Task-driven large mobile device mobile data center resource provisioning prediction method

Abstract

The application discloses a task-driven large mobile equipment mobile data center resource configuration prediction method. The method comprises the following steps: ①data storage and processing method setting, determining the physical structure of the mobile data center, the data storage method and parameters and the data processing method and parameters according to the physical structure and actual use of the mobile equipment and other factors; ②data storage resource configuration prediction, predicting the configuration of the data storage resource according to the data storage method and parameters; ③data processing resource configuration prediction, predicting the configuration of the data processing resource according to the data processing method and parameters; and ④iterative updating, collecting the physical characteristics of the mobile equipment and the mobile data center, the data storage and processing task related parameters, the resource configuration prediction results and the like to form a resource configuration library, and adjusting and updating according to the changes of the equipment and the task. The application can significantly improve the accuracy of the resource configuration prediction result of the large mobile equipment mobile data center.

Description

Technical Field

[0001] This invention relates to a method for predicting the data storage and processing resources required to build a mobile data center on a large mobile device, belonging to the field of mobile device data processing. Background Technology

[0002] The increasing intelligence of large mobile devices such as aircraft and ships is underpinned by advanced technologies including high-precision sensors and high-performance data processing equipment. A mobile data center, formed by various sensors, actuators, connectors, and servers on large mobile devices connected via a dedicated data bus, performs data acquisition, transmission, storage, and processing functions. It acts as the "brain" of the mobile device, comprehensively processing information captured by its "eyes" (sensors, radar, satellites, etc.) to achieve precise and real-time control. The configuration of sensors on a mobile device is typically determined by its intended use and operating environment; the configuration of the data transmission bus is usually determined by the device's structural characteristics; and the configuration of data storage and processing resources depends primarily on the data storage model and the device's task processing requirements. The level of data storage and processing resource configuration significantly impacts the operation and control of the mobile device. Insufficient data storage and processing resources can lead to data loss and delays in intelligent control, while excessive resources can result in increased energy consumption and costs, increased space occupation, increased equipment failure rates, and increased maintenance difficulty. Accurately predicting the demand for data storage and processing resources is a prerequisite for resource allocation. Predicting the resource allocation of large mobile devices requires taking into account factors such as the data acquisition needs, data storage strategies, and data processing tasks of mobile data centers.

[0003] Currently, both military and civilian applications are transforming the organization and management of mobile data centers. Distributed integrated modular control systems have replaced traditional centralized control systems, enabling the classification and modular processing of tasks. Researchers have proposed distributed data storage and processing models, such as the Distributed Integrated Modular Avionics (DIMA) system for aircraft, which utilizes distributed principles to manage and schedule resources in onboard mobile data centers. DIMA has been designated by governments worldwide as the next-generation avionics control system for aircraft. However, current research on DIMA primarily focuses on network architecture design, hardware resource deployment, and security verification, with limited research addressing data storage and processing. While technologies in traditional data centers and edge computing are mature, they are difficult to directly apply to mobile data centers due to the following characteristics: First, mobile data centers have a small network scale, typically consisting of only a few servers and dozens to hundreds of edge nodes such as sensors, with servers directly connected via high-speed data buses, eliminating the need for relaying via switches or gateways. Second, they have fewer task types, primarily performing data analysis tasks for monitoring, evaluating, and making control decisions about their own status, typically including trigger-based tasks for monitoring and early warning and analytical tasks for decision analysis. Third, they have high reliability requirements, as the working environment of mobile devices makes it difficult to perform off-site data backups, and data loss is irreversible. Fourth, they have high real-time requirements, as the intelligent control of mobile devices requires tasks to be completed within a limited time; in some cases, untimely task processing can even lead to the catastrophic damage of the mobile device. Traditional experience-based resource allocation prediction methods do not consider data storage strategies and task types, but only predict the total amount of data and peak computing needs, which are suitable for centralized server configurations. However, mobile data centers need to predict resource allocation for each server of mobile devices, and must comprehensively consider the data storage and processing tasks of each server, such as the collection scale and replication strategy of data storage tasks, the task type of data processing tasks, task allocation and task latency, etc.

[0004] From a data perspective, the intelligent control of large mobile devices essentially involves highly reliable real-time processing of data collected by various sensors. This necessitates sufficient data storage and processing resources for the mobile data center. Predicting mobile data center resource allocation serves two purposes: firstly, it ensures accurate resource allocation, guaranteeing reliable data storage and efficient processing; secondly, it helps users become familiar with the data acquisition scale, data storage strategies, task types, task allocation strategies, and task processing response speeds of the mobile data center, facilitating the maintenance, management, and optimization of the mobile data center's software resources. Summary of the Invention

[0005] This invention provides a task-driven method for predicting resource allocation in mobile data centers for large mobile devices. The method aims to predict the allocation of data storage and processing resources based on the data storage and processing tasks of the mobile data center. It comprehensively considers the data acquisition scale, generation scale, replication strategy, and storage resource vacancy rate of data storage tasks to predict the allocation of data storage resources. It also comprehensively considers the task type, number of tasks, data requirements, allocation strategy, task latency, and processing resource vacancy rate of data processing tasks to predict the allocation of data processing resources, thereby improving the accuracy of mobile data center resource allocation prediction.

[0006] The technical solution adopted in this invention is as follows:

[0007] A task-driven method for predicting mobile data center resource allocation for large mobile devices, specifically including the following steps:

[0008] (1) Data storage and processing method settings: Based on factors such as the physical structure and actual use of mobile devices, determine the physical structure of the mobile data center, data storage methods and parameters, and data processing methods and parameters;

[0009] (2) Data storage resource configuration prediction: Based on the relevant methods and parameters of data storage tasks, such as data collection scale, generation scale, replication strategy and storage resource vacancy rate, the configuration of data storage resources is predicted;

[0010] (3) Data processing resource configuration prediction: Based on the relevant methods and parameters of data processing tasks, such as task type, number of tasks, data requirements, allocation strategy, task latency and idle rate of processing resources, the configuration of data processing resources is predicted.

[0011] (4) Iterative update: Collect the physical characteristics of mobile devices and mobile data centers, data storage and processing task-related parameters, resource configuration prediction results, and the resource configuration's satisfaction with tasks during actual use to form a resource configuration library. During use, resource configuration is adjusted and updated according to changes in devices and tasks, and the resource configuration library can be used to make resource configuration recommendations.

[0012] As a preferred embodiment, the specific steps of step (2) of the task-driven mobile data center resource allocation prediction method for large mobile devices described above are as follows:

[0013] ①Based on the amount of data generated by all sensors within the physical area per unit time and the amount of intermediate data generated by the server itself, the amount of raw data that each server needs to store per unit time.

[0014] ② Calculate the amount of replica data that should be stored on each server per unit time based on the replica storage strategy of the mobile data center;

[0015] ③ Calculate the required storage capacity for each server based on the server storage space vacancy rate and data retention time, and then calculate the number of storage modules required for each server based on the average storage capacity of each storage module.

[0016] As a preferred embodiment, the specific steps of step (3) of the task-driven mobile data center resource allocation prediction method for large mobile devices described above are as follows:

[0017] ① Divide data processing tasks on mobile devices into two categories: periodically executed routine triggering tasks and real-time executed decision analysis tasks. Calculate the number of routine triggering tasks and decision analysis tasks that need to be processed on mobile devices per unit time, the amount of data to be processed and the amount of computation for each task, and the storage location of the data to be processed for each task.

[0018] ② Assign tasks to each server according to the task allocation strategy, and calculate the data transfer time for each task to be processed to be transmitted to the task processing server;

[0019] ③ Calculate the sum of the computational workload of each server's tasks within a unit of time, and the maximum transmission time for each task to complete the data transmission to be processed.

[0020] ④ Based on the task response time threshold allowed by the mobile device and the computing resource vacancy rate, calculate the amount of computing power required by each server, and then, based on the computing power of each data processing module, determine the number of data processing modules that should be configured on each server.

[0021] As a preferred embodiment, the specific steps of step (4) of the task-driven mobile data center resource allocation prediction method for large mobile devices described above are as follows:

[0022] ① Establish a resource allocation database: Collect the characteristics and uses of mobile devices, the physical structure of mobile data centers, parameters related to data storage and processing tasks, resource allocation prediction results, and the degree to which resource allocation meets data storage and processing tasks during actual use of mobile devices, to form a resource allocation database for mobile data centers.

[0023] ② Storage resource configuration update: If the sensor data acquisition frequency is adjusted, data acquisition equipment is added, data copy strategy is changed, or data retention period is changed during actual use of the mobile device, the storage resources are adjusted according to step (2), and the adjusted results are saved in the resource configuration library;

[0024] ③ Processing resource configuration updates: If the task processing requirements change, task types or number of tasks are added during actual use of the mobile device, the processing resources are adjusted according to step (3), and the adjusted results are saved in the resource configuration library;

[0025] ④ Resource configuration recommendation: When predicting resource configuration for new large-scale mobile data centers, if the accuracy requirements are not strict, the resource configuration library can be queried to recommend similar resource configurations based on the physical characteristics of mobile devices and mobile data centers and the needs of the task. The specific similarity measurement method is determined according to the actual application.

[0026] The beneficial effects of this invention are as follows:

[0027] (1) Organize multiple servers of large mobile devices through mobile data centers and use a task-driven approach to accurately predict the resource configuration of each server.

[0028] (2) By comprehensively considering multiple influencing factors of data storage and processing tasks, the accuracy of prediction results for data storage and processing resource allocation in mobile data centers can be improved.

[0029] (3) By setting the resource vacancy rate and iteratively updating the operation, the scalability of the data resource allocation prediction method in different mobile data centers and its adaptability throughout the life cycle of mobile devices are improved. Attached Figure Description

[0030] Figure 1 This invention relates to the network architecture of a task-driven method for predicting resource allocation in mobile data centers for large mobile devices.

[0031] Figure 2 This is the specific process of the task-driven mobile data center resource allocation prediction method for large mobile devices according to the present invention.

[0032] Figure 3 This is the resource allocation prediction process in an embodiment of the present invention. Detailed Implementation

[0033] The technical solution of the present invention will be further described below with reference to the embodiments.

[0034] The resource allocation prediction method proposed in this invention is mainly used in mobile data centers composed of various data acquisition, storage, and processing devices on large mobile devices such as aircraft and ships. It predicts the resource allocation for data storage and processing in mobile data centers to ensure that the data storage and processing resources meet relevant task requirements without causing excessive resource waste. Since the data acquisition, storage, and processing tasks of each server on large mobile devices differ, their storage and processing resources need to be rationally allocated according to the tasks. Centralized allocation can lead to errors in the total resource allocation and uneven resource distribution across servers, resulting in problems such as data loss and processing timeouts. Therefore, it is necessary to predict storage and processing resources based on the characteristics and requirements of each server's data storage and processing tasks to improve the accuracy of the resource allocation prediction results.

[0035] See the network architecture of the mobile data center. Figure 1 Each server is responsible for storing and processing data within a physical area. Each sensor is connected to only one connector, and each connector is connected to only one server. All devices are connected to each other through different types of data buses, and servers are connected to each other through high-speed buses.

[0036] Current resource allocation prediction methods for large mobile devices mainly employ traditional centralized architecture prediction methods, relying on experience to predict the entire mobile device's data storage and processing resource requirements. On the one hand, the mismatch between actual application needs and experience can easily lead to inaccurate prediction results; on the other hand, the inability to determine the resource configuration of each server can result in problems such as data loss or task timeouts on overloaded servers. The proposed task-driven mobile data center resource allocation prediction method for large mobile devices describes the prediction process as follows: Figure 2 As shown. This invention starts from the task that triggers resource demand, comprehensively considers the physical characteristics and task requirements of mobile devices, as well as data storage and processing methods and parameters related to mobile data centers, to improve the accuracy of prediction results. The specific resource allocation prediction method includes the following four steps:

[0037] 1. Data storage and processing method settings

[0038] Based on the physical structure and actual use of mobile devices, determine the data storage and processing methods for mobile data centers, including the number and location of servers, sensor-to-server data transmission, server data copy storage strategy, data processing task allocation strategy, and the types and quantities of data processing tasks.

[0039] 2. Data storage resource configuration prediction: The data storage resource configuration of each server is determined based on the amount of data generated by sensors within the physical area where each server is located, the replica storage strategy, and the types of daily tasks. The specific calculation process is as follows:

[0040] ① Calculate the original data volume: If server P i The amount of data generated per unit time by all sensors within the physical area is Sdata i Server P i The amount of intermediate data generated per unit time is Cdata i Then each server P i The amount of raw data that needs to be stored per unit of time, Rdata i =Sdata i +Cdata i , where 1≤i≤m, and m is the number of servers configured on the mobile device;

[0041] ② Calculate the amount of replica data: If server P i The replica storage vector of the above data is Indicates server P j Stored P i The data is a copy, and the copy data size is... P represents j P not stored i Upload a copy of the data, Indicates server P i Each server P stores its own raw data. i The actual amount of data stored Where 1≤i,j≤m, m is the number of servers configured on the mobile device, and the replica storage vector. The value is determined by the replica storage strategy of the mobile data center;

[0042] ③ Data storage resource configuration: If the server storage space vacancy rate is α and the data retention time is T units of time, then each server P i The storage capacity is Scap i =T×data i / (1-α); If the average storage capacity of each storage module is D, then each server P i The number of storage modules required to be configured is Where 1≤i≤m, m is the number of servers configured on the mobile device, and the values ​​of parameters α and T are set according to the specific purpose of the mobile device.

[0043] 3. Data Processing Resource Allocation Prediction: Based on factors such as the data processing tasks of mobile devices, data storage, and task execution mechanisms, the allocation of data and gooseberry resources is determined. The specific calculation is as follows:

[0044] ①Task Statistics: The vast majority of data processing tasks on mobile devices are sensor data analysis tasks, including periodically executed routine triggering tasks and real-time decision analysis tasks. If the routine triggering tasks and decision analysis tasks that need to be processed on the mobile device per unit time are JP={JP1,JP2,...,JP...} q} and JR = {JR1,JR2,...,JR h} where q and h are the number of regular trigger-type tasks and the number of decision analysis tasks, respectively; it is assumed that the computational load of each task is proportional to the amount of data processed. in and These are the regular triggering tasks JP i and decision analysis tasks JR j The computational complexity coefficient, and Task JP i and JR j The amount of data to be processed is 1≤i≤q, 1≤j≤h; Task JP i and JR j The raw data to be processed comes from server P. j The parts are respectively and but Where 1≤k≤m,

[0045] ②Task Allocation: To reduce the additional overhead of task migration and improve task response speed, each task is typically executed on only one server, and tasks are assigned to the server with the shortest data transfer time; if each regularly triggered task JP i Or decision analysis task JR j On server P k If executed above, it is denoted as Loc(JP) i ) = P k Or Loc(JR) j ) = P k Server P k Processing Task JP i Or JR j Waiting time and The amount of data to be processed is respectively or Transmitted to server P k The time required, i.e. and Where τ is the time required to transmit a unit amount of data per unit length on a high-speed bus between servers, and d s,k This represents two servers P. sand P k The length of the high-speed bus between them, if s = k represents the server P s The distance to itself, i.e., ds,s=0,1≤k,s≤m;

[0046] ③Calculation workload statistics: P per server per unit time k Total computational cost Z(P) k The sum of the computational cost of all tasks assigned to it for execution is... Due to bus preemption during data transmission, each server P s To server P k The time t for the data to be processed in the transmission task s,k For each task in P s The data to be processed is transmitted to P. k The total time, that is Where 1≤k, s≤m; server P k The maximum waiting time before executing a task is That is, other servers send to P k The maximum time for data to be processed in the transmission task;

[0047] ④ Data processing resource configuration: The allowable task response time threshold for mobile devices is Δt, and the idle rate of computing resources is β. Then, for each server P k Minimum computational load required If each data processing module can provide C computational resources, then each server P k The number of data processing modules that should be configured is:

[0048] 4. Iterative updates

[0049] ① Establish a resource allocation database: Collect the characteristics and uses of mobile devices, the physical structure of mobile data centers, parameters related to data storage and processing tasks, resource allocation prediction results, and the degree to which resource allocation meets data storage and processing tasks during actual use of mobile devices, to form a resource allocation database for mobile data centers.

[0050] ② Storage resource configuration update: If the sensor data acquisition frequency is adjusted, data acquisition devices are added, data copying strategy is changed, or data retention period is changed during actual use of the mobile device, the storage resources are adjusted according to step 2, and the results of the adjustment are saved in the resource configuration library;

[0051] ③ Handling resource configuration updates: If the task processing requirements change, task types or number of tasks are added during actual use of the mobile device, the processing resources are adjusted according to step 3, and the adjusted results are saved to the resource configuration library;

[0052] ④ Resource configuration recommendation: When predicting resource configuration for new large-scale mobile data centers, if the accuracy requirements are not strict, the resource configuration library can be queried to recommend similar resource configurations based on the physical characteristics of mobile devices and mobile data centers and the needs of the task. The specific similarity measurement method is determined according to the actual application.

[0053] The resource allocation prediction process in this embodiment is described in [reference]. Figure 3 The specific process is as follows: If a mobile data center of a certain mobile device consists of 3 servers {P1, P2, P3} and 6 sensors {S1, S2, S3, S4, S5, S6}, the resource configuration prediction process of the mobile data center of this mobile device is as follows:

[0054] ① If the data replication storage strategy is P1 data backup to P2, P2 data direct backup to P3, and P3 data direct backup to P3; sensors S1-S4 collect data once per second, with each data collection being 1KB; sensors S5 and S6 collect data once every 10 seconds, with each data collection being 1MB; taking one minute as the time unit, the data collected by servers P1, P2, and P3 are 120KB, 120KB, and 12MB respectively; within one minute, servers P1, P2, and P3 generate intermediate results of 100KB, 500KB, and 8MB respectively, then the original data that P1, P2, and P3 need to store per minute are 220KB, 620KB, and 20MB respectively;

[0055] ② According to the replication strategy, servers P1, P2, and P3 need to store 20MB, 220KB, and 620KB of replica data per minute, respectively. If the data storage period is one year (calculated as 366 days), then the required data storage capacity for servers P1, P2, and P3 is 10404GB, 422GB, and 10605GB, respectively. If the storage resource vacancy rate is 50%, then the data storage capacity that servers P1, P2, and P3 should be configured with is 20808GB, 844GB, and 21210GB, respectively. If each storage device is a 500GB SSD, then the number of storage devices that servers P1, P2, and P3 should be configured with is 42, 2, and 43, respectively.

[0056] ③ If the mobile data center needs to execute 2 triggered tasks {JP1, JP2} and 3 analytical tasks {JR1, JR2, JR3} per second; task JP1 needs to process data from 100KB of server P1 and 200KB of server P2, with computational requirements of 1MIPS and 2MIPS respectively; task JP2 needs to process data from 100KB of server P1, 200KB of server P2 and 1MB of server P3, with computational requirements of 10MIPS, 20MIPS and 100MIPS respectively; task JR1 needs to process data from 100KB of server P1, with a computational requirement of 10MIPS; task JR2 needs to process data from 200KB of server P2, with a computational requirement of 20MIPS; task JR3 needs to process data from 300KB of server P1, 200KB of server P2 and 1MB of server P3, with computational requirements of 30MIPS, 20MIPS and 100MIPS respectively.

[0057] ④ If a task is assigned to the server with the largest amount of data to be processed, then tasks JP1 and JP2 are assigned to servers P2 and P3 respectively, and tasks JR1, JR2, and JR3 are assigned to servers P1, P2, and P3 respectively. Server P1 needs to transmit 100KB of data requested by task JP1 to P2, and transmit 100KB of data from JP2 and 300KB of data from JR3 to P3. Server P2 needs to transmit 200KB of data requested by task JP2 and 200KB of data requested by JR3 to P3. If 100KB of data is transmitted between servers every millisecond, then server P3 needs to wait 2 milliseconds for data from task JP2 from P2, 3 milliseconds for data from task JR3 from P1, and 2 milliseconds for data from task JR3 from P2.

[0058] ⑤ If the completion time of each task is 10 milliseconds, then server P2 needs to complete task JP1 within 9 milliseconds and task JR2 within 10 milliseconds; server P1 needs to complete task JR1 within 10 milliseconds; server P3 needs to complete task JP2 within 8 milliseconds and task JR3 within 7 milliseconds; the computing power required by servers P1, P2 and P3 is 1000 MIPS, 11111 MIPS and 14286 MIPS respectively; if the vacancy rate is 50%, then the computing resources that servers P1, P2 and P3 can provide are 2000 MIPS, 22222 MIPS and 28572 MIPS respectively; if the computing power that each computing module can provide is 3000 MIPS, then the number of computing modules that servers P1, P2 and P3 need to be configured with are 1, 8 and 10 respectively.

[0059] ⑥ Add the characteristics of mobile devices, mobile data center architecture, task requirements and other parameters, resource configuration results and actual usage results of mobile devices to the resource configuration library, and update it according to the parameter adjustments during the usage process.

[0060] The embodiments described herein are merely exemplary implementations. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. These embodiments are provided to fully and completely disclose the invention and to fully convey the scope of the invention to those skilled in the art. The terminology used in the exemplary embodiments illustrated in the accompanying drawings is not intended to limit the invention. In the accompanying drawings, the same units / elements are referred to by the same reference numerals.

[0061] Unless otherwise stated, the terms used herein (including technical terms) have the common understanding meaning to those skilled in the art; in addition, it is understood that terms defined by commonly used dictionaries should be understood to have a meaning consistent with the context of their relevant field, and should not be understood to have an idealized or overly formal meaning.

Claims

1. A task-driven method for predicting mobile data center resource allocation in large mobile devices, characterized in that, The method specifically includes the following steps: (1) Data storage and processing method settings Based on the physical structure and actual use of mobile devices, determine the data storage and processing methods for mobile data centers, including the number and location of servers, sensor-to-server data transmission, server data copy storage strategy, data processing task allocation strategy, and data processing task types and quantities. (2) Data storage resource configuration prediction: The data storage resource configuration of each server is determined based on the amount of data generated by sensors in the physical area where each server is located, the replica storage strategy, and the types of daily tasks. The specific calculation process is as follows: ① Calculate the original data volume: If the server The amount of data generated per unit time by all sensors within the physical area is ,server The amount of intermediate data generated per unit time is Then each server P i The amount of raw data that needs to be stored per unit of time , where 1≤i≤m, and m is the number of servers configured on the mobile device; ② Calculate the amount of replica data: If server P i The replica storage vector of the above data is , Indicates server P j Stored P i The data is a copy, and the copy data size is... , P represents j P not stored i Upload a copy of the data, Indicates server P i Each server P stores its own raw data. i The actual amount of data stored Where 1≤i, j≤m, m is the number of servers configured on the mobile device, and the replica storage vector. The value is determined by the replica storage strategy of the mobile data center; ③ Data storage resource configuration: If the server storage space vacancy rate is α and the data retention time is T units of time, then each server P i The storage capacity is ; If the average storage capacity of each storage module is D, then each server P i The number of storage modules required to be configured is Where 1≤i≤m, m is the number of servers configured on the mobile device, and the values ​​of parameters α and T are set according to the specific purpose of the mobile device; (3) Data processing resource allocation prediction: The data processing resource allocation is determined based on factors such as the data processing tasks of mobile devices, data storage, and task execution mechanisms. The specific calculation is as follows: ① Task Statistics: Data processing tasks on mobile devices are sensor data analysis tasks, including periodically executed routine triggering tasks and real-time decision analysis tasks. If the routine triggering tasks and decision analysis tasks that need to be processed on the mobile device per unit time are respectively... and Where q and h are the number of regular trigger-type tasks and the number of decision analysis-type tasks, respectively; it is assumed that the computational load of each task is proportional to the amount of data processed. , ,in and These are the regular triggering tasks JP i and decision analysis tasks JR j The computational complexity coefficient, and Task JP i and JR j The amount of data to be processed is 1≤i≤q, 1≤j≤h; Task JP i and JR j The raw data to be processed comes from server P. j The parts are respectively and ,but , Where 1 ≤ k ≤ m, , ; ②Task Allocation: To reduce the additional overhead of task migration and improve task response speed, each task is typically executed on only one server, and tasks are assigned to the server with the shortest data transfer time; if each regularly triggered task JP i Or decision analysis task JR j On server P k If executed above, it is recorded as ,or Server P k Processing Task JP i Or JR j Waiting time and The amount of data to be processed is respectively or Transmitted to server P k The time required, i.e. and ,in d is the time required to transfer a unit amount of data per unit length on a high-speed bus between servers. s,k This represents two servers P. s and P k The length of the high-speed bus between them, if s=k represents the server P s Its own distance, that is ; ③Calculation workload statistics: P per server per unit time k Total computational cost Z(P) k The sum of the computational cost of all tasks assigned to it for execution is... Due to bus preemption during data transmission, each server P... s To server P k Time to process data for transmission task For each task in P s The data to be processed is transmitted to P. k The total time, that is Where 1 ≤ k, s ≤ m; server P k The maximum waiting time before executing a task is That is, other servers send to P k The maximum time for data to be processed in the transmission task; ④ Data processing resource configuration: The allowed task response time threshold for mobile devices is [value missing]. If the vacancy rate of computing resources is β, then for each server P k Minimum computational load required ; If each data processing module can provide C computational resources, then each server P k The number of data processing modules that should be configured is: ; (4) Iterative update ① Establish a resource allocation database: Collect the characteristics and uses of mobile devices, the physical structure of mobile data centers, parameters related to data storage and processing tasks, resource allocation prediction results, and the degree to which resource allocation meets data storage and processing tasks during actual use of mobile devices, to form a resource allocation database for mobile data centers. ② Storage resource configuration update: If the sensor data acquisition frequency is adjusted, data acquisition equipment is added, data copy strategy is changed, or data retention period is changed during actual use of the mobile device, the storage resources are adjusted according to step (2), and the adjusted results are saved in the resource configuration library; ③ Processing resource configuration updates: If the task processing requirements change, task types or number of tasks are added during actual use of the mobile device, the processing resources are adjusted according to step (3), and the adjusted results are saved in the resource configuration library; ④ Resource configuration recommendation: When predicting the resource configuration of new large-scale mobile data centers, the resource configuration library is queried, and similar resource configurations are recommended based on the physical characteristics of mobile devices and mobile data centers and the needs of tasks. The specific similarity measurement method is determined according to the actual application.

Images