Distributed service method based on service platform and platform

Abstract

The invention provides a distributed service method and platform based on a service platform, and relates to the technical field of distributed computing power processing. The method comprises the steps of collecting historical computing power periodic supply data of different distributed servers, and performing performance analysis for stable available computing power to form distributed computing power supply data; obtaining a pre-estimated calculation power total amount of the training object set, and performing calculation power supply matching analysis in combination with the distribution calculation power supply data to form distribution calculation power matching data; and acquiring power supply price change data of the power grid, and performing computational power supplementary analysis in combination with the distributed computational power matching data and historical computational power period supply data of different distributed servers to form distributed computational power supplementary data. According to the method, reasonable computing power supply matching is carried out by extracting the distributed computing power based on historical computing power supply characteristics, and efficient and reasonable utilization of total computing power provided by a platform can be realized while a training object is ensured to obtain sufficient computing power.

Description

Technical Field

[0001] The present invention relates to the field of distributed computing power processing technology, and in particular to a distributed service method and platform based on a service platform. Background Art

[0002] In order to cope with the increasing amount of model training data, a single server can no longer meet such needs. With the development of Internet technology, connecting different servers to form a huge computing power provision system to provide sufficient computing power for model training with huge computing power requirements has become the current popular way of providing computing power. Due to the decentralized distribution of servers, this method has also become a distributed computing power provision method.

[0003] Although the distributed computing power provision method solves the huge computing power demand for model training, the computing power that can be provided by different servers varies due to differences in hardware and software performance. At the same time, the time at which computing power can be provided also varies. How to more reasonably allocate the computing power provided by distributed computing power to ensure the computing power demand for model training has become a key issue of concern at present.

[0004] Therefore, it is an urgent problem to be solved to design a distributed service method and platform based on a service platform, to reasonably match the distributed computing power based on the extraction of historical computing power supply characteristics, to ensure that the training objects obtain sufficient computing power while also achieving efficient and reasonable utilization of the total computing power provided by the platform. Summary of the invention

[0005] The purpose of the present invention is to provide a distributed service method based on a service platform. By collecting the historical computing power cycle supply data of different distributed servers within the service platform, the characteristics of the stable available computing power supply are extracted, the computing power supply situation of the distributed servers is determined, and the total estimated computing power demand of all training objects to be provided with computing power by the current platform is combined to perform a reasonable computing power matching analysis to ensure that each training object can obtain sufficient and reasonable computing power supply, and arrange the computing power supply as reasonably as possible to avoid wasting server computing power. At the same time, the periodic volatility of the server computing power supply is fully considered to reserve computing power at low electricity prices to ensure the stability of the computing power supply. This not only realizes the rationality and efficiency of the computing power utilization of the platform's distributed servers, but also can fully meet the computing power needs of the training objects, making the allocation and use of computing power more efficient and reasonable.

[0006] The purpose of the present invention is also to provide a distributed service platform based on a service platform. The platform obtains the historical computing power cycle supply data of the server through a data acquisition unit, and performs a matching analysis of the computing power supply on a computing power analysis unit, thereby realizing the reasonable allocation and utilization of computing power while also considering the computing power storage with low electricity prices, fully realizing the efficiency and rationality of computing power supply, forming an important part of the platform's distributed services, and being an important material basis for realizing the platform's distributed computing power services.

[0007] In a first aspect, the present invention provides a distributed service method based on a service platform, including: collecting historical computing power cycle supply data of different distributed servers, performing performance analysis on stable available computing power, and forming distributed computing power supply data; obtaining the estimated total computing power of the training object set, and performing computing power supply matching analysis in combination with the distributed computing power supply data, and forming distributed computing power matching data; obtaining power grid power supply price change data, and performing computing power supplement analysis in combination with the distributed computing power matching data and the historical computing power cycle supply data of different distributed servers, and forming distributed computing power supplement data.

[0008] In the present invention, the method collects historical computing power cycle supply data of different distributed servers within the service platform, extracts features of stable available computing power supply, determines the computing power supply situation of distributed servers, and performs reasonable computing power matching analysis based on the total estimated computing power demand of all training objects that the current platform is going to provide computing power for, to ensure that each training object can obtain sufficient and reasonable computing power supply, and arranges computing power supply as reasonably as possible to avoid wasting server computing power. At the same time, the periodic volatility of server computing power supply is fully considered to reserve computing power at low electricity prices to ensure the stability of computing power supply. This not only realizes the rationality and efficiency of computing power utilization of the platform's distributed servers, but also can fully meet the computing power needs of the training objects, making the allocation and use of computing power more efficient and reasonable.

[0009] As a possible implementation method, historical computing power cycle supply data of different distributed servers is collected, and performance analysis of stable and available computing power is conducted to form distributed computing power supply data, including: extracting the periodic effective computing power supply change information of the distributed servers based on the historical computing power cycle supply data corresponding to the distributed servers; and conducting computing power availability analysis based on the periodic effective computing power supply change information corresponding to different distributed servers to form distributed computing power supply data.

[0010] In the present invention, in order to reasonably match the computing power of the distributed servers with the training objects, it is first necessary to determine the computing power that the distributed servers can stably provide, and then extract information on the characteristics of the computing power cycle changes by obtaining the historical computing power cycle supply data of the distributed servers, thereby determining the stable computing power that the distributed servers can provide.

[0011] As a possible implementation method, according to the historical computing power cycle supply data corresponding to the distributed server, the periodic effective computing power supply change information of the distributed server is extracted, including: for different distributed servers, the historical periodic computing power change curve corresponding to each period is extracted. , m is the number of different distributed servers, n is the number of different computing cycles of the distributed server numbered m; compare the computing power change curves of different historical cycles under distributed servers , determine the minimum historical computing power supply at each cycle point, and form the periodic effective computing power supply change information corresponding to the distributed server.

[0012] In the present invention, for each training project, the distributed computing power server combines its own computing power to provide as much computing power as possible. Since the computing power of the distributed server has changes in performance, supply time and supply, the computing power it provides is in a constantly changing state during the implementation stage of the entire training project. Therefore, it is first necessary to extract the change information of the computing power provided by the distributed server in the entire training cycle of each training project, which is to lay the data foundation for the subsequent determination of the stable computing power provided by the distributed server. Of course, due to the difference in training projects, the computing power provided by the distributed server each time has differences in size. Therefore, considering the computing power that the distributed server can provide stably, the minimum computing power at each cycle point on all training projects is used as the computing power that can be stably provided to form the periodic effective computing power supply change information of the distributed server. It should be noted that for different training projects, there are certain differences in the training cycle. In order to ensure the comparability and processability of the data, the cycles of different projects can be reasonably reduced to adjust to a unified cycle length. Of course, the corresponding computing power supply on the cycle is also equalized accordingly to achieve this, and the computing power change curve can also be scaled according to the scaling ratio of the cycle to form a unified computing power supply curve.

[0013] As a possible implementation method, according to the periodic effective computing power supply change information corresponding to different distributed servers, computing power availability analysis is performed to form distributed computing power supply data, including: for different distributed servers, according to the corresponding periodic effective computing power supply change information, the historical computing power supply provided at all periodic points is extracted, and the total historical computing power supply over the entire period is determined. ; According to the total historical computing power supply corresponding to the distributed server , determine the corresponding stable available computing power ,in, ; Aggregate the stable and available computing power corresponding to different distributed servers , forming distributed computing power supply data.

[0014] In the present invention, it is considered that the obtained periodic effective computing power supply change information has a certain volatility in the cycle time, and this volatility affects the stability of the computing power provided. Therefore, for the stable available computing power that the distributed server can provide, the total amount of computing power provided in the entire cycle is generally used as the standard. On the one hand, it can avoid fluctuations in the computing power supply when the computing power is continuously provided with stable available computing power, and it can also provide the distributed server with a certain amount of computing power surplus for its own computing power usage needs.

[0015] As a possible implementation method, the estimated total power of the training object set is obtained, and the computing power supply matching analysis is performed in combination with the distributed computing power supply data to form distributed computing power matching data, including: according to the estimated total power of the training object set, the estimated total object computing power corresponding to different training objects is determined. , k represents the number of different training objects; according to the data collection mode of different training objects, the total amount of computing power of the corresponding objects is estimated , determine the minimum computing power split unit corresponding to the training object ; For different training objects, estimate the total amount based on the corresponding object computing power The smallest split unit for sum computing power , and combined with the distributed computing power supply data, conduct computing power supply matching analysis to form distributed computing power matching data.

[0016] In the present invention, after extracting the stable and available computing power data of different distributed servers, a reasonable computing power matching analysis can be performed for the training object. Considering that the provision of computing power is determined by the computing power demand of the training object, it is first necessary to estimate the computing power required for training the training object. This estimate includes both the estimate of the total computing power of all training objects and the estimate of the computing power of each training object. After all, estimating the total computing power can determine how much total computing power needs to be invested by the distributed server, and the computing power estimate for each training object is an important data reference basis for realizing computing power matching. Of course, for different training objects, due to the different types and sizes of the trained data, the minimum demand for computing power is also different, so it is also necessary to determine the minimum split unit of computing power, after all, it is the evaluation basis for the distributed server to provide computing power to the training object.

[0017] As a possible implementation method, for different training objects, the total amount is estimated according to the corresponding object computing power. The smallest split unit for sum computing power , and combined with the distributed computing power supply data, perform computing power supply matching analysis to form distributed computing power matching data, including: according to the minimum computing power split unit corresponding to different training objects , sort the matching order of different training objects in descending order to form the object computing power matching order; according to the object computing power matching order, perform distributed server computing power supply matching analysis on different training objects in turn to form distributed computing power matching data.

[0018] In the present invention, the computing power matching amount corresponding to different distributed servers can be realized according to the estimated computing power demand of the training object and the minimum computing power split unit. It can be understood that for the minimum computing power split unit, the smaller the computing power that can be split, the less computing power will be left when the distributed server provides computing power when supplying computing power for the training object, and the basic remaining amount will not exceed the minimum computing power split unit. When computing power matching is performed, computing power matching is performed once in order of size, so that the reasonable use of the computing power of the distributed server can be ensured.

[0019] As a possible implementation method, according to the object computing power matching order, the distributed server computing power supply matching analysis is performed on different training objects in turn to form distributed computing power matching data, including: according to the object computing power matching order, the minimum computing power splitting unit corresponding to each training object is extracted in turn And the estimated total computing power of the object , perform the following computing power supply matching analysis: the stable available computing power corresponding to all distributed servers , determine the split unit with the minimum computing power Poor computing power , and according to the difference in computing power Sort the distributed servers from small to large to form a distributed server computing power matching order, where: ; Extract in sequence according to the distributed server computing power matching order The number of distributed servers forms the object matching distributed server set corresponding to the training object, and the extracted The number of distributed servers meets the following requirements: , ,express An integer formed by a further tailing method, β represents the minimum allowable computing power difference, and β < 0; excluding the previously matched distributed servers, for the remaining distributed servers, continue to extract the minimum computing power split unit corresponding to the training object according to the object computing power matching order And the estimated total computing power of the object , perform computing power matching analysis until the computing power matching analysis of all training objects is completed; determine the object matching distributed server set corresponding to different training objects to form distributed computing power matching data.

[0020] In the present invention, the computing power of distributed servers is matched for different training objects in the order of arrangement of the splitting units with the smallest computing power. When matching, the stable and available computing power provided by the distributed servers is compared and sorted, and the computing power supply is matched in order from small to large. This ensures that the computing power remaining after the distributed servers supply computing power is minimized, and ensures that the computing power of the distributed servers can be utilized to the maximum extent. For the minimum allowable computing power difference, it is a measure of whether the corresponding distributed server has an impact on the computing power supply after providing the corresponding computing power. Only when there is enough remaining computing power will the volatility of the computing power provided by the server itself not cause an unstable impact on the computing power provided to the training object. The minimum allowable computing power difference can be set according to the actual situation, and can also be determined based on big data analysis. Of course, if there is still enough computing power remaining for the distributed server that has been matched with computing power, it can continue to match to provide computing power for other training objects.

[0021] As a possible implementation method, the power supply price change data of the power grid is obtained, and the distributed computing power matching data and the historical computing power cycle supply data of different distributed servers are combined to conduct computing power supplement analysis to form distributed computing power supplement data, including: according to the distributed computing power matching data, combined with the total historical computing power supply corresponding to different distributed servers The following computing power missing amount analysis is performed based on the periodic effective computing power supply change information to form the computing power missing amount analysis result: for each distributed server in the object matching distributed server set corresponding to different training objects, according to the periodic effective computing power supply change information corresponding to the distributed server, determine the historical computing power supply at the periodic point that is less than the minimum computing power split unit The period of time is marked as the period of insufficient computing power; according to the periodic effective computing power supply change information corresponding to the distributed server, the historical computing power supply in the period of insufficient computing power and the minimum split unit of continuous computing power supply in the same period are determined The computing power difference of the computing power of the distributed server is used to form the missing computing power; according to the periodic effective computing power supply change information corresponding to the distributed server, the historical computing power supply in other periods outside the period of insufficient computing power and the minimum split unit of the continuous computing power supply in the same period are determined. The computing power difference of the computing power is formed to form the self-charging computing power; the following analysis is performed based on the self-charging computing power and the missing computing power: if the self-charging computing power is not less than the missing computing power, the corresponding distributed server is determined as the first-time outsourced computing power server; if the self-charging computing power is less than the missing computing power, the corresponding distributed server is determined as the continuous outsourced computing power server; based on the results of the computing power missing analysis and combined with the power supply price change data of the power grid, a computing power supplement analysis is performed to form the computing power supplement analysis result data.

[0022] In the present invention, although the minimum allowable computing power difference is used to avoid the impact of the volatility of the server computing power supply on the computing power supply when matching the computing power of a single training object, there will still be insufficient computing power in actual situations, especially when the server provides computing power in the early stage. Before that, a certain amount of computing power storage is performed, which causes the computing power supply to fluctuate, resulting in insufficient computing power and no computing power to be supplemented. Therefore, in this case, before the computing power supply is performed, the storage of computing power should be considered in advance. Since the generation of computing power is closely related to the consumption of electric energy resources, in order to control the cost of computing power production, it is necessary to obtain electric energy based on the characteristics of power supply price changes of the power grid for computing power production. Before production, the actual situation of the computing power of the distributed server should be accurately confirmed. One is that there is excess computing power in the subsequent computing power fluctuations, which cannot make up for the lack of computing power caused by the previous computing power fluctuations. This lack is continuous. The other is that there is excess computing power in the subsequent computing power fluctuations. It can make up for the lack of computing power caused by the previous computing power fluctuations. This lack will only occur in the early stage. When there is excess computing power in the later computing power fluctuations, it can be self-sufficient.

[0023] As a possible implementation method, based on the analysis results of the amount of computing power missing, combined with the power grid price change data, a computing power replenishment analysis is conducted to form computing power replenishment analysis result data, including: if the distributed server is a server that borrows computing power for the first time, then according to the computing power shortage cycle period of the server that borrows computing power for the first time and the corresponding historical computing power supply in the period, and combined with the power grid price change data, the earliest charging time point for the first borrowing computing power and the charging period during the low-price electricity period are determined; if the distributed server is a server that borrows computing power continuously, then: according to the continuous external The computing power server's computing power is insufficient during the period and the corresponding historical computing power supply during the period, and combined with the power grid price change data, the earliest charging time for the first borrowed computing power and the charging time during the period of low electricity prices are determined; the computing power insufficient period that can be self-charged is determined based on the difference between the self-charged computing power and the missing computing power, and for the remaining computing power insufficient period that cannot be automatically charged, the earliest charging time for continuous borrowing of computing power and the charging time during the period of low electricity prices are determined based on the corresponding historical computing power supply during the period and the power grid price change data.

[0024] In the present invention, according to the determined distributed servers with different types of computing power shortage, reasonable computing power production and storage are performed in combination with low-cost power supply from the power grid to provide sufficient computing power supplement for subsequent computing power supply fluctuations, thereby ensuring that the distributed servers can continuously and effectively provide stable and sufficiently effective computing power.

[0025] In the second aspect, the present invention provides a distributed service platform based on a service platform, including a data collection unit for collecting historical computing power cycle supply data of different distributed servers; a computing power analysis unit for performing available computing power performance analysis on the historical computing power cycle supply data obtained by the data collection unit to form distributed computing power supply data, performing computing power estimation on a training object set and performing computing power matching analysis in combination with the distributed computing power supply data to form distributed computing power matching data, and performing computing power supplement analysis based on power grid power supply price change data to form distributed computing power supplement data.

[0026] The platform obtains the historical computing power cycle supply data of the server through the data acquisition unit, and performs matching analysis of the computing power supply on the computing power analysis unit. While achieving reasonable allocation and utilization of computing power, it also considers computing power storage with low electricity prices, fully realizing the efficiency and rationality of computing power supply, forming an important part of the platform's distributed services, and is an important material basis for realizing the platform's distributed computing power services.

[0027] The distributed service method and platform based on the service platform provided by the present invention have the following beneficial effects: The method collects historical computing power cycle supply data of different distributed servers within the service platform, extracts the characteristics of stable available computing power supply, determines the computing power supply of distributed servers, and performs reasonable computing power matching analysis based on the total estimated computing power demand of all training objects that the current platform needs to provide computing power. It ensures that each training object can obtain sufficient and reasonable computing power supply, and arranges computing power supply as reasonably as possible to avoid wasting server computing power. At the same time, it fully considers the periodic volatility of server computing power supply to reserve computing power at low electricity prices to ensure the stability of computing power supply. This not only realizes the rationality and efficiency of computing power utilization of the platform's distributed servers, but also can fully meet the computing power needs of the training objects, making the allocation and use of computing power more efficient and reasonable. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments of the present invention are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0029] Figure 1 A step diagram of a distributed service method based on a service platform provided in an embodiment of the present invention; Figure 2 A schematic diagram of the structure of a distributed service platform based on a service platform provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0030] The technical solutions in the embodiments of the present invention will be described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0031] In order to cope with the increasing amount of model training data, a single server can no longer meet such needs. With the development of Internet technology, connecting different servers to form a huge computing power provision system to provide sufficient computing power for model training with huge computing power requirements has become the current popular way of providing computing power. Due to the decentralized distribution of servers, this method has also become a distributed computing power provision method.

[0032] Although the distributed computing power provision method solves the huge computing power demand for model training, the computing power that can be provided by different servers varies due to differences in hardware and software performance. At the same time, the time at which computing power can be provided also varies. How to more reasonably allocate the computing power provided by distributed computing power to ensure the computing power demand for model training has become a key issue of concern at present.

[0033] refer to Figure 1-Figure 2 The embodiment of the present invention provides a distributed service method based on a service platform. The training cluster collects historical computing power cycle supply data of different distributed servers within the service platform, extracts features of stable available computing power supply, determines the computing power supply of distributed servers, and performs reasonable computing power matching analysis based on the total estimated computing power demand of all training objects to be provided with computing power by the current platform, to ensure that each training object can obtain sufficient and reasonable computing power supply, and arranges computing power supply as reasonably as possible to avoid wasting server computing power. At the same time, the periodic volatility of server computing power supply is fully considered to reserve computing power at low electricity prices to ensure the stability of computing power supply. In this way, the rationality and efficiency of computing power utilization of the platform's distributed servers are achieved, and the computing power demand of the training objects can be fully met, making the allocation and use of computing power more efficient and reasonable.

[0034] The distributed service method based on the service platform specifically includes the following steps: S1: Collect historical computing power cycle supply data of different distributed servers, conduct performance analysis on stable available computing power, and form distributed computing power supply data.

[0035] Collect historical computing power cycle supply data of different distributed servers, conduct performance analysis on stable and available computing power, and form distributed computing power supply data, including: extracting the periodic effective computing power supply change information of distributed servers based on the historical computing power cycle supply data corresponding to the distributed servers; and conducting computing power availability analysis based on the periodic effective computing power supply change information corresponding to different distributed servers to form distributed computing power supply data.

[0036] In order to reasonably match the computing power of the distributed servers to the training objects, it is first necessary to determine the computing power that the distributed servers can stably provide. By obtaining the historical computing power cycle supply data of the distributed servers, information on the characteristics of the computing power cycle changes can be extracted to determine the stable computing power that the distributed servers can provide.

[0037] According to the historical computing power cycle supply data corresponding to the distributed server, the periodic effective computing power supply change information of the distributed server is extracted, including: for different distributed servers, the corresponding historical cycle computing power change curve for each cycle is extracted , m is the number of different distributed servers, n is the number of different computing cycles of the distributed server numbered m; compare the computing power change curves of different historical cycles under distributed servers , determine the minimum historical computing power supply at each cycle point, and form the periodic effective computing power supply change information corresponding to the distributed server.

[0038] For each training project, the distributed computing power server combines its own computing power to provide as much computing power as possible. Since the computing power of the distributed server varies in performance, supply time and supply, the computing power it provides is in a state of constant change during the implementation of the entire training project. Therefore, it is necessary to first extract the change information of the computing power provided by the distributed server during the entire training cycle of each training project. This is to lay the data foundation for the subsequent determination of the stable computing power provided by the distributed server. Of course, due to the different training projects, the computing power provided by the distributed server each time is different in size. Therefore, considering the computing power that the distributed server can provide stably, the minimum computing power at each cycle point on all training projects is used as the computing power that can be stably provided to form the periodic effective computing power supply change information of the distributed server. It should be noted that for different training projects, there are certain differences in the training cycle. In order to ensure the comparability and processability of the data, the cycles of different projects can be reasonably reduced to adjust to a unified cycle length. Of course, the corresponding computing power supply in the cycle is also equalized accordingly to achieve this. The computing power change curve can also be scaled according to the scaling ratio of the cycle to form a unified computing power supply curve.

[0039] According to the periodic effective computing power supply change information corresponding to different distributed servers, computing power availability analysis is performed to form distributed computing power supply data, including: for different distributed servers, according to the corresponding periodic effective computing power supply change information, the historical computing power supply provided at all periodic points is extracted, and the total historical computing power supply over the entire period is determined. ; According to the total historical computing power supply corresponding to the distributed server , determine the corresponding stable available computing power ,in, ; Aggregate the stable and available computing power corresponding to different distributed servers , forming distributed computing power supply data.

[0040] Considering that the information on changes in effective computing power supply obtained during the period has certain volatility in the periodic time, and this volatility affects the stability of computing power provision, the total amount of computing power provided during the entire period is generally used as the standard for the stable available computing power that can be provided by distributed servers. On the one hand, it can avoid fluctuations in computing power supply when computing power is continuously provided with stable available computing power, and it can also provide a certain amount of computing power surplus for distributed servers for their own computing power usage needs.

[0041] S2: Obtain the estimated total computing power of the training object set, and perform computing power supply matching analysis in combination with the distributed computing power supply data to form distributed computing power matching data.

[0042] Obtain the estimated total power of the training object set, and perform power supply matching analysis in combination with the distributed power supply data to form distributed power matching data, including: Determine the estimated total object computing power corresponding to different training objects based on the estimated total power of the training object set , k represents the number of different training objects; according to the data collection mode of different training objects, the total amount of computing power of the corresponding objects is estimated , determine the minimum computing power split unit corresponding to the training object ; For different training objects, estimate the total amount based on the corresponding object computing power The smallest split unit for sum computing power , and combined with the distributed computing power supply data, conduct computing power supply matching analysis to form distributed computing power matching data.

[0043] After extracting the stable and available computing power data of different distributed servers, a reasonable computing power matching analysis can be performed for the training object. Considering that the provision of computing power is determined by the computing power demand of the training object, it is necessary to first estimate the computing power required for training the training object. This estimate includes both the total computing power estimate of all training objects and the computing power estimate of each training object. After all, estimating the total computing power can determine how much computing power is required for distributed servers, and estimating the computing power of each training object is an important data reference basis for achieving computing power matching. Of course, for different training objects, due to the different types and sizes of the training data, the minimum demand for computing power is also different, so it is also necessary to determine the minimum split unit of computing power, after all, it is the evaluation basis for the distributed server to provide computing power to the training object.

[0044] For different training objects, estimate the total amount based on the corresponding object computing power The smallest split unit for sum computing power , and combined with the distributed computing power supply data, perform computing power supply matching analysis to form distributed computing power matching data, including: according to the minimum computing power split unit corresponding to different training objects , sort the matching order of different training objects in descending order to form the object computing power matching order; according to the object computing power matching order, perform distributed server computing power supply matching analysis on different training objects in turn to form distributed computing power matching data.

[0045] According to the estimated computing power demand of the training object and the minimum computing power split unit, the computing power matching amount corresponding to different distributed servers can be achieved. It can be understood that for the minimum computing power split unit, the smaller the computing power that can be split, the less computing power will be left when the distributed server provides computing power for the training object, and the basic remaining amount will not exceed the minimum computing power split unit. When matching computing power, the computing power is matched once in order of size, which can ensure the reasonable use of the computing power of the distributed server.

[0046] According to the object computing power matching order, the distributed server computing power supply matching analysis is performed on different training objects in turn to form distributed computing power matching data, including: according to the object computing power matching order, the minimum computing power split unit corresponding to each training object is extracted in turn And the estimated total computing power of the object , perform the following computing power supply matching analysis: the stable available computing power corresponding to all distributed servers , determine the split unit with the minimum computing power Poor computing power , and according to the difference in computing power Sort the distributed servers from small to large to form a distributed server computing power matching order, where: ; Extract in sequence according to the distributed server computing power matching order The number of distributed servers forms the object matching distributed server set corresponding to the training object, and the extracted The number of distributed servers meets the following requirements: , , express An integer formed by a further tailing method, β represents the minimum allowable computing power difference, and β < 0; excluding the previously matched distributed servers, for the remaining distributed servers, continue to extract the minimum computing power split unit corresponding to the training object according to the object computing power matching order And the estimated total computing power of the object , perform computing power matching analysis until the computing power matching analysis of all training objects is completed; determine the object matching distributed server set corresponding to different training objects to form distributed computing power matching data.

[0047] The computing power of distributed servers is matched for different training objects in the order of the arrangement of the smallest splitting units based on computing power. When matching, the stable available computing power provided by the distributed servers is compared and sorted, and the computing power supply is matched in order from small to large. This ensures that the remaining computing power of the distributed servers after the computing power supply is minimized, and ensures that the computing power of the distributed servers can be utilized to the maximum extent. The minimum allowable computing power difference is a measure of whether the computing power volatility of the corresponding distributed servers affects the computing power supply after providing the corresponding computing power. Only when there is enough remaining computing power will the volatility of the computing power provided by the server itself not cause an unstable impact on the computing power provided to the training objects. The minimum allowable computing power difference can be set according to the actual situation, or it can be determined based on big data analysis. Of course, if there is enough computing power remaining for the distributed servers that have already been matched, it can continue to match to provide computing power for other training objects.

[0048] S3: Obtain the power supply price change data of the power grid, and combine it with the distributed computing power matching data and the historical computing power cycle supply data of different distributed servers to conduct computing power replenishment analysis to form distributed computing power replenishment data.

[0049] Obtain the power supply price change data of the power grid, and combine the distributed computing power matching data and the historical computing power cycle supply data of different distributed servers to conduct computing power supplement analysis to form distributed computing power supplement data, including: according to the distributed computing power matching data, combined with the total historical computing power supply corresponding to different distributed servers The following computing power missing amount analysis is performed based on the periodic effective computing power supply change information to form the computing power missing amount analysis result: for each distributed server in the object matching distributed server set corresponding to different training objects, according to the periodic effective computing power supply change information corresponding to the distributed server, determine the historical computing power supply at the periodic point that is less than the minimum computing power split unit The period of time is marked as the period of insufficient computing power; according to the periodic effective computing power supply change information corresponding to the distributed server, the historical computing power supply in the period of insufficient computing power and the minimum split unit of continuous computing power supply in the same period are determined The computing power difference of the computing power of the distributed server is used to form the missing computing power; according to the periodic effective computing power supply change information corresponding to the distributed server, the historical computing power supply in other periods outside the period of insufficient computing power and the minimum split unit of the continuous computing power supply in the same period are determined. The computing power difference of the computing power is formed to form the self-charging computing power; the following analysis is performed based on the self-charging computing power and the missing computing power: if the self-charging computing power is not less than the missing computing power, the corresponding distributed server is determined as the first-time outsourced computing power server; if the self-charging computing power is less than the missing computing power, the corresponding distributed server is determined as the continuous outsourced computing power server; based on the results of the computing power missing analysis and combined with the power supply price change data of the power grid, a computing power supplement analysis is performed to form the computing power supplement analysis result data.

[0050] Although the minimum allowable computing power difference has been used to avoid the impact of the volatility of server computing power supply on computing power supply when matching the computing power of a single training object, there will still be insufficient computing power in actual situations, especially when the server is providing computing power in the early stage. Previously, a certain amount of computing power was stored, which caused the computing power supply to fluctuate and there was no computing power to supplement when the computing power was insufficient. Therefore, in this case, before providing computing power, it is necessary to consider storing computing power in advance. Since the generation of computing power is closely related to the consumption of electric energy resources, in order to control the cost of computing power production, it is necessary to obtain electricity based on the characteristics of power supply price changes of the power grid for computing power production. Before production, the actual situation of the computing power of the distributed server should be accurately confirmed. One is that the excess computing power in the later computing power fluctuations cannot make up for the lack of computing power caused by the previous computing power fluctuations. This lack is continuous. The other is that the excess computing power in the later computing power fluctuations can make up for the lack of computing power caused by the previous computing power fluctuations. This lack only occurs in the early stage. When there is excess computing power in the later computing power fluctuations, it can be self-sufficient.

[0051] According to the results of the computing power shortage analysis and in combination with the grid power supply price change data, a computing power replenishment analysis is performed to form computing power replenishment analysis result data, including: if the distributed server is a server that borrows computing power for the first time, then according to the computing power shortage cycle period of the server that borrows computing power for the first time and the corresponding historical computing power supply in the period, and in combination with the grid power supply price change data, the earliest charging time point for the first borrowing of computing power and the charging time period in the period of low electricity price are determined; if the distributed server is a server that borrows computing power continuously, then according to the computing power shortage cycle period of the server that borrows computing power continuously and the corresponding historical computing power supply in the period, and in combination with the grid power supply price change data, the earliest charging time point for the first borrowing of computing power and the charging time period in the period of low electricity price are determined; according to the difference between the self-charging computing power and the missing computing power, the computing power shortage cycle period that can be self-charged is determined, and for the remaining computing power shortage cycle periods that cannot be automatically charged, according to the corresponding historical computing power supply in the period, combined with the grid power supply price change data, the earliest charging time point for the continuous borrowing of computing power and the charging time period in the period of low electricity price are determined.

[0052] According to the different types of distributed servers that are lacking computing power, reasonable computing power production and storage are carried out in combination with low-cost power supply from the power grid to provide sufficient computing power to supplement subsequent computing power supply fluctuations, ensuring that distributed servers can continue to effectively provide stable and sufficiently effective computing power.

[0053] The present application also provides a distributed service platform based on a service platform, which includes a data collection unit for collecting historical computing power cycle supply data of different distributed servers; a computing power analysis unit for performing available computing power performance analysis on the historical computing power cycle supply data obtained by the data collection unit to form distributed computing power supply data, performing computing power estimation on the training object set and performing computing power matching analysis in combination with the distributed computing power supply data to form distributed computing power matching data, and performing computing power supplement analysis based on the power grid power supply price change data to form distributed computing power supplement data.

[0054] The platform obtains the historical computing power cycle supply data of the server through the data acquisition unit, and performs matching analysis of the computing power supply on the computing power analysis unit. While achieving reasonable allocation and utilization of computing power, it also considers computing power storage with low electricity prices, fully realizing the efficiency and rationality of computing power supply, forming an important part of the platform's distributed services, and is an important material basis for realizing the platform's distributed computing power services.

[0055] In summary, the distributed service method and platform based on the service platform provided by the embodiments of the present invention have the following beneficial effects: This method collects historical computing power cycle supply data of different distributed servers within the service platform, extracts the characteristics of stable available computing power supply, determines the computing power supply of distributed servers, and performs reasonable computing power matching analysis based on the total estimated computing power demand of all training objects that the current platform needs to provide computing power, ensuring that each training object can obtain sufficient and reasonable computing power supply, and arranges computing power supply as reasonably as possible to avoid wasting server computing power. At the same time, it fully considers the periodic volatility of server computing power supply to reserve computing power at low electricity prices to ensure the stability of computing power supply, so as to achieve the rationality and efficiency of computing power utilization of platform distributed servers, and can fully meet the computing power needs of training objects, making the allocation and use of computing power more efficient and reasonable. It is an important part of the platform's distributed services and an important material basis for realizing the platform's distributed computing power services.

[0056] In the embodiments of the present application, "indication" may include direct indication and indirect indication, and may also include explicit indication and implicit indication. The information indicated by a certain information is called information to be indicated. In the specific implementation process, there are many ways to indicate the information to be indicated, such as but not limited to, the information to be indicated can be directly indicated, such as the information to be indicated itself or the index of the information to be indicated. The information to be indicated can also be indirectly indicated by indicating other information, wherein there is an association between the other information and the information to be indicated. It is also possible to indicate only a part of the information to be indicated, while the other parts of the information to be indicated are known or agreed in advance. For example, the indication of specific information can also be achieved by means of the arrangement order of each piece of information agreed in advance (for example, specified by the protocol), thereby reducing the indication overhead to a certain extent. At the same time, the common parts of each piece of information can also be identified and indicated uniformly to reduce the indication overhead caused by indicating the same information separately.

[0057] In addition, the specific indication method may also be various existing indication methods, such as but not limited to the above-mentioned indication methods and various combinations thereof. The specific details of the various indication methods can refer to the prior art and will not be repeated herein. As can be seen from the above, for example, when it is necessary to indicate multiple information of the same type, different indication methods may be used for different information. In the specific implementation process, the desired indication method can be selected according to specific needs. The embodiment of the present application does not limit the selected indication method. In this way, the indication method involved in the embodiment of the present application should be understood to cover various methods that can enable the party to be indicated to obtain the information to be indicated.

[0058] It should be understood that the information to be indicated can be sent as a whole, or divided into multiple sub-information and sent separately, and the sending period and / or sending time of these sub-information can be the same or different. The specific sending method is not limited in the embodiment of the present application. Among them, the sending period and / or sending time of these sub-information can be pre-defined, for example, pre-defined according to a protocol, or can be configured by the sending end device by sending configuration information to the receiving end device.

[0059] "Pre-definition" or "pre-configuration" can be implemented by pre-saving corresponding codes, tables or other methods that can be used to indicate relevant information in the device, and the embodiments of the present application do not limit the specific implementation method. Among them, "saving" can mean saving in one or more memories. The one or more memories can be set separately or integrated in an encoder or decoder, a processor, or a communication device. The one or more memories can also be partially set separately and partially integrated in a decoder, a processor, or a communication device. The type of memory can be any form of storage medium, which is not limited by the embodiments of the present application.

[0060] The "protocol" involved in the embodiments of the present application may refer to a protocol family in the communication field, a standard protocol with a similar protocol family frame structure, or a related protocol used in future communication systems, and the embodiments of the present application do not make specific limitations on this.

[0061] In the embodiments of the present application, descriptions such as "when...", "in the case of...", "if" and "if" all mean that the device will make corresponding processing under certain objective circumstances. It does not limit the time, nor does it require the device to have a judgment action when implementing it, nor does it mean that there are other limitations.

[0062] In the description of the embodiments of the present application, unless otherwise specified, " / " indicates that the objects associated before and after are in an "or" relationship. For example, A / B can represent A or B; "and / or" in the embodiments of the present application is only a description of the association relationship of the associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. These three situations, where A and B can be singular or plural. In addition, in the description of the embodiments of the present application, unless otherwise specified, "multiple" refers to two or more than two. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, c can be single or multiple. In addition, in order to facilitate the clear description of the technical solution of the embodiments of the present application, in the embodiments of the present application, the words "first" and "second" are used to distinguish the same or similar items with basically the same functions and effects. Those skilled in the art will appreciate that words such as "first" and "second" do not limit the quantity and execution order, and words such as "first" and "second" do not necessarily limit the difference. At the same time, in the embodiments of the present application, words such as "exemplary" or "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "exemplary" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a concrete way for easy understanding.

[0063] It should be understood that the processor in the embodiments of the present application may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0064] It should also be understood that the memory in the embodiments of the present application may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic random access memory (DRAM), synchronous DRAM (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0065] The above embodiments can be implemented in whole or in part by software, hardware (such as circuits), firmware or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center by wired (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a tape), an optical medium (for example, a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0066] It should be understood that the term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. A and B can be singular or plural. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship, but it may also indicate an "and / or" relationship. Please refer to the context for specific understanding.

[0067] In this application, "at least one" means one or more, and "plurality" means two or more. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b, or c can mean: a, b, c, ab, ac, bc, or abc, where a, b, and c can be single or multiple.

[0068] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0069] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0070] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0071] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0072] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0073] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0074] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, and other media that can store program codes.

[0075] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A distributed service method based on a service platform, characterized in that: is configured as: Collect historical computing power cycle supply data of different distributed servers, conduct performance analysis on stable available computing power, and form distributed computing power supply data; Obtaining the estimated total computing power of the training object set, and performing computing power supply matching analysis in combination with the distributed computing power supply data to form distributed computing power matching data; Obtain the power supply price change data of the power grid, and combine the distributed computing power matching data and the historical computing power cycle supply data of different distributed servers to perform computing power supplement analysis to form distributed computing power supplement data.

2. The distributed service method based on the service platform according to claim 1, characterized in that: The historical computing power cycle supply data of different distributed servers is collected, and performance analysis is performed on stable and available computing power to form distributed computing power supply data, including: Extracting the periodic effective computing power supply change information of the distributed server according to the historical computing power period supply data corresponding to the distributed server; Based on the periodic effective computing power supply change information corresponding to different distributed servers, computing power availability analysis is performed to form the distributed computing power supply data.

3. The distributed service method based on the service platform according to claim 2, characterized in that: The extracting, according to the historical computing power cycle supply data corresponding to the distributed server, the periodic effective computing power supply change information of the distributed server includes: For different distributed servers, extract the historical cycle computing power change curve of each corresponding cycle , m is the number of different distributed servers, and n represents the number of different computing cycles of the distributed server numbered m; Compare the different historical period computing power change curves under the distributed server , determine the minimum historical computing power supply at each periodic point, and form the periodic effective computing power supply change information corresponding to the distributed server.

4. The distributed service method based on the service platform according to claim 3, characterized in that: The performing of computing power availability analysis according to the periodic effective computing power supply change information corresponding to different distributed servers to form the distributed computing power supply data includes: For different distributed servers, according to the corresponding periodic effective computing power supply change information, extract the historical computing power supply provided at all periodic points, and determine the total historical computing power supply for the entire period ; According to the total amount of historical computing power supply corresponding to the distributed server , determine the corresponding stable available computing power ,in, ; Aggregate the stable available computing power corresponding to different distributed servers , forming the distributed computing power supply data.

5. The distributed service method based on the service platform according to claim 4, characterized in that: The obtaining of the estimated total computing power of the training object set and performing computing power supply matching analysis in combination with the distributed computing power supply data to form distributed computing power matching data includes: According to the estimated total computing power of the training object set, the estimated total computing power of the object corresponding to different training objects is determined. , k represents the number of different training objects; According to the data collection mode of different training objects, combined with the corresponding estimated total computing power of the object , determine the minimum computing power split unit corresponding to the training object ; For different training objects, the total amount of computing power is estimated according to the corresponding object and the minimum splitting unit of computing power , and combined with the distributed computing power supply data, perform computing power supply matching analysis to form the distributed computing power matching data.

6. The distributed service method based on the service platform according to claim 5, characterized in that: For different training objects, the total amount of computing power is estimated according to the corresponding object and the minimum splitting unit of computing power , and combined with the distributed computing power supply data, perform computing power supply matching analysis to form the distributed computing power matching data, including: The minimum splitting unit of the computing power corresponding to different training objects , sorting the matching order of the different training objects in descending order to form an object computing power matching order; According to the object computing power matching order, the computing power supply matching analysis of the distributed server is performed on different training objects in turn to form the distributed computing power matching data.

7. The distributed service method based on the service platform according to claim 6, characterized in that: The step of performing the computing power supply matching analysis of the distributed server on different training objects in sequence according to the object computing power matching order to form the distributed computing power matching data includes: According to the object computing power matching order, the minimum computing power splitting unit corresponding to each training object is extracted in turn. and the estimated total computing power of the object , perform the following computing power supply matching analysis: The stable available computing power corresponding to all distributed servers , determine the split unit with the minimum computing power Poor computing power , and according to the computing power difference The distributed servers are sorted in order from small to large to form a distributed server computing power matching order, where: ; According to the distributed server computing power matching order, extract The number of distributed servers forms an object matching distributed server set corresponding to the training object, and the extracted The number of distributed servers meets the following requirements: , , express An integer formed by rounding up, β represents the minimum allowable computing power difference, and β < 0; Excluding the distributed servers that have been matched before, for the remaining distributed servers, continue to extract the minimum computing power split unit corresponding to the training object according to the object computing power matching order and the estimated total computing power of the object , perform computing power matching analysis until the computing power matching analysis of all the training objects is completed; Determine the object matching distributed server set corresponding to different training objects to form the distributed computing power matching data.

8. The distributed service method based on the service platform according to claim 7, characterized in that: The obtaining of the power supply price change data of the power grid, and combining the distributed computing power matching data and the historical computing power cycle supply data of different distributed servers, performing computing power supplement analysis to form distributed computing power supplement data, includes: According to the distributed computing power matching data, combined with the total historical computing power supply corresponding to different distributed servers The following computing power shortage analysis is performed based on the effective computing power supply change information of the period to form the computing power shortage analysis result: For each of the distributed servers in the object matching distributed server set corresponding to different training objects, according to the periodic effective computing power supply change information corresponding to the distributed server, determine that the historical computing power supply at the periodic point is less than the minimum computing power split unit The period of time is marked as the period of insufficient computing power; According to the periodic effective computing power supply change information corresponding to the distributed server, determine the historical computing power supply during the computing power shortage period and the minimum split unit of the computing power that is continuously supplied during the same period The computing power of the two parties is poor, resulting in a lack of computing power; According to the periodic effective computing power supply change information corresponding to the distributed server, determine the historical computing power supply in other periods outside the period of insufficient computing power and the minimum split unit of the computing power that is continuously supplied in the same period The computing power of the two parties is poor, forming a self-charging computing power; The following analysis is performed based on the self-charging computing power and the missing computing power: If the self-charged computing power is not less than the missing computing power, the corresponding distributed server is determined as a server for first-time borrowing computing power; If the self-charged computing power is less than the missing computing power, the corresponding distributed server is determined as a server for continuous borrowing computing power; Based on the analysis results of the computing power shortage and in combination with the power supply price change data of the power grid, a computing power supplement analysis is performed to form computing power supplement analysis result data.

9. The distributed service method based on the service platform according to claim 8, characterized in that: The computing power replenishment analysis is performed based on the computing power shortage analysis result and combined with the power grid power supply price change data to form computing power replenishment analysis result data, including: If the distributed server is the server that is borrowing computing power for the first time, then according to the computing power shortage period of the server that is borrowing computing power for the first time and the historical computing power supply corresponding to the period, and in combination with the power supply price change data of the power grid, determine the earliest charging time point for the first borrowing computing power and the charging time period during the period of low electricity price; If the distributed server is the continuously borrowed computing power server, then: According to the insufficient computing power cycle period of the continuously borrowed computing power server and the corresponding historical computing power supply in the period, and in combination with the power supply price change data of the power grid, determine the earliest charging time point for the first borrowed computing power and the charging period during the low electricity price period; The period of insufficient computing power during which self-charging can be achieved is determined based on the difference between the self-charging computing power and the missing computing power. For the remaining period of insufficient computing power during which self-charging cannot be achieved, the earliest charging time point for continuously borrowing computing power and the charging period during the period of low electricity prices are determined based on the historical computing power supply corresponding to the period and combined with the power supply price change data of the power grid.

10. A distributed service platform based on a service platform, characterized in that: The distributed service method based on the service platform according to any one of claims 1 to 9 comprises: A data collection unit, used to collect historical computing cycle supply data of different distributed servers; The computing power analysis unit is used to perform available computing power performance analysis on the historical computing power cycle supply data obtained by the data acquisition unit to form distributed computing power supply data, estimate the computing power of the training object set and perform computing power matching analysis in combination with the distributed computing power supply data to form distributed computing power matching data, and perform computing power supplement analysis based on the power grid power supply price change data to form distributed computing power supplement data.

Images