method according to claim 1, characterized in that, The reported information includes: the task provider's quotation and the contractor's quotation; The task provider's quote includes: the unit computing power quote, computing volume, portability, and cross-domain transmission cost index for each highly portable task. The bid from the contractor includes: the unit computing power bid for the park, the remaining computing power, the computing power carrying capacity index, the current period's committed green electricity ratio, and the comprehensive credit score.

6. The method according to claim 5, characterized in that, The collaborative scheduling platform performs a many-to-many dynamic matching of each of the data center parks based on the reported information, including: The collaborative scheduling platform initializes the matching pool based on the reported information; wherein, the matching pool includes: a pool of valid recipients and a pool of valid task providers; The collaborative scheduling platform determines combinations of multiple task parties and recipients based on the matching pool, and calculates a comprehensive priority score for each combination based on the reported information. The collaborative scheduling platform sorts the multiple combinations in descending order of the comprehensive priority score, and matches tasks to each combination in turn based on the remaining computing power of the task party and the remaining computing power of the receiving party. The collaborative scheduling platform will send the matching results to the corresponding task party and the recipient.

7. The method according to claim 1, characterized in that, The steps for updating the comprehensive credit score of the recipient based on the execution result include: If the execution result indicates that the contractor has not completed the task on time, the contractor's comprehensive credit score will be updated based on the actual delay time of the task. If the execution result indicates that the actual green electricity ratio of the contractor is higher than the promised green electricity ratio, the contractor's comprehensive credit score is updated based on the actual green electricity ratio and the promised green electricity ratio.

8. The method according to any one of claims 1-7, characterized in that, The method further includes: The collaborative scheduling platform calculates the global coordination degree based on energy efficiency score, economic score, carbon emission score, and service quality score, and performs adaptive resource reservation adjustments based on the global coordination degree.

9. The method according to claim 8, characterized in that, The steps for adaptive resource reservation adjustment based on the global coordination degree include: If the global coordination degree is greater than the preset threshold, calculate the historical load volatility of each park. The basic reserve ratio for each park is calculated based on the historical load volatility of each park. The reserved computing power for delay-sensitive tasks is updated based on the nature of each park and the basic reserved ratio of each park; wherein, the nature of the parks includes: parks with surplus energy and computing power, and parks with shortage of energy and computing power.

10. A computing and power collaborative intelligent scheduling device for a multi-interconnected data center campus, characterized in that, The device includes: The computing power carrying capacity index calculation module is used to calculate the computing power carrying capacity index of each data center campus. The portability calculation module is used to calculate the portability of local tasks in each of the data center campuses and classify the local tasks based on the portability. The reporting information sending module is used for each of the data center parks to send reporting information to the collaborative scheduling platform; wherein, the reporting information includes: the computing power carrying capacity index, the portability, the classification result of the local task, and the comprehensive credit score; A many-to-many dynamic matching module is used by the collaborative scheduling platform to perform many-to-many dynamic matching of each of the data center parks based on the reported information. The computing power migration and computation task execution module is used to execute computing power migration and computation tasks in each of the data center parks based on the matching results of many-to-many dynamic matching, and update the comprehensive credit score of the receiving party based on the execution results.

Images