CN114610475A - Training method of intelligent resource arrangement model - Google Patents

Training method of intelligent resource arrangement model Download PDF

Info

Publication number
CN114610475A
CN114610475A CN202011401075.2A CN202011401075A CN114610475A CN 114610475 A CN114610475 A CN 114610475A CN 202011401075 A CN202011401075 A CN 202011401075A CN 114610475 A CN114610475 A CN 114610475A
Authority
CN
China
Prior art keywords
resource arrangement
arrangement model
model
encryption information
training
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202011401075.2A
Other languages
Chinese (zh)
Inventor
明中行
杨术
萧伟
李忠鹏
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Huawei Technologies Co Ltd
Shenzhen Research Institute Tsinghua University
Original Assignee
Huawei Technologies Co Ltd
Shenzhen Research Institute Tsinghua University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Huawei Technologies Co Ltd, Shenzhen Research Institute Tsinghua University filed Critical Huawei Technologies Co Ltd
Priority to CN202011401075.2A priority Critical patent/CN114610475A/en
Publication of CN114610475A publication Critical patent/CN114610475A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00—Arrangements for program control, e.g. control units
    • G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/46—Multiprogramming arrangements
    • G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
    • G06F9/5005—Allocation of resources, e.g. of the central processing unit [CPU] to service a request
    • G06F9/5027—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resource being a machine, e.g. CPUs, Servers, Terminals
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F21/00—Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
    • G06F21/60—Protecting data
    • G06F21/602—Providing cryptographic facilities or services
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • G06N20/20—Ensemble learning

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Software Systems (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Medical Informatics (AREA)
  • Computing Systems (AREA)
  • Data Mining & Analysis (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Mathematical Physics (AREA)
  • Artificial Intelligence (AREA)
  • Health & Medical Sciences (AREA)
  • Bioethics (AREA)
  • General Health & Medical Sciences (AREA)
  • Computer Hardware Design (AREA)
  • Computer Security & Cryptography (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The application is applicable to the technical field of computers, and provides a training method of an intelligent resource arrangement model, which comprises the following steps: the method comprises the steps that control equipment obtains first encryption information of a global resource arrangement model and sends the first encryption information to each computing cluster in a Federal learning alliance; receiving second encryption information of the updated local resource arrangement model returned by each computing cluster; and updating the global resource arrangement model according to all the second encryption information to obtain a new global resource arrangement model. According to the method, the distributed computing clusters are independently computed by adopting a federal learning method, the computing results are gathered to the control equipment, and a global resource arrangement model is obtained, and the global resource arrangement model can accurately and quickly determine the resource arrangement strategy, so that the capability of the edge computing clusters is fully exerted, the average completion time of tasks is minimized, the number of the completed tasks is maximized, and the resource utilization rate and the computing efficiency are improved.

Description

Training method of intelligent resource arrangement model
Technical Field
The application belongs to the technical field of computers, and particularly relates to a training method of an intelligent resource arrangement model.
Background
With the rapid development of 5G and intelligent applications, the number of devices and data volume in the network increases rapidly, and the traditional centralized cloud computing is difficult to meet the requirements of new applications which are rapidly developed. In order to provide low-delay and high-quality computing service, the computing capabilities of the network edges are integrated, in the prior art, the computing capabilities of the network edges are integrated, different resources are reasonably distributed according to the performance, parameters and the like of different computing clusters, and a reasonable resource arrangement strategy is determined.
However, when the traditional network resource management is used for resource arrangement, the traditional network resource management is mainly based on manual and optimization algorithms or heuristic methods, the complexity is high, the self-adaptive capacity is poor, and the requirements of massive, dynamic and differentiated edge computing users are difficult to meet. Some artificial intelligence based resource orchestration, while better able, requires global resource information, which is often difficult to achieve for decentralized, self-organizing edge computing clusters.
Disclosure of Invention
The embodiment of the application provides a training method of an intelligent resource arrangement model, which can solve the problems that when resource arrangement is carried out in traditional network resource management, the traditional network resource management is mainly based on a manual and optimization algorithm or a heuristic method, the complexity is high, the self-adaptive capacity is poor, and the massive, dynamic and differentiated requirements of edge computing users are difficult to meet.
In a first aspect, an embodiment of the present application provides a training method for an intelligent resource arrangement model, which is applied to a control device, and the method includes:
acquiring first encryption information of a global resource arrangement model, and sending the first encryption information to each computing cluster in a Federal learning alliance, wherein the first encryption information is used for updating a local resource arrangement model of each computing cluster;
receiving second encryption information of the updated local resource orchestration model returned by each computing cluster;
and updating the global resource arrangement model according to all the second encrypted information to obtain a new global resource arrangement model.
Further, the updating the global resource arrangement model according to all the second encryption information to obtain a new global resource arrangement model includes:
decrypting the second encrypted information according to a preset decryption rule to obtain update information corresponding to each computing cluster;
and training the global resource arrangement model according to all the updating information to obtain a new global resource arrangement model.
Further, the training method of the initial global resource arrangement model comprises the following steps:
acquiring a first sample training set; the first sample training set comprises historical task information and resource arrangement strategy labels corresponding to the historical task information;
and training the original resource arrangement model according to the first sample training set to obtain an initial global resource arrangement model.
Further, the sample training set further includes preset special case task information and a resource arrangement strategy label corresponding to the preset special case task information.
Further, after the updating the global resource arrangement model according to all the second encryption information to obtain a new global resource arrangement model, the method further includes:
and if a model information acquisition request sent by the computing cluster is received, sending the new global resource arrangement model and/or third encryption information corresponding to the new global resource arrangement model to the computing cluster.
In a second aspect, an embodiment of the present application provides a training method for an intelligent resource arrangement model, which is applied to a computing cluster, and the method includes:
acquiring first encryption information sent by control equipment;
updating the local resource arrangement model according to the first encryption information to obtain an updated local resource arrangement model;
and sending second encryption information of the updated local resource arrangement model to the controller, wherein the second encryption information is used for updating the global resource arrangement model of the controller.
Further, the training method of the initial local resource arrangement model comprises the following steps:
acquiring an original resource arrangement model and a second sample training set;
and training the initial intelligent resource arrangement model according to the second sample training set to obtain an initial local resource arrangement model.
Further, the updating the local resource arrangement model according to the first encryption information to obtain an updated local resource arrangement model includes:
performing iterative computation according to a gradient lifting tree algorithm and the first encryption information, and constructing a regression tree corresponding to the local resource arrangement model;
and when the regression tree meets the preset condition, acquiring the current intelligent resource arrangement model as the updated local resource arrangement model.
In a third aspect, an embodiment of the present application provides a method for determining a resource orchestration policy, including:
acquiring task information of a task to be distributed;
inputting the task information into a preset target intelligent resource arrangement model for processing to obtain a resource arrangement strategy corresponding to the task information; wherein the target intelligent resource arrangement model is obtained by the training method of the intelligent resource arrangement model according to the first aspect.
In a fourth aspect, an embodiment of the present application provides a control apparatus, including:
the system comprises a first processing unit, a second processing unit and a third processing unit, wherein the first processing unit is used for acquiring first encryption information of a global resource arrangement model and sending the first encryption information to each computing cluster in a Federal learning alliance, and the first encryption information is used for updating a local resource arrangement model of each computing cluster;
a receiving unit, configured to receive second encrypted information of the updated local resource orchestration model returned by each computing cluster;
and the second processing unit is used for updating the global resource arrangement model according to all the second encryption information to obtain a new global resource arrangement model.
Further, the second processing unit is specifically configured to:
decrypting the second encrypted information according to a preset decryption rule to obtain update information corresponding to each computing cluster;
and training the global resource arrangement model according to all the updating information to obtain a new global resource arrangement model.
Further, the control apparatus further includes:
an obtaining unit, configured to obtain a first sample training set; the sample training set comprises historical task information and a resource arrangement strategy label corresponding to the historical task information;
and the third processing unit is used for training the original resource arrangement model according to the first sample training set to obtain an initial global resource arrangement model.
Further, the sample training set further includes preset special case task information and a resource arrangement strategy label corresponding to the preset special case task information.
Further, the control apparatus further includes:
and the fourth processing unit is configured to send the new global resource arrangement model and/or third encryption information corresponding to the new global resource arrangement model to the computing cluster if a model information acquisition request sent by the computing cluster is received.
In a fifth aspect, an embodiment of the present application provides a computing cluster, including:
a first acquisition unit configured to acquire first encryption information sent by a control device;
the updating unit is used for updating the local resource arrangement model according to the first encryption information to obtain an updated local resource arrangement model;
and the sending unit is used for sending the second encrypted information of the updated local resource arrangement model to the controller, and the second encrypted information is used for updating the global resource arrangement model of the controller.
Further, the computing cluster further includes:
the second acquisition unit is used for acquiring the original resource arrangement model and a second sample training set;
and the training unit is used for training the initial intelligent resource arrangement model according to the second sample training set to obtain an initial local resource arrangement model.
Further, the update unit is specifically configured to:
performing iterative computation according to a gradient lifting tree algorithm and the first encryption information, and constructing a regression tree corresponding to the local resource arrangement model;
and when the regression tree meets the preset condition, acquiring the current intelligent resource arrangement model as the updated local resource arrangement model.
In a sixth aspect, an embodiment of the present application provides an apparatus for determining a resource orchestration policy, including:
the acquisition unit is used for acquiring task information of a task to be distributed;
the processing unit is used for inputting the task information into a preset target intelligent resource arrangement model for processing to obtain a resource arrangement strategy corresponding to the task information; the target intelligent resource arrangement model is obtained by the training method of the intelligent resource arrangement model in the first aspect.
In a seventh aspect, an embodiment of the present application provides a control device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor, when executing the computer program, implements the training method for the intelligent resource orchestration model according to the first aspect.
In an eighth aspect, embodiments of the present application provide a computing cluster, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and when the processor executes the computer program, the processor implements the method for training the intelligent resource orchestration model according to the second aspect.
In a ninth aspect, an embodiment of the present application provides a control device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the method for determining a resource orchestration policy according to the third aspect when executing the computer program.
In a tenth aspect, the present application provides a computer-readable storage medium, which stores a computer program, where the computer program, when executed by a processor, implements the method for training an intelligent resource orchestration model according to the first aspect.
In an eleventh aspect, the present application provides a computer-readable storage medium, where a computer program is stored, and when executed by a processor, the computer program implements the method for determining a resource orchestration policy according to the second aspect.
In a twelfth aspect, the present application provides a computer-readable storage medium, where a computer program is stored, and when executed by a processor, the computer program implements the method for training the intelligent resource orchestration model according to the third aspect.
In the embodiment of the application, a control device acquires first encryption information of a global resource arrangement model, and sends the first encryption information to each computing cluster in a Federal learning alliance, wherein the first encryption information is used for updating a local resource arrangement model of each computing cluster; receiving second encryption information of the updated local resource orchestration model returned by each computing cluster; and updating the global resource arrangement model according to all the second encryption information to obtain a new global resource arrangement model. According to the method, when the number of users is large and the network condition is complex, the distributed computing clusters are independently computed by adopting a federal learning method, the computing results are collected to the control equipment, a global resource arrangement model is obtained, the global resource arrangement model can accurately and quickly determine the resource arrangement strategy, and the resource utilization rate and the computing efficiency are improved.
On the other hand, the computing cluster acquires first encryption information sent by the controller; updating the local resource arrangement model according to the first encryption information to obtain an updated local resource arrangement model; and sending second encryption information of the updated local resource arrangement model to the controller, wherein the second encryption information is used for updating the global resource arrangement model of the controller. According to the method, when the number of users is large and the network condition is complex, the distributed computing clusters are independently computed by adopting a federal learning method, the computing results are collected to the control equipment, a global resource arrangement model is obtained, the global resource arrangement model can accurately and quickly determine the resource arrangement strategy, and the resource utilization rate and the computing efficiency are improved.
On the other hand, the determining device of the resource arrangement strategy acquires task information of the tasks to be distributed; and inputting the task information into a preset target intelligent resource arrangement model for processing to obtain a resource arrangement strategy corresponding to the task information. The target intelligent resource arrangement model used in the method is obtained by independently calculating the dispersed calculation clusters through a federal learning method, summarizing calculation results to control equipment and summarizing calculation through the control equipment. The target intelligent resource arrangement model can accurately and quickly determine the resource arrangement strategy, and improves the resource utilization rate and the calculation efficiency.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the embodiments or the prior art descriptions will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without inventive exercise.
FIG. 1 is a schematic flow chart diagram of a training method for an intelligent resource orchestration model according to a first embodiment of the present application;
FIG. 2 is a schematic flow chart diagram of a method for training an intelligent resource orchestration model according to a second embodiment of the present application;
FIG. 3 is a schematic flow chart of a method for determining a resource orchestration policy according to a third embodiment of the present application;
FIG. 4 is a schematic diagram of a control device provided in a fourth embodiment of the present application;
FIG. 5 is a schematic diagram of a compute cluster provided in a fifth embodiment of the present application;
FIG. 6 is a schematic diagram of a resource orchestration policy determining apparatus according to a sixth embodiment of the present application;
fig. 7 is a schematic diagram of a control apparatus provided in a seventh embodiment of the present application;
FIG. 8 is a schematic diagram of a compute cluster provided in an eighth embodiment of the present application;
fig. 9 is a schematic diagram of a resource orchestration policy determining device according to a ninth embodiment of the present application.
Detailed Description
In the following description, for purposes of explanation and not limitation, specific details are set forth, such as particular system structures, techniques, etc. in order to provide a thorough understanding of the embodiments of the present application. It will be apparent, however, to one skilled in the art that the present application may be practiced in other embodiments that depart from these specific details. In other instances, detailed descriptions of well-known systems, devices, circuits, and methods are omitted so as not to obscure the description of the present application with unnecessary detail.
It will be understood that the terms "comprises" and/or "comprising," when used in this specification and the appended claims, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof.
It should also be understood that the term "and/or" as used in this specification and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.
As used in this specification and the appended claims, the term "if" may be interpreted contextually as "when", "upon" or "in response to" determining "or" in response to detecting ". Similarly, the phrase "if it is determined" or "if a [ described condition or event ] is detected" may be interpreted contextually to mean "upon determining" or "in response to determining" or "upon detecting [ described condition or event ]" or "in response to detecting [ described condition or event ]".
Furthermore, in the description of the present application and the appended claims, the terms "first," "second," "third," and the like are used for distinguishing between descriptions and not necessarily for describing or implying relative importance.
Reference throughout this specification to "one embodiment" or "some embodiments," or the like, means that a particular feature, structure, or characteristic described in connection with the embodiment is included in one or more embodiments of the present application. Thus, appearances of the phrases "in one embodiment," "in some embodiments," "in other embodiments," or the like, in various places throughout this specification are not necessarily all referring to the same embodiment, but rather "one or more but not all embodiments" unless specifically stated otherwise. The terms "comprising," "including," "having," and variations thereof mean "including, but not limited to," unless expressly specified otherwise.
With the rapid development of 5G and intelligent technologies, many new delay-sensitive applications have emerged, for example, industrial control applications remotely control devices via the internet; the equipment monitoring application analyzes the running state of the equipment through the Internet to realize intelligent equipment management; the intelligent driving application intelligently assists the vehicle in making driving decisions by collecting a large amount of data such as videos and sensing. The delay of part remote industrial control application is required to be lower than 20ms, the delay of an intelligent driving scene is required to be lower than 10ms, and the lower the delay of equipment safety monitoring application is required to be, the better the equipment safety monitoring application is.
Conventional centralized cloud computing cannot meet the requirements of the new applications, because the cloud computing center is usually at a certain distance from the user and causes additional delay, and for new intelligent scenarios, such as critical industrial internet applications, remote monitoring, intelligent driving, and the like, fatal influences exist, and huge losses or potential safety hazards may be caused. Therefore, computation processing needs to be performed close to the user data source end to provide a reliable experience for the user.
Edge calculation provides a new direction to solve the above problem. By deploying the computing server at the edge close to the user and providing the information technology service environment and the cloud computing capability at the edge of the network or at the position close to the user, the communication delay can be greatly reduced, the network efficiency is optimized, and the user experience is improved.
However, in the practical application process, the existing edge computing technology faces the challenges of low resource utilization rate and low computing efficiency: edge computing typically provides computing services for multiple users for different applications that have different communication resources and computing resource requirements. However, the existing resource arrangement method relies on user autonomous management or a simple optimization algorithm, and the resource utilization rate is low because the computing units are allocated according to the coarse granularity of the physical machine or the virtual machine.
Therefore, in order to solve the above problems, the present application provides a training method for an intelligent resource arrangement model and a determination method for a resource arrangement strategy.
Referring to fig. 1, fig. 1 is a schematic flowchart of a training method for an intelligent resource orchestration model according to a first embodiment of the present application. An execution subject of the training method of the intelligent resource arrangement model in the embodiment is the control device. The training method of the intelligent resource arrangement model shown in fig. 1 can comprise the following steps:
s101: the method comprises the steps of obtaining first encryption information of a global resource arrangement model, and sending the first encryption information to each computing cluster in a Federal learning alliance, wherein the first encryption information is used for updating a local resource arrangement model of each computing cluster.
In this embodiment, the training of the intelligent resource orchestration model is based on federal learning. The Federal learning technology has the advantages that the adaptability of the Federal learning technology in the aspects of guaranteeing information safety and data privacy during data exchange is ensured, and the Federal learning technology has great advantages in the aspect of developing efficient machine learning among distributed information units. The control equipment and the computing clusters jointly form a Federal learning alliance, each computing cluster can update the local model only based on local data without sharing global information, and the local model information of the computing clusters is gathered through the control equipment, so that a new global resource arrangement model is obtained.
In the control device, an initial global resource arrangement model is stored, and the initial global resource arrangement model may be obtained by pre-training the control device, or may be transplanted to the control device after training of other devices.
Wherein the control device may train the initial global resource orchestration model by: the control equipment acquires a first sample training set; the first sample training set comprises historical task information and corresponding resource arrangement strategy labels. The historical task information may include task-related parameters, such as the number of tasks, resource conditions required by the tasks, time limit for completion of the tasks, task completion requirements, and the like. The historical task information and the resource arrangement strategy label corresponding to the historical task information are preferred resource arrangement strategies corresponding to the historical task information, and the preferred resource arrangement strategies are optimal resource arrangement strategies obtained by considering the factors of resources, bandwidth, service quality, performance and the like of each computing cluster. It will be appreciated that the computing clusters in the resource orchestration strategy should belong to the Federal learning Association mentioned above.
For the richness and the authenticity of the sample, the more the number of the historical task information and the corresponding resource arrangement strategy labels included in the sample training set is, the better the number is.
And after the control equipment acquires the first sample training set, training the original resource arrangement model according to the first sample training set to obtain an initial global resource arrangement model. In the training process, the input of the initial global resource arrangement model is historical task information, and the output of the initial global resource arrangement model is the historical task information and the corresponding resource arrangement strategy.
Further, because the system runs in real time, some sudden situations or situations that have not been met before may be met in the running process, some special situation samples may be set in the training samples, and the sample training set further includes preset special situation task information and resource scheduling policy labels corresponding to the preset special situation task information. The special case samples are used for simulating network states which do not appear in history to train, an optimal solution is obtained according to the network states, and then when the conditions appear in the future, the optimal solution can be adaptively reflected, and an optimized resource scheduling strategy is obtained.
After the initial global resource arrangement model is obtained, the control device updates the initial global resource arrangement model according to the information sent by the computing clusters, and sends the encrypted information corresponding to the updated global resource arrangement model to each computing cluster after the updating. And each computing cluster updates the local resource arrangement model according to the received first encryption information, and sends the encryption information corresponding to the updated local resource arrangement model to the control equipment so as to control the equipment to continuously update the current global resource arrangement model. Therefore, multiple times of cyclic update learning are formed, and the control equipment updates the current global resource arrangement model after acquiring the encryption information sent by the computing cluster each time, so that the final global resource arrangement model is obtained finally.
The encryption information mentioned in this embodiment refers to an intermediate value of the global resource arrangement model in the construction process, and may be understood as a weight coefficient, and the structure of the global resource arrangement model may be determined by the intermediate value. Therefore, after the control device acquires the encryption information of the computing cluster, the current global resource arrangement model can be adjusted through the encryption information of the computing cluster, and after the computing cluster acquires the encryption information of the control device, the local resource arrangement model can also be adjusted through the encryption information of the control device.
It should be noted that, since the federate learning manner is adopted in this embodiment, information is not shared between each computing cluster and the control device, and the above-mentioned intermediate values need to use class labels in the computing process, and they can be used for reconstruction to find class information, there is a risk if the intermediate values are sent directly. In order to ensure security, each computing cluster and the control device cannot directly send and access the intermediate value, the control device first performs an encryption process before sending it to each computing cluster. During encryption, the first encrypted information can be obtained by encrypting the first encrypted information by using a homomorphic encryption algorithm. Homomorphic encryption for example, we define that the expression of a number u in an additivity homomorphic encryption mode is < u >, and we have < u > + < v > -u + v > for two arbitrary numbers u and v according to the additivity principle of homomorphic encryption.
After receiving the first encryption information, the computing cluster needs to update the local resource orchestration model with the intermediate value without directly accessing the intermediate value.
In this embodiment, how the control device obtains the updated global resource arrangement model is described by taking one update in the cyclic update as an example. The control equipment acquires first encryption information of the global resource arrangement model and sends the first encryption information to each computing cluster in the Federal learning alliance, and the first encryption information is used for updating the local resource arrangement model of each computing cluster.
S102: receiving second encryption information of the updated local resource orchestration model returned by each of the computing clusters.
After the control device sends the first encryption information to each computing cluster in the Federal learning alliance, the computing cluster downloads a current global resource arrangement model from the control device to serve as a local resource arrangement model, updates the local resource arrangement model according to the first encryption information to obtain an updated local resource arrangement model, and obtains second encryption information in the updating process. And the computing cluster sends second encryption information to the control equipment, wherein the second encryption information is used for updating the current global resource arrangement model of the control equipment.
The control device receives second encryption information of the updated local resource orchestration model returned by each computing cluster.
S103: and updating the global resource arrangement model according to all the second encryption information to obtain a new global resource arrangement model.
And the control equipment updates the global resource arrangement model according to all the second encryption information to obtain a new global resource arrangement model. In this embodiment, the global resource arrangement model is updated by using the second encryption information, and the global resource arrangement model may be updated by using a gradient lifting tree algorithm, or may be updated by using a neural network training method, which is not limited herein.
In one embodiment, the global resource orchestration model may be updated by neural network training. And after the control equipment acquires the second encrypted information sent by the computing clusters, decrypting the second encrypted information according to a preset decryption rule to obtain the updating information corresponding to each computing cluster.
And training the global resource arrangement model according to all the updating information to obtain a new global resource arrangement model. In this embodiment, the current global resource arrangement model is updated, and the current global resource arrangement model is trained in a machine learning manner to obtain an updated global resource arrangement model. During the training process, the encryption information may be used to adjust the training parameters.
In the training process, special case samples can be set to simulate network states which do not appear in history for training, an optimal solution is obtained according to the network states, and then when the situations appear in the future, the optimal solution can be reacted in a self-adaptive mode, and an optimal resource arrangement strategy is obtained.
After S103, the method may further include: and if a model information acquisition request sent by the computing cluster is received, sending the new global resource arrangement model and/or third encryption information corresponding to the new global resource arrangement model to the computing cluster. Here, after the control device obtains the new global resource arrangement model, it may obtain an intermediate value corresponding to the new global resource arrangement model, encrypt the intermediate value to obtain third encryption information, and when the computing cluster sends a model information obtaining request, send the new global resource arrangement model and/or the third encryption information corresponding to the new global resource arrangement model to the computing cluster.
In the embodiment of the application, a control device acquires first encryption information of a global resource arrangement model, and sends the first encryption information to each computing cluster in a Federal learning alliance, wherein the first encryption information is used for updating a local resource arrangement model of each computing cluster; receiving second encryption information of the updated local resource orchestration model returned by each computing cluster; and updating the global resource arrangement model according to all the second encryption information to obtain a new global resource arrangement model. According to the method, under the condition that the number of users is large and the network condition is complex, the distributed computing clusters are independently computed by adopting a federal learning method, the computing results are gathered to the control equipment, a global resource arrangement model is obtained, the global resource arrangement model can accurately and quickly determine the resource arrangement strategy, the capacity of the edge computing clusters is fully exerted, high-quality computing support is provided for delay sensitive application, resources are dynamically and intelligently arranged according to task characteristics, the average completion time of tasks is minimized, the number of the tasks is maximized, the problem in the aspect of computing efficiency is solved, and the resource utilization rate and the computing efficiency are improved.
Referring to fig. 2, fig. 2 is a schematic flowchart of a training method of an intelligent resource orchestration model according to a second embodiment of the present application. In this embodiment, an execution subject of the training method for the intelligent resource arrangement model is a computing cluster. The training method of the intelligent resource arrangement model shown in FIG. 2 can comprise the following steps:
s201: first encryption information sent by the control equipment is obtained.
In this embodiment, the execution agent is a computing cluster, which is a computer system that cooperates to perform computing work with a high degree of closeness through a set of loosely integrated computer software and/or hardware connections. A computing cluster may be viewed in a sense as a computer or a server. The computing cluster in this embodiment should belong to the federal learning alliance mentioned in the first embodiment.
The computing cluster may preset an initial local resource arrangement model, and the initial local resource arrangement model may be obtained by pre-training the computing cluster, or may be transplanted to the control device after training of other devices.
The training method of the initial local resource arrangement model comprises the following steps: and acquiring an original resource arrangement model and a second sample training set. And the second sample training set comprises historical task information and corresponding resource arrangement strategy labels. The historical task information can comprise common characteristic data samples of all computing clusters in the Federal learning alliance, and although the data of different computing clusters are independent of each other, due to some common characteristics contained in the edge computing nodes, some data samples containing the common characteristics can be found in the data of most clusters. These samples may be identified by their unique identifiers. At this time, data across the computing cluster may be encrypted according to an existing privacy protection mechanism, and then a data sample containing a common feature may be selected.
Specifically, the common characteristics are basic attribute characteristics of the computing cluster, such as CPU type, GPU type, memory size, storage space size, load condition, computational resource condition, and the like.
And the computing cluster trains the initial intelligent resource arrangement model according to the second sample training set to obtain an initial local resource arrangement model. In the training process, the input of the initial local resource arrangement model is historical task information, and the output of the initial local resource arrangement model is the historical task information and the corresponding resource arrangement strategy.
And acquiring a new global resource arrangement model from the control device as a local resource arrangement model. The control device can actively send the new global resource arrangement model to the computing cluster each time when acquiring the new global resource arrangement model, and can also send a request to the control device when the computing cluster needs to update the local resource arrangement model.
After the control device obtains the global resource arrangement model, first encryption information corresponding to the global resource arrangement model is sent to each computing cluster. The computing cluster obtains first encryption information sent by the control device, and the first encryption information is used for updating the local resource arrangement model by the computing cluster.
S202: and updating the local resource arrangement model according to the first encryption information to obtain an updated local resource arrangement model.
In the first embodiment, the encrypted information actually refers to an intermediate value of the global resource arrangement model in the construction process, and may be understood as a weighting coefficient. The control equipment obtains the first encryption information through homomorphic encryption during the first encryption information, and after the computing cluster obtains the first encryption information, the computing cluster obtains an intermediate value through utilizing an additivity principle of homomorphic encryption, and updates the local resource arrangement model to obtain an updated local resource arrangement model.
In one embodiment, the local resource arrangement model may be updated by using a gradient lifting tree algorithm, so as to obtain an updated local resource arrangement model. And performing iterative computation according to the gradient lifting tree algorithm and the first encryption information, and constructing a regression tree corresponding to the local resource arrangement model. And when the regression tree meets the preset condition, acquiring the current intelligent resource arrangement model as the updated local resource arrangement model.
Specifically, the basic idea is as follows: given a data set X e R with n samples and d-dimensional featuresn×dConstructing K regressions by using XGboost algorithmTree to predict local resource orchestration model output:
Figure BDA0002816954570000151
wherein f isk(xi): the output value of each of the k regression trees,
Figure BDA0002816954570000152
and (5) integrating predicted values of k regression trees.
To learn the regression tree model in the above formula, XGboost adds a tree f in t iterationstTo minimize the following losses:
Figure BDA0002816954570000153
wherein,
Figure BDA0002816954570000154
wherein,
Figure BDA0002816954570000155
the predicted loss function of the regression tree,
gi、hi：
Figure BDA0002816954570000156
the first order partial derivative and the second order partial derivative,
ft: a modified regression tree for reducing calculation errors and minimizing prediction losses,
φ(t): the objective function of the t-th iteration is that the iteration process is to minimize phi(t)In the above-described manner, the process (a),
γ: a split threshold.
When the model builds the regression tree on the t-th iteration, starting from depth 0, one split (split) is added to one leaf node at a time until the tree reaches the maximum depth. Further, the best split is determined using the following form:
Figure BDA0002816954570000161
in the above equation, ILAnd IRThe node samples of all the classified subtrees are obtained. The classification that maximizes the split information is then selected as the optimal split. When the model obtains an optimal tree structure, the optimal weight of the leaf node j can be given according to the following formula:
Figure BDA0002816954570000162
wherein,
Figure BDA0002816954570000163
optimal weight value of leaf node with regression tree number j, IjIs the sample space of leaf j.
Because the splitting candidate set and the optimal leaf node weight only depend on giAnd hiTherefore, giAnd hiThe intermediate value is the above-mentioned intermediate value, and the second encryption information is obtained after homomorphic encryption is performed.
S203: and sending second encryption information of the updated local resource arrangement model to the controller, wherein the second encryption information is used for updating the global resource arrangement model of the controller.
And the computing cluster sends second encrypted information of the updated local resource arrangement model to the controller, wherein the second encrypted information is used for updating the global resource arrangement model of the controller. And the control equipment updates the global resource arrangement model according to the second encryption information sent by all the computer clusters to obtain a new global resource arrangement model.
In this embodiment, the computing cluster obtains first encryption information sent by the controller; updating the local resource arrangement model according to the first encryption information to obtain an updated local resource arrangement model; and sending second encryption information of the updated local resource arrangement model to the controller, wherein the second encryption information is used for updating the global resource arrangement model of the controller. According to the method, under the conditions that the number of users is large and the network condition is complex, the distributed computing clusters are independently computed by adopting a federal learning method, the computing results are gathered to the control equipment, a global resource arrangement model is obtained, the global resource arrangement model can accurately and quickly determine the resource arrangement strategy, the capability of the edge computing clusters is fully exerted, high-quality computing support is provided for delay sensitive application, resources are dynamically and intelligently arranged according to task characteristics, the average completion time of tasks is minimized, the number of the completed tasks is maximized, the problem in the aspect of computing efficiency is solved, and the resource utilization rate and the computing efficiency are improved.
Referring to fig. 3, fig. 3 is a schematic flowchart of a method for determining a resource orchestration policy according to a third embodiment of the present application. In this embodiment, an execution subject of the method for determining a resource orchestration policy is a device having a function of determining a resource orchestration policy, such as a desktop computer and a server. The method for determining the resource orchestration policy as shown in fig. 3 may include:
s301: and acquiring task information of the task to be distributed.
The device obtains task information of the tasks to be allocated, and the task information may include task related parameters, such as the number of tasks, the resource condition required by the tasks, the completion time limit of the tasks, the task completion requirement, and the like.
S302: inputting the task information into a preset target intelligent resource arrangement model for processing to obtain a resource arrangement strategy corresponding to the task information; wherein, the target intelligent resource arrangement model is obtained by the training method of the intelligent resource arrangement model of any one of claims 1 to 5.
The device stores the target intelligent resource arrangement model in advance, wherein the obtaining manner of the target intelligent resource arrangement model may refer to the training methods of the intelligent resource arrangement models in the first embodiment and the second embodiment, and details are not repeated here.
And inputting the task information into a preset target intelligent resource arrangement model for processing to obtain a resource arrangement strategy corresponding to the task information.
In this embodiment, the determining device of the resource arrangement policy obtains task information of a task to be allocated; and inputting the task information into a preset target intelligent resource arrangement model for processing to obtain a resource arrangement strategy corresponding to the task information. The target intelligent resource arrangement model used in the method is obtained by independently calculating the dispersed calculation clusters through a federal learning method, summarizing calculation results to the control equipment and summarizing calculation by the control equipment. The target intelligent resource arrangement model can accurately and quickly determine the resource arrangement strategy, and improves the resource utilization rate and the calculation efficiency.
Referring to fig. 4, fig. 4 is a schematic diagram of a control device according to a fourth embodiment of the present application. The units are included for performing the steps in the corresponding embodiment of fig. 1. Please refer to fig. 1 for the related description of the corresponding embodiment. For convenience of explanation, only the portions related to the present embodiment are shown. Referring to fig. 4, the control device 4 includes:
a first processing unit 410, configured to obtain first encryption information of a global resource arrangement model, and send the first encryption information to each computing cluster in a federal learning alliance, where the first encryption information is used to update a local resource arrangement model of each computing cluster;
a receiving unit 420, configured to receive second encryption information of the updated local resource orchestration model returned by each of the computing clusters;
the second processing unit 430 is configured to update the global resource arrangement model according to all the second encryption information, so as to obtain a new global resource arrangement model.
Further, the second processing unit 430 is specifically configured to:
decrypting the second encrypted information according to a preset decryption rule to obtain update information corresponding to each computing cluster;
and training the global resource arrangement model according to all the updating information to obtain a new global resource arrangement model.
Further, the control device 4 further includes:
an obtaining unit, configured to obtain a first sample training set; the sample training set comprises historical task information and resource arrangement strategy labels corresponding to the historical task information;
and the third processing unit is used for training the original resource arrangement model according to the first sample training set to obtain an initial global resource arrangement model.
Further, the sample training set further includes preset special case task information and a resource arrangement strategy label corresponding to the preset special case task information.
Further, the control device 4 further includes:
and the fourth processing unit is configured to send the new global resource arrangement model and/or third encryption information corresponding to the new global resource arrangement model to the computing cluster if a model information acquisition request sent by the computing cluster is received.
Referring to fig. 5, fig. 5 is a schematic diagram of a computing cluster according to a fifth embodiment of the present application. The units are included for performing the steps in the corresponding embodiment of fig. 2. Please refer to fig. 2 for a related description of the embodiment. For convenience of explanation, only the portions related to the present embodiment are shown. Referring to fig. 5, the computing cluster 5 includes:
a first obtaining unit 510, configured to obtain first encryption information sent by a control device;
an updating unit 520, configured to update the local resource arrangement model according to the first encryption information, to obtain an updated local resource arrangement model;
a sending unit 530, configured to send second encryption information of the updated local resource orchestration model to the controller, where the second encryption information is used to update a global resource orchestration model of the controller.
Further, the computing cluster 5 further includes:
the second acquisition unit is used for acquiring the original resource arrangement model and a second sample training set;
and the training unit is used for training the initial intelligent resource arrangement model according to the second sample training set to obtain an initial local resource arrangement model.
Further, the updating unit 520 is specifically configured to:
performing iterative computation according to a gradient lifting tree algorithm and the first encryption information, and constructing a regression tree corresponding to the local resource arrangement model;
and when the regression tree meets the preset condition, acquiring the current intelligent resource arrangement model as the updated local resource arrangement model.
Referring to fig. 6, fig. 6 is a schematic diagram of a device for determining a resource orchestration policy according to a sixth embodiment of the present application. The units are included for performing the steps in the corresponding embodiment of fig. 3. Please refer to the related description of the embodiment in fig. 3. For convenience of explanation, only the portions related to the present embodiment are shown. Referring to fig. 6, the determining means 6 of the resource orchestration policy includes:
an obtaining unit 610, configured to obtain task information of a task to be allocated;
the processing unit 620 is configured to input the task information into a preset target intelligent resource arrangement model for processing, so as to obtain a resource arrangement policy corresponding to the task information; wherein the target intelligent resource arrangement model is obtained by the training method of the intelligent resource arrangement model according to the first aspect.
Fig. 7 is a schematic diagram of a control device according to a seventh embodiment of the present application. As shown in fig. 7, the control device 7 of this embodiment includes: a processor 70, a memory 71 and a computer program 72, such as a training program for an intelligent resource orchestration model, stored in said memory 71 and executable on said processor 70. The processor 70, when executing the computer program 72, implements the steps in the above-described embodiments of the training method for intelligent resource orchestration models, such as the steps 101-103 shown in fig. 1. Alternatively, the processor 70, when executing the computer program 72, implements the functions of the modules/units in the above-mentioned device embodiments, such as the functions of the modules 410 to 430 shown in fig. 4.
Illustratively, the computer program 72 may be partitioned into one or more modules/units that are stored in the memory 71 and executed by the processor 70 to accomplish the present application. The one or more modules/units may be a series of computer program instruction segments capable of performing specific functions, which are used to describe the execution of the computer program 72 in the control device 7. For example, the computer program 72 may be divided into a first processing unit, a receiving unit, and a second processing unit, and the specific functions of each unit are as follows:
the system comprises a first processing unit, a second processing unit and a third processing unit, wherein the first processing unit is used for acquiring first encryption information of a global resource arrangement model and sending the first encryption information to each computing cluster in a Federal learning alliance, and the first encryption information is used for updating a local resource arrangement model of each computing cluster;
a receiving unit, configured to receive second encryption information of the updated local resource orchestration model returned by each computing cluster;
and the second processing unit is used for updating the global resource arrangement model according to all the second encryption information to obtain a new global resource arrangement model.
The control device may include, but is not limited to, a processor 70, a memory 71. It will be appreciated by those skilled in the art that fig. 7 is only an example of a control device 7 and does not constitute a limitation of the control device 7 and may comprise more or less components than shown, or some components may be combined, or different components, e.g. the control device may also comprise input output devices, network access devices, buses, etc.
The Processor 70 may be a Central Processing Unit (CPU), other general purpose Processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA) or other Programmable logic device, discrete Gate or transistor logic, discrete hardware components, etc. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
The memory 71 may be an internal storage unit of the control device 7, such as a hard disk or a memory of the control device 7. The memory 71 may also be an external storage device of the control device 7, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash memory Card (Flash Card), and the like provided on the control device 7. Further, the control device 7 may also include both an internal storage unit and an external storage device of the control device 7. The memory 71 is used for storing the computer program and other programs and data required by the control device. The memory 71 may also be used to temporarily store data that has been output or is to be output.
Fig. 8 is a schematic diagram of a computing cluster according to an eighth embodiment of the present application. As shown in fig. 8, the computing cluster 8 of this embodiment includes: a processor 80, a memory 81, and a computer program 82, such as a training program for an intelligent resource orchestration model, stored in the memory 81 and executable on the processor 80. The processor 80, when executing the computer program 82, implements the steps in the above-described embodiments of the training method for intelligent resource orchestration models, such as steps 201 to 203 shown in fig. 2. Alternatively, the processor 80, when executing the computer program 82, implements the functions of the modules/units in the above-described device embodiments, such as the functions of the modules 510 to 530 shown in fig. 5.
Illustratively, the computer program 82 may be partitioned into one or more modules/units that are stored in the memory 81 and executed by the processor 80 to accomplish the present application. The one or more modules/units may be a series of computer program instruction segments capable of performing certain functions, which are used to describe the execution of the computer program 82 in the computing cluster 8. For example, the computer program 82 may be divided into a first acquiring unit, an updating unit, a second processing unit, and a sending unit, and the specific functions of each unit are as follows:
a first acquisition unit configured to acquire first encryption information sent by a control device;
the updating unit is used for updating the local resource arrangement model according to the first encryption information to obtain an updated local resource arrangement model;
and the sending unit is used for sending the second encrypted information of the updated local resource arrangement model to the controller, and the second encrypted information is used for updating the global resource arrangement model of the controller.
The computing cluster may include, but is not limited to, a processor 80, a memory 81. Those skilled in the art will appreciate that fig. 8 is merely an example of a computing cluster 8 and does not constitute a limitation of computing cluster 8, and may include more or fewer components than shown, or some components in combination, or different components, e.g., the computing cluster may also include input-output devices, network access devices, buses, etc.
The Processor 80 may be a Central Processing Unit (CPU), other general purpose Processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), an off-the-shelf Programmable Gate Array (FPGA) or other Programmable logic device, discrete Gate or transistor logic, discrete hardware components, etc. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
The storage 81 may be an internal storage unit of the computing cluster 8, such as a hard disk or a memory of the computing cluster 8. The memory 81 may also be an external storage device of the computing cluster 8, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash memory Card (Flash Card), and the like, which are provided on the computing cluster 8. Further, the computing cluster 8 may also include both internal storage units and external storage devices of the computing cluster 8. The memory 81 is used for storing the computer programs and other programs and data required by the computing cluster. The memory 81 may also be used to temporarily store data that has been output or is to be output.
Fig. 9 is a schematic diagram of a resource orchestration policy determining device according to a ninth embodiment of the present application. As shown in fig. 9, the determining device 9 of the resource orchestration policy of this embodiment includes: a processor 90, a memory 91 and a computer program 92 stored in said memory 91 and executable on said processor 90, such as a determining program of a resource orchestration policy. The processor 90, when executing the computer program 92, implements the steps in the above-described embodiments of the method for determining a resource orchestration policy, such as the steps 301 to 302 shown in fig. 3. Alternatively, the processor 90, when executing the computer program 92, implements the functions of the modules/units in the above-mentioned device embodiments, such as the modules 610 to 620 shown in fig. 6.
Illustratively, the computer program 92 may be partitioned into one or more modules/units that are stored in the memory 91 and executed by the processor 90 to accomplish the present application. The one or more modules/units may be a series of computer program instruction segments capable of performing specific functions, which are used to describe the execution process of the computer program 92 in the determining device 9 of the resource orchestration policy. For example, the computer program 92 may be divided into a receiving unit, a first processing unit, a second processing unit, a third processing unit, a fourth processing unit, and an executing unit, and the specific functions of each unit are as follows:
the device comprises a receiving unit, a processing unit and a processing unit, wherein the receiving unit is used for receiving a timing request which comprises a timing duration and a processing task;
the first processing unit is used for acquiring first increasing time of a monotone increasing clock and determining first virtual alarm clock time according to the first increasing time and the timing duration;
the second processing unit is used for creating a virtual timer according to the first virtual alarm clock time and the processing task, and determining a system alarm clock task according to the first virtual alarm clock time;
the third processing unit is used for acquiring second increasing time of the monotone increasing clock when the system clock triggers a system alarm clock task;
the fourth processing unit is configured to acquire a virtual alarm clock time corresponding to each virtual timer, and determine the virtual timer with the virtual alarm clock time being less than the second incremental time as a target timer;
and the execution unit is used for executing the processing task corresponding to the target timer.
The determining device of the resource orchestration policy may include, but is not limited to, the processor 90 and the memory 91. It will be appreciated by those skilled in the art that fig. 9 is only an example of the determining device 9 of the resource orchestration policy, does not constitute a limitation to the determining device 9 of the resource orchestration policy, and may comprise more or less components than those shown, or some components in combination, or different components, for example, the determining device of the resource orchestration policy may further comprise an input-output device, a network access device, a bus, etc.
The Processor 90 may be a Central Processing Unit (CPU), other general purpose Processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA) or other Programmable logic device, discrete Gate or transistor logic, discrete hardware components, etc. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
The storage 91 may be an internal storage unit of the resource orchestration policy determining device 9, such as a hard disk or a memory of the resource orchestration policy determining device 9. The memory 91 may also be an external storage device of the determining device 9 of the resource arranging policy, for example, a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash memory Card (Flash Card), or the like, which is equipped on the determining device 9 of the resource arranging policy. Further, the determining device 9 of the resource orchestration policy may also include both an internal storage unit and an external storage device of the determining device 9 of the resource orchestration policy. The memory 91 is used for storing the computer program and other programs and data required by the determining device of the resource orchestration policy. The memory 91 may also be used to temporarily store data that has been output or is to be output.
It should be noted that, for the information interaction, execution process, and other contents between the above-mentioned devices/units, the specific functions and technical effects thereof are based on the same concept as those of the embodiment of the method of the present application, and specific reference may be made to the part of the embodiment of the method, which is not described herein again.
It will be apparent to those skilled in the art that, for convenience and brevity of description, only the above-mentioned division of the functional units and modules is illustrated, and in practical applications, the above-mentioned function distribution may be performed by different functional units and modules according to needs, that is, the internal structure of the apparatus is divided into different functional units or modules to perform all or part of the above-mentioned functions. Each functional unit and module in the embodiments may be integrated in one processing unit, or each unit may exist alone physically, or two or more units are integrated in one unit, and the integrated unit may be implemented in a form of hardware, or in a form of software functional unit. In addition, specific names of the functional units and modules are only for convenience of distinguishing from each other, and are not used for limiting the protection scope of the present application. For the specific working processes of the units and modules in the system, reference may be made to the corresponding processes in the foregoing method embodiments, which are not described herein again.
An embodiment of the present application further provides a timing device of a virtual timer, where the timing device of the virtual timer includes: at least one processor, a memory, and a computer program stored in the memory and executable on the at least one processor, the processor implementing the steps of any of the various method embodiments described above when executing the computer program.
The embodiments of the present application further provide a computer-readable storage medium, where a computer program is stored, and when the computer program is executed by a processor, the computer program implements the steps in the above-mentioned method embodiments.
The embodiments of the present application provide a computer program product, which when running on a mobile terminal, enables the mobile terminal to implement the steps in the above method embodiments when executed.
The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, all or part of the processes in the methods of the embodiments described above can be implemented by a computer program, which can be stored in a computer-readable storage medium and can implement the steps of the embodiments of the methods described above when the computer program is executed by a processor. Wherein the computer program comprises computer program code, which may be in the form of source code, object code, an executable file or some intermediate form, etc. The computer readable medium may include at least: any entity or device capable of carrying computer program code to a photographing apparatus/terminal apparatus, a recording medium, computer Memory, Read-Only Memory (ROM), Random Access Memory (RAM), an electrical carrier signal, a telecommunications signal, and a software distribution medium. Such as a usb-disk, a removable hard disk, a magnetic or optical disk, etc. In certain jurisdictions, computer-readable media may not be an electrical carrier signal or a telecommunications signal in accordance with legislative and patent practice.
In the above embodiments, the descriptions of the respective embodiments have respective emphasis, and reference may be made to the related descriptions of other embodiments for parts that are not described or illustrated in a certain embodiment.
Those of ordinary skill in the art will appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware or combinations of computer software and electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application.
In the embodiments provided in the present application, it should be understood that the disclosed apparatus/device and method may be implemented in other ways. For example, the above-described apparatus/device embodiments are merely illustrative, and for example, the division of the modules or units is only one logical division, and there may be other divisions when actually implemented, for example, a plurality of units or components may be combined or may be integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may be in an electrical, mechanical or other form.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
The above-mentioned embodiments are only used for illustrating the technical solutions of the present application, and not for limiting the same; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; such modifications and substitutions do not substantially depart from the spirit and scope of the embodiments of the present application and are intended to be included within the scope of the present application.

Claims (10)

1. A training method of an intelligent resource arrangement model is applied to a control device, and comprises the following steps:
acquiring first encryption information of a global resource arrangement model, and sending the first encryption information to each computing cluster in a Federal learning alliance, wherein the first encryption information is used for updating a local resource arrangement model of each computing cluster;
receiving second encryption information of the updated local resource orchestration model returned by each computing cluster;
and updating the global resource arrangement model according to all the second encryption information to obtain a new global resource arrangement model.
2. The method for training an intelligent resource arrangement model according to claim 1, wherein the updating the global resource arrangement model according to all the second encryption information to obtain a new global resource arrangement model comprises:
decrypting the second encrypted information according to a preset decryption rule to obtain update information corresponding to each computing cluster;
and training the global resource arrangement model according to all the updating information to obtain a new global resource arrangement model.
3. The method of training an intelligent resource orchestration model according to claim 1, wherein the method of training the initial global resource orchestration model comprises:
acquiring a first sample training set; the first sample training set comprises historical task information and corresponding resource arrangement strategy labels;
and training the original resource arrangement model according to the first sample training set to obtain an initial global resource arrangement model.
4. The training method of an intelligent resource orchestration model according to claim 3, wherein the sample training set further comprises preset special case task information and a resource orchestration strategy label corresponding thereto.
5. The method for training an intelligent resource arrangement model according to claim 1, wherein after the global resource arrangement model is updated according to all the second encryption information to obtain a new global resource arrangement model, the method further comprises:
and if a model information acquisition request sent by the computing cluster is received, sending the new global resource arrangement model and/or third encryption information corresponding to the new global resource arrangement model to the computing cluster.
6. A training method of an intelligent resource arrangement model is applied to a computing cluster, and comprises the following steps:
acquiring first encryption information sent by a controller;
updating the local resource arrangement model according to the first encryption information to obtain an updated local resource arrangement model;
and sending second encryption information of the updated local resource arrangement model to the controller, wherein the second encryption information is used for updating the global resource arrangement model of the controller.
7. The method of training an intelligent resource orchestration model according to claim 6, wherein the method of training the initial local resource orchestration model comprises:
acquiring an original resource arrangement model and a second sample training set;
and training the initial intelligent resource arrangement model according to the second sample training set to obtain an initial local resource arrangement model.
8. The method for training the intelligent resource arrangement model according to claim 6, wherein the updating the local resource arrangement model according to the first encryption information to obtain the updated local resource arrangement model includes:
performing iterative computation according to a gradient lifting tree algorithm and the first encryption information, and constructing a regression tree corresponding to the local resource arrangement model;
and when the regression tree meets the preset condition, acquiring the current intelligent resource arrangement model as the updated local resource arrangement model.
9. A method for determining a resource orchestration policy, comprising:
acquiring task information of a task to be distributed;
inputting the task information into a preset target intelligent resource arrangement model for processing to obtain a resource arrangement strategy corresponding to the task information; wherein the target intelligent resource arrangement model is obtained by the training method of the intelligent resource arrangement model of any one of claims 1 to 5.
10. A computer-readable storage medium, in which a computer program is stored which, when being executed by a processor, carries out the method according to any one of claims 1 to 5, 6 to 8, or 9.
CN202011401075.2A 2020-12-03 2020-12-03 Training method of intelligent resource arrangement model Pending CN114610475A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011401075.2A CN114610475A (en) 2020-12-03 2020-12-03 Training method of intelligent resource arrangement model

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011401075.2A CN114610475A (en) 2020-12-03 2020-12-03 Training method of intelligent resource arrangement model

Publications (1)

Publication Number Publication Date
CN114610475A true CN114610475A (en) 2022-06-10

Family

ID=81856583

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011401075.2A Pending CN114610475A (en) 2020-12-03 2020-12-03 Training method of intelligent resource arrangement model

Country Status (1)

Country Link
CN (1) CN114610475A (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115202908A (en) * 2022-09-09 2022-10-18 杭州海康威视数字技术股份有限公司 Privacy computation request response method and device based on dynamic arrangement
CN117314063A (en) * 2023-09-15 2023-12-29 国投曹妃甸港口有限公司 Emergency personnel coordination method and system in production process
CN117573382A (en) * 2024-01-17 2024-02-20 国网浙江省电力有限公司丽水供电公司 Data acquisition task arrangement method and device

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109165683A (en) * 2018-08-10 2019-01-08 深圳前海微众银行股份有限公司 Sample predictions method, apparatus and storage medium based on federation's training
CN109871702A (en) * 2019-02-18 2019-06-11 深圳前海微众银行股份有限公司 Federal model training method, system, equipment and computer readable storage medium
CN111008709A (en) * 2020-03-10 2020-04-14 支付宝(杭州)信息技术有限公司 Federal learning and data risk assessment method, device and system

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109165683A (en) * 2018-08-10 2019-01-08 深圳前海微众银行股份有限公司 Sample predictions method, apparatus and storage medium based on federation's training
CN109871702A (en) * 2019-02-18 2019-06-11 深圳前海微众银行股份有限公司 Federal model training method, system, equipment and computer readable storage medium
CN111008709A (en) * 2020-03-10 2020-04-14 支付宝(杭州)信息技术有限公司 Federal learning and data risk assessment method, device and system

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
KEWEI CHENG等: "SecureBoost: A Lossless Federated Learning Framework", pages 3 - 6, Retrieved from the Internet <URL:https://arxiv.org/pdf/1901.08755.pdf> *
张君如;赵晓焱;袁培燕;: "面向用户隐私保护的联邦安全树算法", 计算机应用, no. 10, 1 June 2020 (2020-06-01) *

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115202908A (en) * 2022-09-09 2022-10-18 杭州海康威视数字技术股份有限公司 Privacy computation request response method and device based on dynamic arrangement
CN115202908B (en) * 2022-09-09 2023-01-03 杭州海康威视数字技术股份有限公司 Privacy computation request response method and device based on dynamic arrangement
CN117314063A (en) * 2023-09-15 2023-12-29 国投曹妃甸港口有限公司 Emergency personnel coordination method and system in production process
CN117573382A (en) * 2024-01-17 2024-02-20 国网浙江省电力有限公司丽水供电公司 Data acquisition task arrangement method and device
CN117573382B (en) * 2024-01-17 2024-03-29 国网浙江省电力有限公司丽水供电公司 Data acquisition task arrangement method and device

Similar Documents

Publication Publication Date Title
US20230177349A1 (en) Federated learning optimizations
CN110084377B (en) Method and device for constructing decision tree
AU2021232839B2 (en) Updating Attribute Data Structures to Indicate Trends in Attribute Data Provided to Automated Modelling Systems
NL2030226B1 (en) Methods, systems, articles of manufacture and apparatus to optimize resources in edge networks
US20210097449A1 (en) Memory-efficient system for decision tree machine learning
Najafizadeh et al. Multi-objective Task Scheduling in cloud-fog computing using goal programming approach
CN114610475A (en) Training method of intelligent resource arrangement model
US20220261661A1 (en) Methods, systems, articles of manufacture and apparatus to improve job scheduling efficiency
JP2021517295A (en) High-efficiency convolutional network for recommender systems
CN110705719B (en) Method and apparatus for performing automatic machine learning
WO2022171066A1 (en) Task allocation method and apparatus based on internet-of-things device, and network training method and apparatus
JP2020533692A (en) Methods, systems, and computer programs for updating training data
US20200311539A1 (en) Cloud computing data compression for allreduce in deep learning
CN109788489A (en) A kind of base station planning method and device
US20220300418A1 (en) Maximizing resource bandwidth with efficient temporal arbitration
CN110460662A (en) The processing method and system of internet of things data
Moghaddasi et al. An enhanced asynchronous advantage actor-critic-based algorithm for performance optimization in mobile edge computing-enabled internet of vehicles networks
CN112215326A (en) Distributed AI system
Mangas et al. WoTemu: An emulation framework for edge computing architectures based on the Web of Things
Bantouna et al. Network load predictions based on big data and the utilization of self-organizing maps
Marin et al. Reaching for the clouds: contextually enhancing smartphones for energy efficiency
CN114679283A (en) Block chain data request processing method and device, server and storage medium
US11410023B2 (en) Lexicographic deep reinforcement learning using state constraints and conditional policies
CN115002215B (en) Cloud government enterprise oriented resource allocation model training method and resource allocation method
US10417228B2 (en) Apparatus and method for analytical optimization through computational pushdown

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination