CN113992678A - Calculation migration method for offshore MEC load balancing and resource allocation joint optimization - Google Patents
Calculation migration method for offshore MEC load balancing and resource allocation joint optimization Download PDFInfo
- Publication number
- CN113992678A CN113992678A CN202111257708.1A CN202111257708A CN113992678A CN 113992678 A CN113992678 A CN 113992678A CN 202111257708 A CN202111257708 A CN 202111257708A CN 113992678 A CN113992678 A CN 113992678A
- Authority
- CN
- China
- Prior art keywords
- server
- task
- calculation
- tasks
- matching
- 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.)
- Withdrawn
Links
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
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Computer Hardware Design (AREA)
- General Engineering & Computer Science (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a calculation migration method for combined optimization of marine MEC load balancing and resource allocation, which aims to solve the problem of marine energy limitation and ensure that performance optimization is realized under the condition of stable network, so that the energy consumption, time delay and service quality are balanced. The first stage of the method is unloading decision, which ensures whether the task is unloaded to the edge server and the server capable of unloading in the time delay allowable range; and the second stage is resource allocation, and the task of selecting unloading adopts a multiple maximum matching algorithm based on the idea of Gale-Shapley to allocate computing resources. Simulation experiment results show that the method can obtain the optimal unloading decision and resource allocation results, eliminate the optimization effect when the task selects the server under the condition of meeting the task time delay, avoid the server from being overloaded, and improve the energy consumption efficiency and the overall performance of the offshore edge computing system.
Description
Technical Field
The invention relates to the technical field of offshore edge computing, in particular to a computing migration method for offshore MEC load balancing and resource allocation joint optimization.
Background
In the last decade, mobile intelligent terminals such as mobile phones, pads and wearable devices have gradually become important tools for mass study, entertainment, travel, participation in social networks and understanding news due to portability and convenience. The rapidly developed information and communication technology for the Internet of things assists industrial and agricultural production, and the development of the digital industry is deeply promoted. The popularization of the 5G network makes users increasingly demand low-delay, low-power consumption and high-reliability services. However, resources such as processing capability, battery capacity, and storage capacity of the mobile terminal are limited, and real-time response requirements cannot be met. Meanwhile, the rapid development of the intelligent Internet also creates new opportunities for maritime activities. Due to uncertainty of marine environment, network base station construction is difficult, and the development of marine communication needs to be strengthened. In recent years, with the increasing frequency of maritime activities and the rapid development of maritime economy, the development of new-generation maritime communication systems has become a focus of academic and industrial interest.
At present, the basic coverage mode of the global marine communication network is mainly a communication network which is constructed by heterogeneous structures such as shore-based mobile communication, maritime wireless communication, satellite communication and underwater acoustic communication. In order to solve the problems of incompatible communication systems, inconsistent communication bandwidth, blind areas in coverage, lack of efficient and uniform management mechanism, and the like in the communication networks, a Mobile Edge Computing (MEC) offloading technology becomes an effective scheme for solving the problems. The main idea is that by deploying computing nodes near a data source, the network transmission delay and energy consumption overhead are reduced by resource sharing among the nodes, data are prevented from leaving an edge network, and data security and privacy protection are improved. Therefore, the introduction of edge computing offload into marine communication networks has three advantages: (1) temporary data are processed at the edge of a wireless access network near a user, so that the calculation speed is increased, and the energy loss of the network is reduced; (2) the data processing can be performed more timely when the terminal device is close to the terminal device, so that the quick response capability of the service is enhanced; (3) private data do not need to be uploaded, and the edge nodes transmit the data in an encrypted mode, so that the risk of leakage of network data is reduced. However, the introduction of edge computing and computational offloading techniques into the ocean network has been challenging:
(1) the energy supply is insufficient. The power supply modes of terminal equipment, servers and the like of the internet in a marine scene mainly comprise batteries and cables, however, the coverage range of the cables is limited, the batteries of the marine equipment are difficult to replace, and the energy consumption of nodes can cause the great fluctuation of the whole network.
(2) The resource allocation is not uniform. The deployment of computing resources of edge servers in the ocean presents 'long tail distribution', the servers with strong capacity execute tasks frequently, and the servers with weak capacity are idle, so that the stability of the network is influenced.
Disclosure of Invention
Aiming at the offshore edge computing environment, (1) task unloading is divided into two stages, wherein the first stage is unloading decision, and the computing task makes a decision according to self attributes, server load and the like; and the second stage is resource allocation, marks servers meeting task requirements, converts the servers into multiple matching problems of bipartite graphs, and solves the multiple matching problems through a multiple maximum matching algorithm based on the idea of Gale-Shapley algorithm to balance the load of the edge network. (2) And the task execution time delay is taken as a constraint, the task processing energy consumption and the calculation cost are comprehensively evaluated, the calculation resources are reasonably distributed, and the communication quality and efficiency of the network during calculation unloading are ensured.
The technical scheme for realizing the purpose of the invention is as follows:
a calculation migration method for combined optimization of offshore MEC load balancing and resource allocation improves energy consumption efficiency and overall performance of a system by solving an optimal unloading decision and a resource allocation scheme;
in the two-stage calculation migration method, the marine network composition constructed by the model mainly comprises 3 parts, wherein the first part is a core cloud network layer and comprises a large number of high-performance cloud servers for providing services for users; the second part is an edge server layer, all edge servers are divided into regions according to relative distances, each region consists of an unmanned aerial vehicle, an airship, a marine vernier and a land base station, and the unmanned aerial vehicle, the airship, the marine vernier and the land base station all have servers with moderate and heterogeneous performances; the third part is a terminal equipment layer, each area is composed of yachts, steamships, aircraft carriers and the like, and the yachts, the steamships, the aircraft carriers and the like have equipment with certain processing capacity or make unloading decision on the current calculation task according to the property of the task;
in the two-stage calculation migration method, task unloading is divided into two stages, the first stage is unloading decision, and a calculation task makes a decision according to self attributes, server load and the like; the second stage is resource allocation, marks servers meeting task requirements, converts the servers into multiple matching problems of bipartite graphs, and solves the multiple matching problems through a multiple maximum matching algorithm based on the idea of Gale-Shapley algorithm to balance edge network loads;
in the two-stage calculation migration method, task execution time delay is taken as constraint, channel resources and calculation cost are comprehensively evaluated, channels are reasonably distributed, waste of the channel resources is reduced, and communication quality and efficiency of a network during calculation unloading are guaranteed;
the calculation migration method for the offshore MEC load balancing and resource allocation combined optimization comprises the following steps:
1) the terminal node generates a calculation task and records the task attribute into a temporary array;
2) traversing all servers in sequence according to the attributes of each computing task, and storing energy consumption and task execution time delay of the computing task under each server into a temporary array;
3) taking task execution time delay as a constraint, judging whether the computing tasks need to be unloaded, and recording the server which can be unloaded by each computing task and the required energy consumption;
4) constructing a server dynamic price model according to the initial calculation cost, the unit calculation cost, the current calculation capacity and the number of processed tasks of the server; constructing a comprehensive evaluation model of the calculation tasks according to the energy consumption and the task execution delay of the calculation tasks under each server;
5) and searching the optimal matching of the calculation tasks selected to be unloaded in the optional server to convert into a multiple matching problem of bipartite graphs, and solving the multiple matching problem, wherein the optimal matching is finally obtained according to the idea of a Gale-Shapley algorithm in the graph theory in combination with the energy consumption, the time delay and the current price of the calculation tasks temporarily stored in the array.
In the step 4), the server constructs a server dynamic price model according to the initial calculation cost, the unit calculation cost, the current calculation capacity and the number of processed tasks; the method comprises the following steps that a calculation task comprehensive evaluation model is constructed by a calculation task according to energy consumption and task execution time delay generated under each server, and specifically comprises the following steps:
4-1) setting initial price Cost of server0And Price per unit of computing resourcej；
4-2) constructing a dynamic price model of the server, and judging whether the current server is in a high load or idle state:
in the ocean, the computing resource deployment of the edge server is easy to generate 'long tail distribution', the server with strong capability executes tasks frequently, the server with weak capability is idle, the stability of the network is influenced, and the dynamic price model of the server is constructed as follows:
among them, Cost0An initial cost for the server; pricejCalculating the price of the resource for the server unit; ratio (R)jA rate of use of computing resources for a current server; surplusjComputing resources left for the current server;
total computing resources for the current server; miCalculating the task quantity of the task; n is a task set processed by the server;
4-3) according to energy consumption and task execution time delay of the calculation task under each server, constructing a comprehensive evaluation model of the calculation task as follows:
s.t.C1:αij∈{0,1},λ+μ＝1
C2:Pi≤Pi,max
where C1 is the unload decision, αijIs a binary variable; c2 is the constraint of the terminal transmission power; c3 is the maximum delay requirement of the terminal; c4 is the constraint of terminal and server power; c5 is the restriction that the terminal can only select one server to send tasks; c6 is the maximum number of simultaneous online processing tasks of the server;
respectively representing the time delay and the energy consumption after the normalization processing.
Step 5), searching optimal matching of the calculation tasks selected to be unloaded in an optional server, converting the optimal matching into a multiple matching problem of bipartite graphs, and solving the multiple matching problem, wherein the optimal matching is finally obtained according to the idea of a Gale-Shapley algorithm in a graph theory and the energy consumption, time delay and current price of the calculation tasks temporarily stored in an array; the method specifically comprises the following steps:
5-1) traversing the calculation tasks and the optional edges thereof in sequence, matching the task with the server if only one optional edge of one task is connected with the server, and calculating and updating the current cost of the server according to the formula (1);
5-2) improving the thought of sequential point selection in sequence, matching task nodes from less to more according to the connectivity in sequence, selecting the edge with the lowest cost as the current matching by using the thought of a greedy algorithm in an approximate algorithm, calculating and updating the current cost of the server according to a formula (1), and storing the current cost into a temporary array;
5-3) matching the remaining nodes, selecting the edge with the least cost as the current matching, if the number of tasks received by the selected server reaches a threshold value, generating conflict and needing coordination, reserving the task with the connectivity degree of 1 in the tasks matched by the server, canceling the matching with the largest cost in the remaining tasks, and enabling the computing task to cancel the optional edge and try to re-match. And if the conflict still exists, executing the step 5-1) to the step 5-2) to recurse in sequence, and when the optional edges of the task do not meet the matching condition, giving up the task and judging that the task processing fails.
The calculation migration method for the combined optimization of the load balancing of the marine MEC and the resource allocation can realize the network load balancing and fully utilize the calculation resources, thereby avoiding the downtime of a server caused by overload and improving the system throughput and the network stability. Through simulation, the method can obtain the optimal unloading decision and resource allocation result, eliminate the optimization effect when the task selects the server under the condition of meeting the task time delay, avoid the overload of the server, and improve the energy consumption efficiency and the overall performance of the offshore edge computing system.
Drawings
FIG. 1 is a model of a marine edge computing network;
FIG. 2 is a flow chart of the method of the present invention;
FIG. 3 is a graph comparing server throughput under different algorithms;
FIG. 4 is a graph comparing success rates of task offloading of servers under different algorithms;
FIG. 5 is a comparison graph of server load ratios under different algorithms;
FIG. 6 is a comparison graph of the number of processing tasks of the server under different algorithms;
FIG. 7 is a comparison of total server energy consumption for different algorithms;
FIG. 8 is a comparison graph of energy efficiency ratios of servers under different algorithms;
FIG. 9 is a graph of the comprehensive analysis of the performance of the method of the present invention.
Detailed Description
The invention will be further described with reference to the following drawings and examples, but the invention is not limited thereto;
example (b):
assuming a marine edge computing scenario, where a marine vessel is used as an edge terminal, a base station, an unmanned aerial vehicle, an airship and a floating platform are used as edge servers, as shown in fig. 1, an MEC computing device is assumed to have n ship terminals U ═ {1,2,3, …, n } and m edge servers E ═ {1,2,3, …, m }, each edge server is surrounded by k orthogonal sub-channels C ═ {1,2,3, …, k }, data needs a certain time and energy consumption for transmission through a channel, T time slots are denoted as T ═ {1,2,3, …, T }, where the length of each time slot is τ;
suppose that:
1) the ship terminal generates 0-20 unequal computing tasks in a unit time slot, and the computing tasks can be split into a plurality of independent subtasks, namely zetaiThe task generated by the ith terminal is represented, whether the task can be unloaded is taken as a judgment basis, the task is divided and represented as
In order to be the non-unloadable portion,
is a removable part. Computing task composed of
The five attributes are respectively: calculating the size of a task (input data quantity), the number of CPU cycles required by the task, the transmission bandwidth of the task, the maximum delay of the task and the processing condition of the task;
2) edge server includes
The four attributes are respectively: the method comprises the steps of firstly, setting a server initial price, a server current price, a server CPU capacity and a server receiving task threshold;
1. constructing a marine edge computing network model, as shown in fig. 1, initializing the system as follows:
1) the first part is a terminal equipment layer, computing tasks are randomly generated in each area by equipment with certain processing capacity, such as yachts, ships, aircraft carriers and the like, and unloading decisions can be made on the current computing tasks according to task properties. Among tasks generated by the ship terminal, time delay sensitive tasks such as emergency risk avoidance and the like and tasks such as navigation calculation and the like which have high requirements on equipment calculation capacity exist. However, the computing and storage capacities of the ship terminal equipment are limited, tasks with large task quantity and low delay sensitivity can be unloaded to the edge server for processing, but tasks with high delay requirements and small required computing power can be processed locally;
2) the second part is an edge server layer, all edge servers are divided into regions according to relative distances, each region consists of an unmanned aerial vehicle, an airship, a marine vernier and a land base station, and the server has moderate performance and is heterogeneous;
3) the third part is a core cloud network layer and comprises a large number of high-performance cloud servers for providing services for users.
2. A computing migration method for offshore MEC load balancing and resource allocation joint optimization, as shown in fig. 2, specifically includes the following steps:
1) selecting unloaded computing tasks to enter an unloading waiting queue; selecting a locally processed computing task to enter a local processing queue;
2) sequentially dequeuing the calculation tasks in the unloading waiting queue, sequentially traversing the edge servers, recording the optional servers in a temporary array according to each calculation task, and simultaneously recording temporary energy consumption and time delay in the array;
3) taking task execution time delay as a constraint, judging whether the computing tasks need to be unloaded, and recording the server which can be unloaded by each computing task and the required energy consumption;
4) constructing a server dynamic price model according to the initial calculation cost, the unit calculation cost, the current calculation capacity and the number of processed tasks of the server; and constructing a comprehensive evaluation model of the calculation tasks according to the energy consumption and the task execution delay of the calculation tasks under each server. The method specifically comprises the following steps:
4-1) setting initial price Cost of server0And Price per unit of computing resourcej；
4-2) constructing a dynamic price model of the server, and judging whether the current server is in a high load or idle state:
in the ocean, the computing resource deployment of the edge server is easy to generate 'long tail distribution', the server with strong capability executes tasks frequently, the server with weak capability is idle, the stability of the network is influenced, and the dynamic price model of the server is constructed as follows:
among them, Cost0An initial cost for the server; pricejCalculating the price of the resource for the server unit; ratio (R)jA rate of use of computing resources for a current server; surplusjComputing resources left for the current server;
total computing resources for the current server; miCalculating the task quantity of the task; n is a task set processed by the server;
4-3) according to energy consumption and task execution time delay of the calculation task under each server, constructing a comprehensive evaluation model of the calculation task as follows:
s.t.C1:αij∈{0,1},λ+μ＝1
C2:Pi≤Pi,max
where C1 is the unload decision, αijIs a binary variable; c2 is the constraint of the terminal transmission power; c3 is the maximum delay requirement of the terminal; c4 is the constraint of terminal and server power; c5 is the restriction that the terminal can only select one server to send tasks; c6 is the maximum number of simultaneous online processing tasks of the server;
respectively representing the time delay and the energy consumption after normalization processing;
5) and searching the optimal matching in the optional server for the calculation task selected to be unloaded, and converting the optimal matching into a multiple matching problem of the bipartite graph to solve. And finally obtaining optimal matching according to the idea of the Gale-Shapley algorithm in the graph theory by combining the energy consumption and the time delay of the computing task temporarily stored in the array and the current price of the server. The method comprises the following steps:
5-1) traversing the calculation tasks and the optional edges thereof in sequence, and matching the task with the server if only one optional edge of one task is connected with the server. Calculating and updating the current cost of the server according to the formula (1);
and 5-2) improving the thought of sequential point selection in sequence, matching task nodes from less to more according to the connectivity in sequence, and selecting the edge with the lowest cost as the current matching by utilizing the thought of a greedy algorithm in an approximation algorithm. Meanwhile, the current cost of the server is calculated and updated according to the formula (1), and the current cost is stored in a temporary array;
5-3) matching the remaining nodes, selecting the edge with the least cost as the current matching, if the number of tasks received by the selected server reaches a threshold value, generating conflict and needing coordination, reserving the task with the connectivity degree of 1 in the tasks matched by the server, canceling the matching with the largest cost in the remaining tasks, and enabling the computing task to cancel the optional edge and try to re-match. If a conflict still occurs, steps 5-1) to 5-2) are performed in a recursive sequence. When the optional edges of the task do not meet the matching requirement, the task is abandoned, and the task processing failure is judged;
6) to verify the validity of the model, 2 controls were set up in the experiment:
modulo two plus random walk algorithm (Mod 2-RE): when a task migration request is initiated, use A ═ a1,a2,…,am]Indicating that the server receives the decision to process the task, A0For the initial decision, A1To randomly generate a binary matrix, the update formula for A is as follows:
wherein the content of the first and second substances,
method for representing modulo two addition
If a is [0,0,1,0,0,1,0,1], it means that the task can select server 3, 6, 8, and then select the three servers
Selecting an optimal server from the servers for unloading;
greedy algorithm based on CPU power (Greedy-CPU): during task migration, greedy selection of servers with high calculation capacity for processing each task within a time allowed range;
wherein the experimental simulation parameters are shown in the table 1;
3. analysis of Experimental Effect
1) Server throughput contrastive analysis under different algorithms
FIG. 3 compares the performance of the method in terms of server throughput under the GS-Match algorithm, the Mod2-RE algorithm and the Greedy-CPU algorithm. As shown in fig. 3, when the number of tasks, the task attribute, the server attribute, and the like are the same, the Mod2-RE algorithm is unstable due to high randomness, the number of tasks allocated to a single server is in between, and the total number of tasks processed by the server in the time slot is 34; the Greedy-CPU algorithm shows a trend of optimization, the servers carry out task allocation from strong computing to weak computing, the number of tasks allocated by a single server is 41, and the total number of tasks processed by the servers in the time slot is; the number of tasks distributed by a single server under the GS-Match algorithm is between, and the total number of tasks processed by the server in a unit time slot is 44; therefore, the GS-Match algorithm can improve the throughput of the system;
2) server task unloading success rate comparison analysis under different algorithms
FIG. 4 compares the performance of the method in terms of task unloading success rate under GS-Match algorithm, Mod2-RE algorithm and Greedy-CPU algorithm. Under 10 time slots, the success rates of the Mod2-RE algorithm and the Greeny-CPU algorithm are totally equal; the success rate of the GS-Match algorithm is generally improved by 18 percent and 14 percent respectively, so that the success rate of task unloading can be effectively improved;
3) server load contrast analysis under different algorithms
FIG. 5 compares the performance of the method in terms of server load under GS-Match algorithm, Mod2-RE algorithm and Greedy-CPU algorithm. As can be seen in FIG. 7, in 10 time slots, the server load degree under the Greedy-CPU algorithm is 2.0-3.0, the server load degree under the Mod2-RE algorithm is 1.0-2.0, the server load degree under the GS-Match algorithm is 0.4-0.5, the variance is respectively reduced by 3 times and 6 times, and the load degrees among the servers can be effectively balanced.
In conclusion, the GS-Match algorithm model can greatly improve the load balance degree and the task success rate of the server, eliminate the server overload caused by the optimization phenomenon and realize the full utilization of computing resources.
4) Server processing task number comparison analysis under different algorithms
FIG. 6 compares the performance of the method in terms of the number of processing tasks of the server under the GS-Match algorithm, the Mod2-RE algorithm and the Greedy-CPU algorithm in 10 time slots. As seen in FIG. 6, the number of tasks successfully processed by the server based on the GS-Match algorithm in 10 time slots is higher than that of the Mod2-RE algorithm and the Greeny-CPU algorithm. The reason is that the GS-Match algorithm reasonably distributes the unloading tasks according to the formula (1), thereby effectively avoiding the waste of computing resources.
5) Server energy efficiency ratio comparative analysis under different algorithms
Fig. 7 and fig. 8 compare the performance of the method of the invention in terms of energy consumption efficiency of the server under the GS-Match algorithm, the Mod2-RE algorithm and the Greedy-CPU algorithm. As shown in fig. 7, the total server energy consumption of the three algorithms increases linearly with the increase of tasks, and the energy consumption is nearly the same. As can be seen from FIG. 8, the energy consumption efficiency of the server based on the GS-Match algorithm is superior to that of the Mod2-RE algorithm and the Greedy-CPU algorithm and is stabilized at about 0.5, the Mod2-RE algorithm and the Greedy-CPU algorithm fluctuate between 0.3 and 0.5, and the energy efficiency ratio is respectively improved by 15% and 12%.
In conclusion, the unloading decision of the GS-Match algorithm can process more calculation tasks under the condition of the same energy consumption, the resources of the server are fully utilized, and the overload of the server and the waste of the calculation resources are avoided.
6) Comprehensive analysis of the Performance of the method of the invention
Fig. 9 and table 2 randomly send task requests by the terminal in different observation periods, and evaluate the comprehensive performance of the constructed model in a task interaction manner. The method mainly analyzes the task acceptance rate and rejection rate and the success rate and failure rate of processing in 15 time slots under the marine environment. The average acceptance rate of the tasks in the whole network is 92.10%, the rejection rate is 7.90%, the success rate in the process of accepting the tasks is 96.00%, and the failure rate is 4.00%.
The migration method of the invention can realize network load balance and fully utilize computing resources, thereby avoiding the downtime of the server caused by overload and improving the system throughput and the network stability.
To summarize:
aiming at the problems of network stability and limited computing resources of a low-delay and high-reliability marine communication network, the article discusses a joint optimization problem of migration decision and resource allocation. Considering the balance degree of the server and the number of tasks capable of processing more, an optimization problem with the total energy consumption and the total cost of an optimization system as a target is provided, and in the first stage, the calculation task carries out unloading decision according to the self attribute, the server load and the like; and the computation task mark participating in the unloading in the second stage meets the server of the computation task mark, the problem is converted into a multiple matching problem of bipartite graphs, and reasonable server distribution is carried out through a multiple maximum matching algorithm of a Gale-sharley idea. Finally, an optimal unloading decision and resource allocation result are obtained, the optimization effect when the task selects the server is eliminated, the server is prevented from being overloaded, and the energy consumption efficiency and the overall performance of the system are improved.
TABLE 1 simulation parameters
TABLE 2 comprehensive analysis of properties
Claims (3)
1. A calculation migration method for combined optimization of offshore MEC load balancing and resource allocation is characterized in that:
in the two-stage calculation migration method, the marine network composition constructed by the model mainly comprises 3 parts, wherein the first part is a core cloud network layer and comprises a large number of high-performance cloud servers for providing services for users; the second part is an edge server layer, all edge servers are divided into regions according to relative distances, each region consists of an unmanned aerial vehicle, an airship, a marine vernier and a land base station, and the unmanned aerial vehicle, the airship, the marine vernier and the land base station all have servers with moderate and heterogeneous performances; the third part is a terminal equipment layer, each area is composed of yachts, steamships, aircraft carriers and the like, and the yachts, the steamships, the aircraft carriers and the like have equipment with certain processing capacity or make unloading decision on the current calculation task according to the property of the task;
in the two-stage calculation migration method, task unloading is divided into two stages, the first stage is unloading decision, and a calculation task makes a decision according to self attributes, server load and the like; the second stage is resource allocation, marks servers meeting task requirements, converts the servers into multiple matching problems of bipartite graphs, and solves the multiple matching problems through a multiple maximum matching algorithm based on the idea of Gale-Shapley algorithm to balance edge network loads;
in the two-stage calculation migration method, task execution time delay is taken as constraint, channel resources and calculation cost are comprehensively evaluated, channels are reasonably distributed, waste of the channel resources is reduced, and communication quality and efficiency of a network during calculation unloading are guaranteed;
the calculation migration method for the offshore MEC load balancing and resource allocation combined optimization comprises the following steps:
1) the terminal node generates a calculation task and records the task attribute into a temporary array;
2) traversing all servers in sequence according to the attributes of each computing task, and storing energy consumption and task execution time delay of the computing task under each server into a temporary array;
3) taking task execution time delay as a constraint, judging whether the computing tasks need to be unloaded, and recording the server which can be unloaded by each computing task and the required energy consumption;
4) constructing a server dynamic price model according to the initial calculation cost, the unit calculation cost, the current calculation capacity and the number of processed tasks of the server; constructing a comprehensive evaluation model of the calculation tasks according to the energy consumption and the task execution delay of the calculation tasks under each server;
5) and searching the optimal matching of the calculation tasks selected to be unloaded in the optional server to convert into a multiple matching problem of bipartite graphs, and solving the multiple matching problem, wherein the optimal matching is finally obtained according to the idea of a Gale-Shapley algorithm in the graph theory in combination with the energy consumption, the time delay and the current price of the calculation tasks temporarily stored in the array.
2. The method for computing migration of combined optimization of offshore MEC load balancing and resource allocation as recited in claim 1, wherein in step 4), the server constructs a dynamic price model of the server according to initial computation cost, unit computation cost, current computation capability and number of tasks processed; the method comprises the following steps that a calculation task comprehensive evaluation model is constructed by a calculation task according to energy consumption and task execution time delay generated under each server, and specifically comprises the following steps:
4-1) setting initial price Cost of server0And Price per unit of computing resourcej；
4-2) constructing a dynamic price model of the server, and judging whether the current server is in a high load or idle state:
in the ocean, the computing resource deployment of the edge server is easy to generate 'long tail distribution', the server with strong capability executes tasks frequently, the server with weak capability is idle, the stability of the network is influenced, and the dynamic price model of the server is constructed as follows:
among them, Cost0An initial cost for the server; pricejCalculating the price of the resource for the server unit; ratio (R)jA rate of use of computing resources for a current server; surplusjComputing resources left for the current server;
total computing resources for the current server; miCalculating the task quantity of the task; n is a task set processed by the server;
4-3) according to energy consumption and task execution time delay of the calculation task under each server, constructing a comprehensive evaluation model of the calculation task as follows:
s.t.C1:αij∈{0,1},λ+μ＝1
C2:Pi≤Pi,max
C3:
C4:
C5:
C6:
where C1 is the unload decision, αijIs a binary variable; c2 is the constraint of the terminal transmission power; c3 is the maximum delay requirement of the terminal; c4 is the constraint of terminal and server power; c5 is the restriction that the terminal can only select one server to send tasks; c6 is the maximum number of simultaneous online processing tasks of the server; t isi *j，
Respectively representing the time delay and the energy consumption after the normalization processing.
3. The method for computing migration of combined optimization of offshore MEC load balancing and resource allocation as recited in claim 1, wherein in step 5), the multiple matching problem that the optimal matching is found and converted into bipartite graph in the optional server for solving is performed, and the optimal matching is finally obtained according to the idea of Gale-Shapley algorithm in graph theory in combination with the computing task energy consumption, time delay and current price of the server temporarily stored in the array; the method specifically comprises the following steps:
5-1) traversing the calculation tasks and the optional edges thereof in sequence, matching the task with the server if only one optional edge of one task is connected with the server, and calculating and updating the current cost of the server according to the formula (1);
5-2) improving the thought of sequential point selection in sequence, matching task nodes from less to more according to the connectivity in sequence, selecting the edge with the lowest cost as the current matching by using the thought of a greedy algorithm in an approximate algorithm, calculating and updating the current cost of the server according to a formula (1), and storing the current cost into a temporary array;
5-3) matching the remaining nodes, selecting the edge with the least cost as the current matching, if the number of tasks received by the selected server reaches a threshold value, generating conflict and needing coordination, reserving the task with the connectivity of 1 in the tasks matched by the server, canceling the matching with the largest cost in the remaining tasks, and canceling the optional edge by the calculation task and trying to match again; and if the conflict still exists, executing the step 5-1) to the step 5-2) to recurse in sequence, and when the optional edges of the task do not meet the matching condition, giving up the task and judging that the task processing fails.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111257708.1A CN113992678A (en) | 2021-10-27 | 2021-10-27 | Calculation migration method for offshore MEC load balancing and resource allocation joint optimization |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111257708.1A CN113992678A (en) | 2021-10-27 | 2021-10-27 | Calculation migration method for offshore MEC load balancing and resource allocation joint optimization |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113992678A true CN113992678A (en) | 2022-01-28 |
Family
ID=79742809
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111257708.1A Withdrawn CN113992678A (en) | 2021-10-27 | 2021-10-27 | Calculation migration method for offshore MEC load balancing and resource allocation joint optimization |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113992678A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114385272A (en) * | 2022-03-24 | 2022-04-22 | 山东省计算中心（国家超级计算济南中心） | Ocean task oriented online adaptive computing unloading method and system |
| CN114615705A (en) * | 2022-03-11 | 2022-06-10 | 广东技术师范大学 | Single user resource allocation strategy method based on 5G network |
| CN114844900A (en) * | 2022-05-05 | 2022-08-02 | 中南大学 | Edge cloud resource cooperation method based on uncertain demand |
| CN115665797A (en) * | 2022-10-31 | 2023-01-31 | 齐鲁工业大学 | Offshore unloading and resource allocation method for mobile edge computing |
-
2021
- 2021-10-27 CN CN202111257708.1A patent/CN113992678A/en not_active Withdrawn
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114615705A (en) * | 2022-03-11 | 2022-06-10 | 广东技术师范大学 | Single user resource allocation strategy method based on 5G network |
| CN114615705B (en) * | 2022-03-11 | 2022-12-20 | 广东技术师范大学 | Single-user resource allocation strategy method based on 5G network |
| CN114385272A (en) * | 2022-03-24 | 2022-04-22 | 山东省计算中心（国家超级计算济南中心） | Ocean task oriented online adaptive computing unloading method and system |
| CN114844900A (en) * | 2022-05-05 | 2022-08-02 | 中南大学 | Edge cloud resource cooperation method based on uncertain demand |
| CN114844900B (en) * | 2022-05-05 | 2022-12-13 | 中南大学 | Edge cloud resource cooperation method based on uncertain demand |
| CN115665797A (en) * | 2022-10-31 | 2023-01-31 | 齐鲁工业大学 | Offshore unloading and resource allocation method for mobile edge computing |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113992678A (en) | Calculation migration method for offshore MEC load balancing and resource allocation joint optimization | |
| CN113950103B (en) | Multi-server complete computing unloading method and system under mobile edge environment | |
| CN111930436B (en) | Random task queuing unloading optimization method based on edge calculation | |
| CN110096362B (en) | Multitask unloading method based on edge server cooperation | |
| CN110413392A (en) | The method of single task migration strategy is formulated under a kind of mobile edge calculations scene | |
| CN111163178B (en) | Game theory-based service deployment and task unloading method in edge computing | |
| CN108924254B (en) | User-centered distributed multi-user computing task unloading method | |
| CN110519370B (en) | Edge computing resource allocation method based on facility site selection problem | |
| US11968122B2 (en) | Joint optimization method and system for delay and spectrum occupation in cloud-edge collaborative network | |
| CN113961264B (en) | Intelligent unloading algorithm and system for video monitoring cloud edge cooperation | |
| CN111263401A (en) | Multi-user cooperative computing unloading method based on mobile edge computing | |
| CN112040512B (en) | Mist computing task unloading method and system based on fairness | |
| CN113254095A (en) | Task unloading, scheduling and load balancing system and method of cloud edge combined platform | |
| CN112888005A (en) | MEC-oriented distributed service scheduling method | |
| CN111158893B (en) | Task unloading method, system, equipment and medium applied to fog computing network | |
| CN114153515B (en) | Highway internet of vehicles task unloading algorithm based on 5G millimeter wave communication | |
| CN113365290B (en) | Game theory calculation unloading method based on greedy strategy in world fusion network | |
| CN111324429A (en) | Micro-service combination scheduling method based on multi-generation ancestry reference distance | |
| CN116582173B (en) | Method, device and storage medium for processing data by satellite-based distributed network | |
| CN117579701A (en) | Mobile edge network computing and unloading method and system | |
| CN110177383B (en) | Efficiency optimization method based on task scheduling and power allocation in mobile edge calculation | |
| CN116744367A (en) | Unloading method based on double-layer unloading mechanism and multi-agent algorithm under Internet of vehicles | |
| CN115955685A (en) | Multi-agent cooperative routing method, equipment and computer storage medium | |
| CN114745386B (en) | Neural network segmentation and unloading method in multi-user edge intelligent scene | |
| CN113573280B (en) | Vehicle edge calculation cost-effective optimization method, system, equipment and terminal |
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 | ||
| WW01 | Invention patent application withdrawn after publication | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20220128 |