CN113162965B - Low-delay Map and Reduce joint scheduling method for heterogeneous MapReduce cluster - Google Patents
Low-delay Map and Reduce joint scheduling method for heterogeneous MapReduce cluster Download PDFInfo
- Publication number
- CN113162965B CN113162965B CN202110020740.1A CN202110020740A CN113162965B CN 113162965 B CN113162965 B CN 113162965B CN 202110020740 A CN202110020740 A CN 202110020740A CN 113162965 B CN113162965 B CN 113162965B
- Authority
- CN
- China
- Prior art keywords
- reduce
- map
- servers
- time delay
- task allocation
- 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
- 238000000034 method Methods 0.000 title claims abstract description 39
- 238000004364 calculation method Methods 0.000 claims abstract description 33
- 230000001934 delay Effects 0.000 claims abstract description 9
- 239000013598 vector Substances 0.000 claims abstract description 6
- 238000004891 communication Methods 0.000 claims description 16
- 230000005540 biological transmission Effects 0.000 claims description 12
- 230000009286 beneficial effect Effects 0.000 abstract description 2
- 238000004422 calculation algorithm Methods 0.000 description 15
- 238000011161 development Methods 0.000 description 5
- 238000005457 optimization Methods 0.000 description 5
- 230000007423 decrease Effects 0.000 description 3
- 238000005265 energy consumption Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000012804 iterative process Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000005094 computer simulation Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000010295 mobile communication Methods 0.000 description 1
- 238000005192 partition Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/10—Protocols in which an application is distributed across nodes in the network
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/10—Protocols in which an application is distributed across nodes in the network
- H04L67/1001—Protocols in which an application is distributed across nodes in the network for accessing one among a plurality of replicated servers
- H04L67/1004—Server selection for load balancing
- H04L67/1008—Server selection for load balancing based on parameters of servers, e.g. available memory or workload
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/10—Protocols in which an application is distributed across nodes in the network
- H04L67/1001—Protocols in which an application is distributed across nodes in the network for accessing one among a plurality of replicated servers
- H04L67/1004—Server selection for load balancing
- H04L67/101—Server selection for load balancing based on network conditions
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Computer Hardware Design (AREA)
- General Engineering & Computer Science (AREA)
- Multi Processors (AREA)
Abstract
The invention discloses a heterogeneous MapReduce cluster-oriented low-delay Map and Reduce combined scheduling method, which comprises the following steps: the client submits the MapReduce operation to the heterogeneous cluster, and the control center distributes the operation to each server and completes distributed computation so as to minimize the total time delay of operation execution. The method comprises the steps that calculation task allocation exists in a Map stage and a Reduce stage of a MapReduce frame, according to the principle that all servers can finish calculation at the same time, the task allocation at one position is fixed to optimize the other position, iterative solution is conducted on two task allocation vectors until execution time delays of all the servers are equal, the optimal calculation task allocation strategies of the two stages are obtained, and the total time delay of operation completion is minimum. The method can be used for MapReduce distributed computation under heterogeneous clusters to obtain the beneficial effect of low time delay.
Description
Technical Field
The invention relates to the field of wireless communication, in particular to a heterogeneous MapReduce cluster-oriented low-delay Map and Reduce joint scheduling method.
Background
With the popularization of mobile equipment and the development of mobile internet, the development trend of internet of everything is presented in the future, the 5G serving as a new generation mobile communication technology really realizes the conversion of communication from human to everything in the aspect of network technology, and especially the advantages of the 5G communication in the aspects of low time delay, large network bandwidth and the like can radically promote the development of the core fields of the internet of things such as artificial intelligence and block chains. Considering that most data services in the internet of things have the characteristics of large data volume, complex types and the like, and the resource shortage of a core network, the requirement of low time delay makes it unrealistic for users to intensively process oversized computing tasks, and the development of distributed computing in the field of wireless communication is greatly promoted. The MapReduce is used as a distributed computing framework specially designed for large-scale parallel data processing, is very suitable for a plurality of edge servers to process computing tasks of various types in a distributed manner, and helps a user terminal to efficiently solve computing problems.
The execution flow of the MapReduce framework is summarized as follows: after the client submits the operation, the control center in the main node divides the operation into servers to perform Map operation and generates a certain amount of intermediate values, then the intermediate values need to be communicated with other servers for transmission, and finally each server performs Reduce operation on the collected intermediate value data and outputs the intermediate value data to the HDFS.
In order to meet the basic requirements of large-scale computing tasks on time delay, people research the problem of resource scheduling in a MapReduce framework under different conditions, and a good resource scheduling strategy can effectively reduce time delay cost. Specifically, the optimization objectives can be classified into load optimization, energy consumption optimization, time delay optimization, and the like. The method can be divided into single-task and multi-task distributed computing according to the number of submitted jobs. Most research works are developed based on MapReduce in a homogeneous environment, however, with the development of a communication network, a heterogeneous cluster environment is closer to an actual situation, a traditional average distribution mechanism is not applicable any more, and the problems of load imbalance, resource waste, high time delay and the like are caused to a great extent.
Disclosure of Invention
In order to overcome the defects of the prior art, the invention aims to minimize the delay tailing of a single server according to the principle that all servers can complete calculation at the same time, so that the total delay for completing calculation tasks is minimized, and further a heterogeneous MapReduce cluster-oriented low-delay Map and Reduce combined scheduling method is designed.
The low-delay Map and Reduce joint scheduling method for the heterogeneous MapReduce cluster comprises the following steps:
acquiring information such as the size of an executed job, the number of cluster servers, the transmission rate of communication among the servers, the clock period of unit bit data calculated by the servers, the CPU frequency, the number of working slots of Map and Reduce, the data volume output by unit Map tasks and the like;
obtaining system time delay of each server serving as a Reduce computing node to complete Reduce computing according to the information, and defining the problem of minimizing total time delay of job completion;
according to the minimization of the delay tailing of a single server, the parallel time between the servers is improved, and the optimality condition which meets the condition that all the servers can complete calculation at the same time is obtained:
according to the principle that the system time delays of all servers for completing Reduce calculation are equal, fixing one task distribution variable to be unchanged, and carrying out iterative solution on the other task distribution variable:
and (3) fixing a Map stage task allocation variable m, updating a Reduce stage task allocation variable q:
normalizing the task allocation variables in the Reduce stage at the current moment according to the equality constraint:
fixing the updated Reduce stage task allocation variable q to update the Map stage task allocation variable m:
normalizing the task allocation variables in the Map stage at the current moment according to the equality constraint:
and alternately solving two task allocation variables by iteration based on the method, so that the total time delay of the operation is monotonously reduced, and the iteration is stopped until the difference between the two time delays is within the threshold range, thereby obtaining the optimal task allocation vector.
The invention has the beneficial effects that:
according to the invention, by minimizing the trailing time delay of a single server and improving the parallel time between servers, a low-time-delay Map and Reduce combined scheduling method for a heterogeneous MapReduce cluster is designed, the computing capacity of different servers in the cluster and the transmission rate between the servers are considered, the task allocation before the Map stage and the task allocation at the Reduce stage are optimized, and the total time delay of job completion is minimized.
Drawings
FIG. 1 is a system model of a heterogeneous MapReduce cluster-oriented low-latency Map and Reduce joint scheduling method.
Fig. 2 is a graph comparing the performance of the method of the present invention with four comparison algorithms in terms of latency for different numbers of edge servers.
FIG. 3 is a graph comparing the performance of the method of the present invention with four comparison algorithms in terms of latency for different job sizes.
Fig. 4 is a graph comparing the performance of the method of the present invention in terms of time delay with four comparison algorithms, with different average calculation frequencies, respectively, fixed at 100 and 200.
Fig. 5 is a graph comparing the performance of the method of the present invention with the four comparison algorithms in terms of delay for different average transmission rates, respectively, fixed at 100 and 300.
FIG. 6 shows the cluster server count at 30 and the threshold at 10 -3 Under the condition of (1), the method of the invention is used for a time delay relationship graph in an iteration process.
Detailed Description
The invention is further illustrated below with reference to the figures and examples.
A system model adopting a heterogeneous MapReduce cluster-oriented low-delay Map and Reduce joint scheduling method is shown in FIG. 1, and a user distributes a calculation task to servers in a cluster according to a certain task distribution rule in Map and Reduce phases. According to the method, the current optimal value of one task distribution variable is obtained in an iterative mode under the condition that the trailing time delay of a single server is minimized and one task distribution variable is fixed, the current optimal value of the other task distribution variable is obtained in an alternate mode until the difference between the time delay of the two times before and after the task distribution variable is stopped within the threshold range, and therefore the optimal task distribution strategies of the total time delay of the operation completion and the Map stage and the Reduce stage are obtained.
The low-delay Map and Reduce joint scheduling method for the heterogeneous MapReduce cluster comprises the following steps:
step 1: acquiring information such as the size of an executed job, the number of cluster servers, the transmission rate of communication among the servers, the clock period of unit bit data calculated by the servers, the CPU frequency, the number of working slots of Map and Reduce, the data volume output by unit Map tasks and the like;
step 2: calculating to obtain system time delay which is calculated when each server is used as a Reduce calculation node according to the information, and defining the problem of minimizing the total time delay when the operation is finished;
and 3, step 3: according to the time delay tailing of the minimized single server, the parallel time between the servers is improved, and the optimality condition that all the servers can complete Reduce calculation at the same time is obtained;
and 4, step 4: and according to the optimality condition met by the total delay tau after the operation is finished, fixing one task distribution variable to be unchanged, and simultaneously carrying out iterative solution on the other task distribution variable, wherein the two task distribution variables are alternately iterated to make the total delay after the operation is finished monotonously reduced until the difference between the two time delays is within the threshold value range, and the iteration is stopped.
As a better implementation example, first, information such as the size of a job submitted by a client, the number of cluster servers, the calculation frequency of each server, and the transmission rate of communication between servers is obtained, and each server can obtain corresponding communication and calculation delay according to the following calculation method:
wherein,
and
respectively the calculation delay of the server l in the Map stage and the calculation delay of the server k in the Reduce stage.
And collecting partition data transmitted by each node except the server k for the communication delay of the server k in the Shuffle stage.
The calculation time delay of the Map stage and the communication time delay of the Shuffle stage are overlapped, for a server k serving as a Reduce calculation node, the Reduce operation can be executed only when the maps of all servers and their own Shuffle are finished, and the system time delay t corresponding to the calculation is finished k (m, q) is:
comparing the system time delay of all the servers for completing Reduce calculation, taking the maximum value as the total time delay tau for completing the operation, and meeting the requirement
And then the problem of minimizing the total time delay of job completion is established:
minτ
at the same time, the problem needs to satisfy the following constraints: all the intermediate values generated by the unit Map task are distributed to each server; the size of the tasks allocated on each server is non-negative. The constraints of the optimization problem are established using the following formula:
m l ,q k ≥0
based on this, the objective function can be further converted into:
wherein
The system time delay is the system time delay when all the servers finish Reduce calculation when the server l is used as a Map calculation node.
In order to minimize the total time delay of the completion of the operation, the basic idea of improving the parallel time of the servers according to the time delay tailing of the minimized single server is obtained, and the optimality conditions met by the total time delay of the completion of the operation are as follows:
according to the principle that the system time delays of all servers for completing Reduce calculation are equal, fixing one task distribution variable to be unchanged, and carrying out iterative solution on the other task distribution variable until the current is optimal:
and (3) fixing a Map stage task allocation variable m, updating a Reduce stage task allocation variable q:
normalizing the task allocation variables in the Reduce stage at the current moment according to the equality constraint:
fixing the updated Reduce stage task allocation variable q, updating the Map stage task allocation variable m:
normalizing the task allocation variables at the Map stage at the current moment according to equality constraint:
based on the method, the two task allocation variables are alternately and iteratively solved, so that the total time delay of the completion of the operation is monotonically reduced, and the iteration is stopped until the difference between the time delays of the previous time and the next time is within the threshold range, thereby obtaining the optimal task allocation vector.
As a better implementation example, a job with a size D of 1GB needs to be submitted to the MapReduce framework for calculation, the number N of default servers in the cluster is 30, the total number of working slots of each server is 4, that is, any server k in the cluster satisfies the requirement
In addition, assume that the clock period, CPU frequency, and transmission rate of the channel at which the server calculates the unit number of bits are respectively obeyed [200,600]、[1,2]、[10,30]The output data amount gamma of the unit Map task is 200, and each group of data is averaged by running 1000 times of computer simulation to be used as a final result.
Four comparison methods are introduced, wherein the comparison algorithm 1 is a traditional resource scheduling method in which the server performance in the default cluster is the same, and the data allocation in the Map stage and the Reduce stage adopts average allocation; the comparison algorithm 2 is a resource scheduling method for distributing Map tasks for the servers according to the computing performance of the servers in the cluster; the comparison algorithm 3 is a resource scheduling method for dividing Reduce tasks according to the calculation performance of each server in the cluster; the comparison algorithm 4 is a resource scheduling method for jointly optimizing the task allocation of each server in the Map stage and the power allocation of each server with the aim of minimizing the total energy consumption for executing the tasks. The invention designs a low-delay resource scheduling method by jointly optimizing task allocation variables in Map and Reduce stages.
Firstly, the size of the submitted job is fixed to be 1GB, and the number of servers in the cluster is changed to observe the performance of each algorithm in the time delay performance. Fig. 2 shows the relationship that the total time delay of job completion of the five algorithms changes within the range of [30,100] along with the number of servers in the cluster, and it can be seen that the method of the present invention has an obvious reduction in the total time delay compared with the four comparison algorithms, and as the number of servers in the cluster increases, the total time delay of job completion gradually decreases, and the performance advantage of the method of the present invention is more significant.
Then, the number of the servers in the cluster is fixed to be 30, and the size of the submitted job is changed to observe the performance of each algorithm in the time delay performance. Fig. 3 shows the influence of different sizes of calculation tasks on the time delay in five algorithms, the calculation tasks are increased from 1GB to 10GB, and it is seen from the figure that the increase of the job task amount can significantly increase the time delay, and the size of the calculation tasks is the most direct factor affecting the time delay under the condition of keeping other parameters unchanged.
And then, respectively changing the average calculation frequency and the average transmission rate of the cluster server under different gamma coefficients to observe the performance of each algorithm in the time delay performance. Fig. 4 shows that as the expected value of the CPU clock frequency distribution increases, each algorithm shows a trend of time delay decrease, the average value of the CPU clock frequency represents the overall computing capability of the cluster, the increase of the CPU clock frequency decreases the computation time delay of Map and Reduce stages, and the smaller the γ value is, the smaller the total data amount of the communication is represented to a certain extent, and the smaller the total time delay is. Fig. 5 shows that when the value γ is smaller, the total job completion delay is first reduced with the increase of the mean transmission rate, and then remains unchanged, because the increase of the transmission rate to a certain extent reduces the communication delay in the Shuffle stage to a level that does not participate in the calculation, and this further improves the communication capability between the servers in the cluster without reducing the total task execution delay. And when the value of gamma is larger, the communication time delay at the Shuffle stage is higher, and at the moment, increasing the transmission rate between the servers within a certain range inevitably reduces the total time delay of task execution. Likewise, the method of the present invention achieves the lowest latency compared to other methods.
Finally, FIG. 6 shows that at a cluster server number of 30, a threshold of 10 -3 Under the condition of time delay relationship in the iterative process, the obvious step represents that two task distribution variables are alternately solved in the iterative process, and the optimal total time delay of operation completion can be quickly converged.
By combining the performance comparison, the resource scheduling method provided by the invention for jointly optimizing the task allocation of Map and Reduce is superior to other comparison methods in time delay performance, and has more outstanding advantages in large-scale networks.
Claims (1)
1. The low-delay Map and Reduce joint scheduling method for the heterogeneous MapReduce cluster is characterized by comprising the following steps of:
1) the following information is acquired: submitted job size D (bits), number N of servers in cluster, and transmission rate v of communication between servers in network lk (Mbps) and the server calculates the clock period mu of the unit bit data k (cycles/bit), CPU frequency
Number of working grooves of Map and Reduce tasks
And the data quantity gamma output by the unit Map task, wherein l and k are the serial numbers of the servers;
2) minimizing the delay tailing of a single server, and improving the parallel time of the servers to obtain task allocation vectors m and q at Map and Reduce stages;
the method obtains the optimal total time delay for completing the operation and the optimal task allocation vector, and comprises the following steps:
when any one server k is obtained as a Reduce computing node according to the acquired information, the system time delay t corresponding to the Reduce computing is completed k (m，q)；
Wherein, beta is the time delay ratio of the Map task and the Reduce task of the server computing unit, m l And q is k Respectively distributing tasks of a server l in a Map stage and a server k in a Reduce stage, comparing system time delays of each server for completing Reduce calculation, and taking the maximum value as a total operation completion time delay tau;
obtaining another equivalent representation of total job completion time delay tau based on the above
Wherein
When a server l is used as a Map computing node, all servers finish the system time delay of Reduce computing;
minimizing the delay tailing of a single server, improving the parallel time of the servers, and obtaining the total operation completion delay tau to meet the optimality condition that all the servers can complete Reduce calculation at the same time:
t k (m * ，q * )＝s l (m * ，q * )，
l∈{1，...，N}
according to the principle that the system time delays of all servers for completing Reduce calculation are equal, fixing one task distribution variable to be unchanged, and carrying out iterative solution on the other task distribution variable until the current is optimal:
fixing a Map stage task allocation variable m, updating a Reduce stage task allocation variable q:
normalizing the task allocation variables in the Reduce stage at the current moment according to the equality constraint:
fixing the updated Reduce stage task allocation variable q, updating the Map stage task allocation variable m:
normalizing the task allocation variables at the Map stage at the current moment according to equality constraint:
and alternately solving two task allocation variables by iteration based on the method, so that the total time delay of the operation is monotonously reduced, and the iteration is stopped until the difference between the two time delays is within the threshold range, thereby obtaining the optimal task allocation vector.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110020740.1A CN113162965B (en) | 2021-01-07 | 2021-01-07 | Low-delay Map and Reduce joint scheduling method for heterogeneous MapReduce cluster |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110020740.1A CN113162965B (en) | 2021-01-07 | 2021-01-07 | Low-delay Map and Reduce joint scheduling method for heterogeneous MapReduce cluster |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113162965A CN113162965A (en) | 2021-07-23 |
| CN113162965B true CN113162965B (en) | 2022-09-20 |
Family
ID=76878426
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110020740.1A Active CN113162965B (en) | 2021-01-07 | 2021-01-07 | Low-delay Map and Reduce joint scheduling method for heterogeneous MapReduce cluster |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113162965B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114844781B (en) * | 2022-05-20 | 2023-05-09 | 南京大学 | Method and system for optimizing Shuffle performance for encoding MapReduce under Rack architecture |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102567086A (en) * | 2010-12-30 | 2012-07-11 | 中国移动通信集团公司 | Task scheduling method, equipment and system |
| CN103631657A (en) * | 2013-11-19 | 2014-03-12 | 浪潮电子信息产业股份有限公司 | Task scheduling algorithm based on MapReduce |
| CN111381950A (en) * | 2020-03-05 | 2020-07-07 | 南京大学 | Task scheduling method and system based on multiple copies for edge computing environment |
| US10880366B1 (en) * | 2018-07-30 | 2020-12-29 | Splunk Inc. | Enabling stateless status checks using collector redirection |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8751639B2 (en) * | 2011-04-27 | 2014-06-10 | Rackspace Us, Inc. | Event queuing and distribution system |
-
2021
- 2021-01-07 CN CN202110020740.1A patent/CN113162965B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102567086A (en) * | 2010-12-30 | 2012-07-11 | 中国移动通信集团公司 | Task scheduling method, equipment and system |
| CN103631657A (en) * | 2013-11-19 | 2014-03-12 | 浪潮电子信息产业股份有限公司 | Task scheduling algorithm based on MapReduce |
| US10880366B1 (en) * | 2018-07-30 | 2020-12-29 | Splunk Inc. | Enabling stateless status checks using collector redirection |
| CN111381950A (en) * | 2020-03-05 | 2020-07-07 | 南京大学 | Task scheduling method and system based on multiple copies for edge computing environment |
Non-Patent Citations (2)
| Title |
|---|
| Venkata Swamy Martha ; Weizhong Zhao.h-MapReduce: A Framework for Workload Balancing in MapReduce.《IEEE》.2013,全文. * |
| 异构Hadoop集群下的负载自适应反馈调度策略;潘佳艺;《计算机工程与科学》;20170330;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113162965A (en) | 2021-07-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112286677B (en) | Resource-constrained edge cloud-oriented Internet of things application optimization deployment method | |
| CN111475274B (en) | Cloud collaborative multi-task scheduling method and device | |
| CN112118287B (en) | Network resource optimization scheduling decision method based on alternative direction multiplier algorithm and mobile edge calculation | |
| CN111722910B (en) | Cloud job scheduling and resource allocation method | |
| CN111813506B (en) | Resource perception calculation migration method, device and medium based on particle swarm optimization | |
| CN111163143B (en) | Low-delay task unloading method for mobile edge calculation | |
| CN110351754A (en) | Industry internet machinery equipment user data based on Q-learning calculates unloading decision-making technique | |
| CN112788605B (en) | Edge computing resource scheduling method and system based on double-delay depth certainty strategy | |
| CN112650581A (en) | Cloud-side cooperative task scheduling method for intelligent building | |
| CN112214301B (en) | Smart city-oriented dynamic calculation migration method and device based on user preference | |
| CN113918240A (en) | Task unloading method and device | |
| Zhu et al. | Computing offloading strategy using improved genetic algorithm in mobile edge computing system | |
| CN111694664A (en) | Calculation unloading distribution method of edge server | |
| CN113992677A (en) | MEC calculation unloading method for delay and energy consumption joint optimization | |
| Wu et al. | A mobile edge computing-based applications execution framework for Internet of Vehicles | |
| CN115396953A (en) | Calculation unloading method based on improved particle swarm optimization algorithm in mobile edge calculation | |
| CN113162965B (en) | Low-delay Map and Reduce joint scheduling method for heterogeneous MapReduce cluster | |
| CN114925852A (en) | Federal learning training method and system based on model segmentation and resource allocation | |
| CN113573363A (en) | MEC calculation unloading and resource allocation method based on deep reinforcement learning | |
| CN112596910A (en) | Cloud computing resource scheduling method in multi-user MEC system | |
| CN114064294B (en) | Dynamic resource allocation method and system in mobile edge computing environment | |
| CN110266757A (en) | A kind of edge calculations method for optimizing resources of oriented mission collaboration scene | |
| CN113473542A (en) | Time delay minimization resource allocation method and device for wireless energy supply edge computing network | |
| CN117579701A (en) | Mobile edge network computing and unloading method and system | |
| Yao et al. | Multi-objective cooperative computation offloading for MEC in UAVs hybrid networks via integrated optimization framework |
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 |