CN109947551B - Multi-turn task allocation method, edge computing system and storage medium thereof - Google Patents
Multi-turn task allocation method, edge computing system and storage medium thereof Download PDFInfo
- Publication number
- CN109947551B CN109947551B CN201910207642.1A CN201910207642A CN109947551B CN 109947551 B CN109947551 B CN 109947551B CN 201910207642 A CN201910207642 A CN 201910207642A CN 109947551 B CN109947551 B CN 109947551B
- Authority
- CN
- China
- Prior art keywords
- task
- computing
- time
- distribution
- module
- 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.)
- Active
Links
Images
Abstract
The invention discloses a multi-round task allocation method, an edge computing system and a storage medium thereof, wherein the method comprises the steps of sending a task and decomposing a single task into a plurality of subtasks; on the basis of multi-turn distribution, considering system overhead and task completion time, providing a joint optimization model for multi-stage optimization to obtain the actual distribution task size of each turn of each terminal in each stage; the distribution task is distributed to each terminal node in turn, and a plurality of stages are repeatedly carried out until the tasks are completely finished.
Description
Technical Field
The invention belongs to the field of task allocation methods, and particularly relates to a multi-turn task allocation method, an edge computing system and a storage medium thereof.
Background
In recent years, cloud computing has become a research hotspot in the industrial and academic circles, and the disadvantages of high network transmission delay, excessive data volume to be processed and the like have attracted attention. With the push of cloud services and the internet of things, we assume that edge networks are changing from data consumers to data producers and data consumers. A large amount of data does not need to be transmitted to the cloud end for processing, and is distributed to terminals close to the edge of the network to complete tasks. Therefore, time delay can be reduced, the method is well suitable for real-time scenes such as current hot virtual reality and image recognition, and pressure caused by big data processing is effectively relieved.
Taking a video analysis task as an example, mobile phone camera shooting and network camera shooting are widely applied to the life of people at present, and when people need to find lost children and old people or track suspect people, cloud computing is not suitable any more. Because the video file is large and the privacy is protected, the video file capturing the target person can be transmitted to a plurality of terminal users nearby, the target person can be found as soon as possible, and therefore the long transmission time and the privacy leakage required for transmitting the video file to the cloud are avoided.
How to allocate tasks to available terminals nearby for efficient processing becomes a key problem in edge computing research, and at present, edge computing task allocation research points mainly aim at task offloading, and researchers propose corresponding offloading strategies from different index angles to perform task allocation.
Disclosure of Invention
Aiming at the technical problems in the related art, the invention provides a multi-turn task allocation method, an edge computing system and a storage medium thereof, which are suitable for a scheduling strategy of a video analysis task in a high-dynamic mobile terminal cooperation environment and are used for solving the problems of long task completion time and high cost in a traditional model.
In the present invention, we mainly consider the following three main features of mobile devices, including:
high mobility, the mobile device's residence time in a particular physical area (e.g., a coffee shop) is uncertain and not regularly recyclable.
High real-time, time is money, especially for emergency tasks (such as finding lost children).
The heterogeneity of devices, the capabilities and trustworthiness of each device before the task starts, is unknown, and the computing power and status may also change dynamically during task execution.
In order to achieve the technical purpose, the technical scheme of the invention is realized as follows:
a multi-turn task allocation method is characterized by comprising the following steps:
s1: sending the tasks, and decomposing a single task into a plurality of subtasks;
s2: on the basis of multi-turn distribution, considering system overhead and task completion time, providing a joint optimization model to obtain the actual distribution task size of each turn of each terminal;
s3: and distributing the distribution tasks to each terminal node in turn.
Preferably, step S2 specifically includes:
s2.1, acquiring available terminal node information to form a cooperation group;
s2.2 when all terminal nodes in the cooperation group are calculated to be the optimal calculation capacity, the task amount delta is completed1=NvRequired time T1I.e. the shortest possible time;
s2.3 predicting the next T by using an ARIMA time series model according to historical computing power and overhead information1The computing power and computing overhead of the available terminals vary over time;
s2.4, substituting the predicted computing capacity and computing cost into an objective function, and solving the problem at T by using CPLEX software1An allocation scheme obtained for minimizing the objective function over time;
s2.5 if in the first stage T1The number of tasks actually performed N within a time periodv1<NvThen the next phase is continued. The remaining task completion amount Δ is calculated using the same method2=Nv-Nv1Desired time T2;
S.2.6 similarly, a phase l optimization is carried out until Δl=0。
Preferably, step S2.2 is specifically: obtaining the computing capacity b of each terminal node by utilizing ARIMA time series model predictionjtAnd computing overhead cjtWherein the computing power b of each of said available terminalsjtAnd computing overhead cjtAre heterogeneous and dynamically changing over time.
Preferably, the step S2.4 is specifically: then, the acquired historical data of each terminal node is formed into a new setSolving the optimal solution by using CPLEX software to enable an objective function
min E(Δk,Tk)=C(Δk,Tk)+β*L(Δk,Tk)
And the minimum is in the k stage, wherein beta is a weight, and the ratio of system overhead to completion time is adjusted.
Compared with the prior art, the invention is suitable for the scheduling strategy of the video analysis task in the high-dynamic mobile terminal cooperation environment, is used for solving the problems of long task completion time and high cost in the traditional model,
drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without creative efforts.
FIG. 1 is a diagram of a Cloudlet-based task distributor of the present invention;
FIG. 2 is a flow chart of the task assignment algorithm of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments that can be derived by one of ordinary skill in the art from the embodiments given herein are intended to be within the scope of the present invention.
Considering the characteristics of high mobility, high real-time performance, equipment heterogeneity and the like of a mobile terminal in the system, a multi-turn task allocation scheme based on prediction is designed, when the amount of a single calculation task is too large, the single task is decomposed into a plurality of subtasks, the subtasks are respectively sent to nearby available mobile equipment in turns, and finally, the result is only required to be returned to the Cloudlet.
On the basis of multi-round distribution, a joint optimization model is provided by considering system overhead and task completion time, and the method specifically comprises the following steps:
(1) meterWhen all the terminal nodes in the group are calculated to have the optimal computing capacity (the historical completed task number is maximum), the completed task quantity NvRequired time T1I.e. the shortest possible time.
(2) And (3) predicting the computing capacity and computing overhead change of the terminal in the next period of time by using an ARIMA time series model according to the historical computing capacity (CPU availability) and the overhead information (computing overhead) of the terminal node participating in the task cooperation.
Each turn is separated by 5 minutes and the more computationally powerful the terminal requires more computational overhead.
(3) Substituting the predicted computing power and computing cost into the objective function, calculated at T1The amount of tasks completed within a period of time delta1=NvThe distribution scheme obtained by minimizing the objective function utilizes CPLEX software to solve the optimal solution, and the objective function is as follows:
min E(Δ1,T1)=C(Δ1,T1)+β*L(Δ1,T1)
satisfy in that
(4) Each turn comprises a plurality of turns if in the first stage T1The number of tasks actually performed N within a time periodv1<NvThen the next phase is continued. The remaining task completion amount Δ is calculated using the same method2=Nv-Nv1Desired time T2Substituting into the objective function of the 2 nd stage as shown below;
min E(Δ2,T2)=C(Δ2,T2)+β*L(Δ2,T2)
(5) similarly, a phase optimization is performed until Δl=0。
With reference to fig. 1, the present invention is based on Cloudlet task distribution, and there are three modules in the system, namely a scheduling module, a prediction module and a distribution module;
the system time is discretized into a plurality of time slots or rounds. For a task JmAnd the method is completed by a plurality of available terminal nodes j in a coordinated mode. At each time slot t, each terminal node has computing power bjt(CPU utilization) and overhead cjt。
Wherein the system overhead calculation formula of the k stage isThe completion time of the k-th stage is calculated by the formulaThe size of the residual task amount is used to represent the length of the calculation time, and when the task amount allocated at the stage is larger, the residual tasks are fewer, the completion time may be shorter, and the calculation overhead is larger at this time. The idea of a trade-off between computational overhead and completion time is compatible.
The specific flow of the present invention is shown in fig. 2, firstly, when a task requester sends a request to a nearby Cloudlet, and after the Cloudlet receives the request, a scheduling module in the Cloudlet broadcasts a message to peripheral terminal nodes to organize NsEach available terminal node forms a cooperation group, and the size of the task completion amount is NvThe task of (2).
Here, we assume that the available terminal stays in proximity of the Cloulet for a certain period of time and the Cloudlet can acquire historical computing power information of all terminal nodes, and the completion time of the total task is counted as the number of turns as
Subsequently, when all the terminal nodes in the cooperation group are calculated to be the optimal computing capacity, the task amount delta is completed1=NvRequired time T1I.e. the shortest possible time. Predicting the next T by utilizing an ARIMA time series model according to historical computing power and overhead information1The calculation capacity and calculation overhead change of the available terminal in time are respectively recorded asAnd
substituting the predicted computing capacity and computing cost into an objective function, and solving at T by using CPLEX software1Distributing the calculated task amount to each terminal node for a distribution scheme obtained by minimizing the objective function within time;
then if in the first stage T1The number of tasks actually performed N within a time periodv1<NvThen the next phase is continued. The remaining task completion amount Δ is calculated using the same method2=Nv-Nv1Desired time T2;
Finally, the same procedure is used for the phase l optimization up to Δl=0。
The invention can dynamically adjust the weight in the objective function to meet the requirements of task requesters on different time and expenses. And the objective function is minimized by the algorithm and is close to the god algorithm.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.
Claims (6)
1. A multi-turn task allocation method is characterized by comprising the following steps:
s1: sending the tasks, and decomposing a single task into a plurality of subtasks;
s2: on the basis of multi-turn distribution, considering system overhead and task completion time, providing a joint optimization model to obtain the actual distribution task size of each turn of each terminal;
s3: distributing the distribution tasks to each terminal node in turn;
the step S2 specifically includes:
s2.1, acquiring available terminal node information to form a cooperation group;
s2.2 when all terminal nodes in the cooperation group are calculated to be the optimal calculation capacity, the task amount delta is completed1=NvRequired time T1I.e. the shortest possible time;
s2.3 predicting the next T by using an ARIMA time series model according to historical computing power and overhead information1The computing power and computing overhead of the available terminals vary over time;
s2.4, substituting the predicted computing capacity and computing cost into an objective function, and solving the problem at T by using CPLEX software1An allocation scheme obtained for minimizing the objective function over time;
s2.5 if in the first stage T1The number of tasks actually performed N within a time periodv1<NvThen continue the next stage; the remaining task completion amount Δ is calculated using the same method2=Nv-Nv1Desired time T2;
S.2.6 phase optimization Using the same method until Δl=0。
2. A method for assigning tasks for multiple rounds as claimed in claim 1, wherein the step S2.2 is specifically: obtaining the computing capacity b of each terminal node by utilizing ARIMA time series model predictionjtAnd computing overhead cjtWherein the computing power b of each of said available terminalsjtAnd computing overhead cjtAre heterogeneous and dynamically changing over time.
3. A method for assigning multiple rounds of tasks according to claim 1 or 2, characterized in that said historical computing power is specifically referred to CPU availability.
4. A method for distributing multiple rounds of tasks according to claim 1 or 2, wherein the step S2.4 is specifically: then, the acquired historical data of each terminal node is formed into a new setSolving the optimal solution by using CPLEX software to enable an objective function minE (delta)k,Tk)=C(Δk,Tk)+β*L(Δk,Tk) And the minimum is in the k stage, wherein beta is a weight, and the ratio of system overhead to completion time is adjusted.
5. An edge computing system, comprising:
the edge requester queue module is used for sending a request and receiving a result to the Cloudlet module;
a Cloudlet module for task distribution using the multi-round task distribution method of any one of claims 1 to 4;
the edge node is used for receiving the subtasks according to the distribution result and feeding back information to the Cloudlet module;
the Cloudlet module comprises a scheduling module, a prediction module and a distribution module, wherein the scheduling module is used for broadcasting information to peripheral terminal nodes after receiving a request sent by a task requester and organizing at least one available terminal node to form a cooperation group; the prediction module predicts the next period of time T by utilizing an ARIMA model according to the historical computing capacity and the historical computing cost of each available terminal nodekThe calculation capacity and the calculation overhead of the internal terminal nodes are changed, and an optimal distribution scheme is obtained; and the distribution module distributes the data to each terminal node in turn according to the optimal distribution scheme, and the stage I is repeatedly carried out until the task is completed.
6. A storage medium having stored therein instructions executable by a processor, the storage medium comprising: the processor-executable instructions, when executed by a processor, are for performing the multi-round task assignment method of any one of claims 1-4.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910207642.1A CN109947551B (en) | 2019-03-19 | 2019-03-19 | Multi-turn task allocation method, edge computing system and storage medium thereof |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910207642.1A CN109947551B (en) | 2019-03-19 | 2019-03-19 | Multi-turn task allocation method, edge computing system and storage medium thereof |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109947551A CN109947551A (en) | 2019-06-28 |
CN109947551B true CN109947551B (en) | 2021-04-23 |
Family
ID=67009088
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910207642.1A Active CN109947551B (en) | 2019-03-19 | 2019-03-19 | Multi-turn task allocation method, edge computing system and storage medium thereof |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109947551B (en) |
Families Citing this family (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110650194A (en) * | 2019-09-23 | 2020-01-03 | 中国科学技术大学 | Task execution method based on edge calculation in computer network |
CN112596892B (en) * | 2020-11-23 | 2021-08-31 | 中标慧安信息技术股份有限公司 | Data interaction method and system of multi-node edge computing equipment |
CN112598112B (en) * | 2020-12-04 | 2021-09-10 | 深圳大学 | Resource scheduling method based on graph neural network |
CN112835691A (en) * | 2021-01-05 | 2021-05-25 | 华翔翔能科技股份有限公司 | Edge data processing method and system for communication of Internet of things |
CN115277031A (en) * | 2021-04-13 | 2022-11-01 | 华为技术有限公司 | Data processing method and device |
CN113407322B (en) * | 2021-06-21 | 2022-05-06 | 平安国际智慧城市科技股份有限公司 | Multi-terminal task allocation method and device, electronic equipment and readable storage medium |
CN113626107B (en) * | 2021-08-20 | 2024-03-26 | 中南大学 | Mobile computing unloading method, system and storage medium |
Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102592247A (en) * | 2010-12-15 | 2012-07-18 | 索尼公司 | Task management system, task management method, and program |
CN104618506A (en) * | 2015-02-24 | 2015-05-13 | 庄奇东 | Crowd-sourced content delivery network system, method and device |
CN106815081A (en) * | 2017-01-23 | 2017-06-09 | 武汉万般上品信息技术有限公司 | A kind of based on mobile cloud computing and have the multi-task planning mechanism of budgetary restraints |
CN108509276A (en) * | 2018-03-30 | 2018-09-07 | 南京工业大学 | A kind of video task dynamic migration method in edge calculations environment |
CN108521446A (en) * | 2018-03-20 | 2018-09-11 | 深圳大学 | Dispatching method, device, equipment and the storage medium of cloud computing resources |
CN108804227A (en) * | 2018-05-23 | 2018-11-13 | 大连理工大学 | The method of the unloading of computation-intensive task and best resource configuration based on mobile cloud computing |
CN108845886A (en) * | 2018-07-27 | 2018-11-20 | 中南大学 | Cloud computing energy consumption optimization method and system based on phase space |
CN109358953A (en) * | 2018-09-20 | 2019-02-19 | 中南大学 | A kind of multitask application discharging method in thin cloud |
Family Cites Families (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US9418035B2 (en) * | 2012-10-22 | 2016-08-16 | Intel Corporation | High performance interconnect physical layer |
CN104754063B (en) * | 2015-04-17 | 2018-02-16 | 北京邮电大学 | Local cloud computing resource scheduling method |
US10146592B2 (en) * | 2015-09-18 | 2018-12-04 | Salesforce.Com, Inc. | Managing resource allocation in a stream processing framework |
US10346206B2 (en) * | 2016-08-27 | 2019-07-09 | International Business Machines Corporation | System, method and computer program product for resource management in a distributed computation system |
CN107133091A (en) * | 2017-05-08 | 2017-09-05 | 武汉轻工大学 | The cloud workflow task dispatching method being classified based on top-down task |
CN108132837B (en) * | 2018-01-02 | 2022-04-15 | 中国工商银行股份有限公司 | Distributed cluster scheduling system and method |
-
2019
- 2019-03-19 CN CN201910207642.1A patent/CN109947551B/en active Active
Patent Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102592247A (en) * | 2010-12-15 | 2012-07-18 | 索尼公司 | Task management system, task management method, and program |
CN104618506A (en) * | 2015-02-24 | 2015-05-13 | 庄奇东 | Crowd-sourced content delivery network system, method and device |
CN106815081A (en) * | 2017-01-23 | 2017-06-09 | 武汉万般上品信息技术有限公司 | A kind of based on mobile cloud computing and have the multi-task planning mechanism of budgetary restraints |
CN108521446A (en) * | 2018-03-20 | 2018-09-11 | 深圳大学 | Dispatching method, device, equipment and the storage medium of cloud computing resources |
CN108509276A (en) * | 2018-03-30 | 2018-09-07 | 南京工业大学 | A kind of video task dynamic migration method in edge calculations environment |
CN108804227A (en) * | 2018-05-23 | 2018-11-13 | 大连理工大学 | The method of the unloading of computation-intensive task and best resource configuration based on mobile cloud computing |
CN108845886A (en) * | 2018-07-27 | 2018-11-20 | 中南大学 | Cloud computing energy consumption optimization method and system based on phase space |
CN109358953A (en) * | 2018-09-20 | 2019-02-19 | 中南大学 | A kind of multitask application discharging method in thin cloud |
Non-Patent Citations (1)
Title |
---|
"移动云计算中任务卸载调度策略的研究";郭琪;《中国优秀硕士学位论文全文数据库 信息科技辑》;20180215(第02期);全文 * |
Also Published As
Publication number | Publication date |
---|---|
CN109947551A (en) | 2019-06-28 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109947551B (en) | Multi-turn task allocation method, edge computing system and storage medium thereof | |
Long et al. | Edge computing framework for cooperative video processing in multimedia IoT systems | |
CN110647391B (en) | Edge computing method and system for satellite-ground cooperative network | |
CN109669768B (en) | Resource allocation and task scheduling method for edge cloud combined architecture | |
Harvey et al. | Edos: Edge assisted offloading system for mobile devices | |
CN113784373B (en) | Combined optimization method and system for time delay and frequency spectrum occupation in cloud edge cooperative network | |
CN113992945B (en) | Multi-server multi-user video analysis task unloading method based on game theory | |
WO2016145137A1 (en) | Scaling the lte control plane for future mobile access | |
CN113992678A (en) | Calculation migration method for offshore MEC load balancing and resource allocation joint optimization | |
CN115103450A (en) | Multi-service time slot distribution method and equipment | |
CN113132497B (en) | Load balancing and scheduling method for mobile edge operation | |
CN109495565B (en) | High-concurrency service request processing method and device based on distributed ubiquitous computing | |
CN113315669B (en) | Cloud edge cooperation-based throughput optimization machine learning inference task deployment method | |
CN113220459B (en) | Task processing method and device | |
CN113747507A (en) | Computing resource management method and device for 5G ultra-dense network | |
CN112817753A (en) | Task processing method and device, storage medium and electronic device | |
Happ et al. | On the impact of clustering for IoT analytics and message broker placement across cloud and edge | |
CN110971707A (en) | Distributed service caching method in mobile edge network | |
CN116233007A (en) | Heterogeneous mobile edge computing load balancing oriented hierarchical distributed matching method | |
Chaturvedi et al. | A Novel method of Improving Spectrum sensing Management system for CR-IoT Networks | |
CN114610503A (en) | Multi-application online computing unloading resource scheduling method and system based on overall resource requirements of task graph in edge computing | |
CN114866790A (en) | Live streaming scheduling method and device | |
Han et al. | Opportunistic coded distributed computing: An evolutionary game approach | |
CN109889598B (en) | Self-organizing computing platform based on mobile node and task scheduling method thereof | |
CN101997720A (en) | Transregional monitoring flow on demand method and video management server |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |