CN113923223A - User allocation method with low time cost in edge environment - Google Patents

User allocation method with low time cost in edge environment Download PDF

Info

Publication number
CN113923223A
CN113923223A CN202111347569.1A CN202111347569A CN113923223A CN 113923223 A CN113923223 A CN 113923223A CN 202111347569 A CN202111347569 A CN 202111347569A CN 113923223 A CN113923223 A CN 113923223A
Authority
CN
China
Prior art keywords
user
allocation
strategy
edge server
time cost
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.)
Granted
Application number
CN202111347569.1A
Other languages
Chinese (zh)
Other versions
CN113923223B (en
Inventor
王福田
王德春
汤进
章程
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Anhui University
Original Assignee
Anhui University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Anhui University filed Critical Anhui University
Priority to CN202111347569.1A priority Critical patent/CN113923223B/en
Publication of CN113923223A publication Critical patent/CN113923223A/en
Application granted granted Critical
Publication of CN113923223B publication Critical patent/CN113923223B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

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
    • 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
    • H04L67/1021—Server selection for load balancing based on client or server locations
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/14—Network analysis or design
    • 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
    • H04L67/1012—Server selection for load balancing based on compliance of requirements or conditions with available server resources

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Data Exchanges In Wide-Area Networks (AREA)

Abstract

The invention discloses a low-time-cost user allocation method in an edge environment, which comprises the steps of obtaining a resource request of a user and residual resources of an edge server, and sequencing and segmenting the user; initializing iteration parameters for user iteration of each segment; setting an initial user allocation strategy; generating a new allocation strategy; calculating the time cost of the new distribution strategy and comparing the time cost with the time cost of the current distribution strategy, if the time cost of the new distribution strategy is smaller, replacing the current strategy with the new strategy, otherwise, replacing the current strategy according to a certain probability; updating parameters and judging the condition of ending the circulation; judging the quality of a strategy generated by each section of user iteration; looping the previous step, and performing iteration of the next section of user; and obtaining the optimal distribution strategy after iteration. The Mongolian reduces the time cost of the user in the edge computing environment, and meets the requirement of low time cost for the user edge side to execute the task.

Description

User allocation method with low time cost in edge environment
Technical Field
The invention belongs to the technical field of user allocation optimization, and particularly relates to a low-time-cost user allocation method in a marginal environment.
Background
In recent years, the number of cloud and mobile network connection terminal devices (including mobile phones, wearable devices, sensors, and a wide range of internet of things devices) has increased dramatically, and due to the limited computing power and battery power of mobile and internet of things devices, their computing tasks are often offloaded to servers of cloud application providers. However, as the number of mobile and internet of things devices grows exponentially, cloud computing becomes increasingly difficult to handle the huge workload and network congestion. This makes it difficult to provide low latency and reliable connections to end users, especially those who wish to obtain real-time responses from applications.
To address this problem, edge computing, often referred to as mobile edge computing, MEC, or multiple access edge computing, provides execution resources (computation and storage) for applications, and the network is close to the end user, often inside or at the border of the operator's network, is proposed as a new distributed computing paradigm.
The edge infrastructure may be managed or hosted by a communication service provider or other type of service provider. Some use cases require different applications to be deployed at different sites. In such a scenario, a distributed cloud is very useful, which can be seen as an execution environment for applications on multiple sites, including connectivity managed as a solution. The primary benefits provided by edge computing solutions include low latency, high bandwidth, device processing and data offloading, and trusted computing and storage. In the edge environment, each base station is equipped with a certain amount of computing resources, allowing mobile users to be provided with computing power at the internet access level.
However, while edge computing provides new opportunities, it also presents many new challenges, such as edge user allocation issues, with edge servers as intermediate provisioning stations with limited computing resources.
The terminal user is changed from a data consumer to a data generator and a data consumer, a large amount of data interacts with the cloud, so that delay and reliability are difficult, the cloud server cannot well meet the requirements of the user, the edge server is close to the edge side of the user, reliability and delay are guaranteed, but limited resources of the edge server cannot accommodate and allocate a large number of users, and therefore a reasonable allocation strategy is needed to guarantee maximization of resource utilization; there are studies to guarantee the resource utilization rate by adjusting the deployment position of the edge server, which have disadvantages in that the maximization of the allocated users cannot be guaranteed, and studies to improve the resource utilization rate while guaranteeing the maximization of the allocated users, but cannot achieve the optimization in terms of time cost. The coverage and resource amount of the edge server are limited, and the geographic positions of the users are randomly distributed, so that it is difficult to ensure that the users are most distributed and the time cost is lowest.
At present, two difficulties exist in the user allocation method in the edge environment:
firstly, edge servers and users are in a one-to-many relationship, the selection of the users by one edge server affects the overall user allocation effect, particularly, under the condition that the size of users with limited resources is large, the residual resources of some servers are less, the resource requirements of users in all coverage ranges cannot be met, selective allocation is needed under the condition that resource constraint conditions are met, the resource requirements of some users are larger, and the influence on other users is reduced as much as possible in the allocation process;
secondly, when enough users are distributed, the requirement of minimum time cost, different architectures of edge servers, different calculation speeds and the highest calculation speed of the edge service cannot be met, and all the users cannot be distributed under the constraints of coverage and resources; under the condition of limited resources, the edge server distributed to more users may not have the fastest calculation speed, and the edge server distributed with the fastest calculation speed may have lower total time cost than the edge server distributed with a plurality of users; .
Disclosure of Invention
The purpose of the invention is as follows: the invention aims to solve the defects in the prior art, and provides a low-time-cost user allocation method in an edge environment, which can obtain an optimal user allocation strategy with lower time cost under the conditions of limited edge server resources and limited edge server signal coverage area, reduce the time cost of a user in an edge computing environment, meet the requirement of low time cost for executing tasks at the edge side of the user, and reduce the cost of an APP service provider.
The technical scheme is as follows: the invention discloses a low-time-cost user allocation method in a marginal environment, which comprises the following steps of:
(1) obtaining a user resource request, residual resources of each edge server and longitude and latitude of each user and each edge server to obtain a user set U ═ U1，…，umAnd the set of edge servers S ═ S1，…，snAnd then, ordering the users according to the size of the requested resource, dividing the user set into L sections with equal quantity and size according to the size of the resource, wherein each section of users is
Figure BDA0003354776480000021
Figure BDA0003354776480000022
1, { …, L }; m and n respectively refer to the total number of the acquired users and the total number of the edge servers;
user uiResource requests are represented as u by a vectori＝<vcpu，memory，…，weight>；
Edge server sjIs represented by a vector
Figure BDA0003354776480000031
Wherein n is the total number of users requesting to calculate resources, m is the total number of edge servers, and vcpu represents the number of cpu micro-cores; memory represents a memory; weight represents the workload of the user; range represents the effective coverage area of the edge server signal;
Figure BDA0003354776480000032
representing edge servers sjThe larger the fcpu value, the better the performance, the faster the calculation speed(ii) a Calculating the geographical distance between each user and each edge server through longitude and latitude information, wherein the user uiAnd edge server sjFor the geographical distance dijIs represented by dijWhen the coverage area is less than or equal to s, the user is represented in the edge server sjIn the range of sjIn case the remaining resources of (1) are sufficient, user uiCan be distributed to edge servers sj. Due to the limited edge server coverage distance, the assigned user must be within the signal coverage of the corresponding edge server;
(2) initializing iteration parameters, including an initial temperature INIT _ TEM, a temperature decay RATE RATE, a termination temperature FINNAL _ TEM, an inner circulation time IN _ LOOP, a continuous non-updating current maximum solution time LIMIT and an outer circulation time OUT _ LOOP;
the time for outputting the optimal distribution strategy is controlled through the iteration parameters, and the optimal distribution strategy can be better output in a shorter time through the iteration parameters; wherein, the parameter INIT _ TEM is a larger integer, the RATE is not less than 0.90, the value of FINNAL _ TEM is about 1E-5, the IN _ LOOP, LIMIT and OUT _ LOOP are determined according to the data scale, the IN _ LOOP is not less than the total number of users;
(3) initial allocation strategy a ═ { a ═ aij(ii) a i belongs to {1, …, n }, j belongs to {1, …, m } }, wherein an initialized allocation decision A is assigned to an initial allocation strategy of a first outer loop, and then iteration is performed on segmented users respectively to generate a better allocation strategy, namely, before the iteration of a first section of users is started, an initial allocation strategy is generated firstly, and then the initial allocation strategy for each section of user iteration behind is the optimal allocation strategy generated after the iteration of the previous section of users;
in the process, a user needs to be within the coverage range of the edge server, the residual resources of the edge server can meet the requirements of the user, then an initial allocation strategy is generated according to the processing speed of the edge server, the alternative edge servers of each user are sorted according to the processing speed, an unallocated user is randomly selected, the user is allocated to the edge server with the highest processing speed in the alternative edge servers, if the processing speed of the edge server still cannot meet the resource constraint, the next alternative edge server is judged until the user is allocated, and if the alternative edge server which does not meet the resource constraint finally, the next alternative edge server is allocated to the cloud end;
(4) generating a new allocation strategy and calculating the total time cost T of the current allocation strategytotle
Randomly selecting a user to re-formulate an allocation strategy on the basis of the current allocation strategy, judging whether the adjusted server can be allocated to the user without the allocated server when the allocation strategy is re-formulated after the operation is finished, and simultaneously judging whether the changed user can be continuously allocated to other edge servers covering the user;
wherein, if the user is assigned to the edge server, the time cost is
Figure BDA0003354776480000041
Namely, it is
Figure BDA0003354776480000042
Denoted as user uiAt edge server sjThe time cost of;
if the user does not have the assignable edge server, the user is assigned to the cloud end to execute the task, and the time cost is
Figure BDA0003354776480000043
Figure BDA0003354776480000044
Figure BDA0003354776480000045
CapacityiRepresenting a user UiThe amount of computing power provided by the resources in demand,
Figure BDA0003354776480000046
Figure BDA0003354776480000047
representative user uiD represents the total number of resource types, k is {1, …, D };
(5) updating allocation decisions based on changes in time cost
Computing a new classification policy AnewTotal time cost of
Figure BDA0003354776480000048
And a current allocation policy A for generating a new allocation decisioncurrentTotal time cost of
Figure BDA0003354776480000049
Comparing, the time cost is low, the time cost is high when replacing, if the time cost is equal or the time cost of the new distribution strategy is higher, whether the distribution strategy is updated or not is selected according to the corresponding probability P, and therefore the local optimal solution can be skipped out with a certain probability;
Figure BDA00033547764800000410
Figure BDA00033547764800000411
Figure BDA00033547764800000412
where delete is the difference between the time cost of the newly generated allocation policy and the time cost of the current allocation policy,
Figure BDA00033547764800000413
representing the total time cost of the newly generated allocation policy,
Figure BDA00033547764800000414
represents the total time cost of the current allocation policy; total time cost TtotleThe sum of the time cost for each user to use the server includes the user allocated to the edge server and the user allocated to the cloud, each allocation strategy generates corresponding time cost, the time cost is an important index for evaluating the quality of the allocation strategy, and the time cost can be reduced by adjusting the allocation strategy of the user.
If delete ≦ 0 and allocationcur≤allocationnewThen the current allocation policy is updated with the newly generated allocation policy, wherein allocationcurIs the total number of users allocated by the current allocation policy, allocationnewRepresenting the total number of users allocated by the newly generated allocation policy, otherwise, randomly generating an r e [0,1 ∈]If r is greater than or equal to P, the condition of delete is less than or equal to a and allocation is judgednew≥allocationcur-b, updating the current allocation policy if the condition holds;
because the distribution strategy worse than the current solution is accepted according to a certain probability p, the local optimal solution is skipped, and when the worse distribution strategy is accepted, the range of the worse distribution strategy is narrowed in order to narrow the search range; a and b control the extent of the worse allocation strategy accepted, a representing the total time cost of the new allocation strategy accepted
Figure BDA0003354776480000051
Above
Figure BDA0003354776480000052
B represents the maximum value that the total number of users allocated by the newly generated allocation strategy is less than the total number of users allocated by the current allocation strategy;
the above conditions may also be described as if the newly generated allocation policy is worse than the current policy, then if
Figure BDA0003354776480000053
Not exceeding
Figure BDA0003354776480000054
And then allocatenewAt a minimum, allocatationcur-b；
(6) Judging whether the circulation process meets the termination condition
The end conditions of the cycle are controlled by six parameters, INIT _ TEM, RATE, FINNAL _ TEM, IN _ LOOP, LIMIT and OUT _ LOOP; judging whether to accept the poor distribution decision through the probability P, and finally outputting the optimal distribution strategy;
Figure BDA0003354776480000055
(7) after the iteration of each section of user is finished, judging the advantages and disadvantages of the distribution strategy generated by the iteration of the previous section of user and the distribution strategy generated by the iteration of the current section of user, and taking the more optimal distribution strategy as the initial distribution strategy of the iteration of the next section of user.
Further, the step (1) of the edge server sjIs calculated at a speed of
Figure BDA0003354776480000056
The computing speed of the cloud server is fcloud,
Figure BDA0003354776480000057
because the allocation policy maker needs to obtain the resource demand of the user and the resource size of the edge server at the same time, and the processing speeds of the two servers are different,
Figure BDA0003354776480000058
the processing speed of the cloud server is the fastest, but the time cost of data transmission distributed to a cloud end by a user is higher due to the fact that the cloud server is far away from the user, the time cost is increased due to transmission quality and the like, and the data transmission time cost between the edge server and the user can be ignored due to the fact that the edge server is close to the user side on the network edge side.
Further, a in the step (3)ijThe distribution strategy of the user is represented, the value is 0 or 1, 0 represents that the user is distributed to the cloud, and 1 representsSubscriber allocation to edge servers sj(ii) a Namely:
Figure BDA0003354776480000061
for any allocation policy a ═ aij(ii) a i ∈ {1, …, n }, j ∈ {1, …, m } }, all must satisfy the constraint
Figure BDA0003354776480000062
That is, the edge server and the users are in a one-to-many relationship, one edge server can be allocated with a plurality of users, and one user can be allocated to only one edge server;
corresponding coverage constraint and resource limitation constraint must be satisfied in the process of allocating servers by a user, only an edge server which covers the user and has enough resources can become an alternative server of the user, and a final allocation strategy must satisfy that the sum of the resource requirements of each type of the users allocated to the edge server is smaller than the size of the residual resources of the edge server;
wherein the coverage constraint is:
Figure BDA0003354776480000063
the resource limitation constraint is:
Figure BDA0003354776480000064
dijrepresenting the geographical position distance between the user and the edge server, and adopting the longitude and latitude distance between the user and the edge server; rangejRepresenting the signal coverage of a server j, wherein n represents the total number of users applying for computing resources, and m represents the total number of servers;
Figure BDA0003354776480000065
representing user uiThe requirements of the kth class of resources of (1),
Figure BDA0003354776480000066
representing edge servers sjClass k ofThe remaining amount of resources, all allocated to the edge server sjThe sum of the demands of various resources of the user is less than the residual quantity of various resources of the edge server j, such as vcpu, memory and the like, the resource types are D-class in total, and the computing power provided by the demand resources is CapacityiThe more resources required by the user, the stronger the corresponding computing power, and the smaller the time cost; different computing capacities can be obtained when users have different requirements;
Figure BDA0003354776480000067
further, the specific process of generating the new allocation policy in step (4) is as follows:
randomly selecting a user, and then changing and adjusting the user UselectIf the other alternative edge servers of the user meet the constraint conditions, the user is allocated to a certain alternative server, otherwise, an alternative server is randomly selected, the user allocated on the server is allocated to the other alternative servers until the server meets the constraint conditions for allocating the selected user, the selected user is allocated to the server, and whether the original edge server of the selected user can reallocate other unallocated users is judged;
wherein, Uselect＝Random({ui，i∈[1,n]})；
YselectFor user uiThe number of alternative servers of (a); then:
Figure BDA0003354776480000071
in the above formula, Yselect>There are two cases to consider when 1, one is the reselected server snewResource can be allocated to user UselectIn this case the allocation policy can be adjusted directly and the edge server s before the change checkedoldWhether other unassigned users can be reassigned; alternatively, s isnewHas not been allocated to the user U enoughselectThis is the need to hold snewThe user on the upper allocation withdraws e users and releases the resources until there are enough resources to allocate the user UselectNamely:
Figure BDA0003354776480000072
at snewE users dropped on UselectAfter allocation, it is checked whether allocation to other alternative servers can continue.
UselectIs assigned to the network
Figure BDA0003354776480000073
The value of (a) becomes 0 and,
Figure BDA0003354776480000074
becomes 1, i.e. user uiBy distribution to edge servers soldBecomes distributed to the edge server snew(ii) a Other users who change the policy may change the value of the decision variable accordingly.
Further, the step (6) generates a new distribution strategy and judges whether updating of the current distribution strategy is performed through an inner LOOP, LIMIT and OUT LOOP return to an initial value when the new distribution strategy directly replaces the current strategy, LIMIT is decreased by 1 when the current distribution strategy is not directly replaced, OUT LOOP is decreased by 1 when LIMIT is equal to 0, IN LOOP is decreased by 1 each time of the inner LOOP, and initial temperature of the outer LOOP is decreased by INIT _ TEM × RATE each time, and the outer LOOP is not ended until INIT _ TEM is less than or equal to finan _ TEM or the number of outer LOOPs is used up;
wherein, whether to terminate the circulation is judged by adopting the following formula;
Figure BDA0003354776480000075
ending the segmentation iteration may also be decided, for example, by i ≧ 5.
Furthermore, after the outer loop is finished in the step (7), the best allocation strategy at the beginning of the outer loop and the best allocation strategy generated when the inner loop is finished at this time need to be judged, so that the best allocation strategy becomes the best allocation strategy of the outer loop at the next time, iteration is performed again, the inner loop is updated until the outer loop is finished, and the best allocation strategy is finally output; the best allocation strategy after each segmentation iteration is the initial allocation strategy of the next segmentation iteration; where he decides whether the allocation strategy is good or bad by the overall time cost.
Has the advantages that: compared with the prior art, the invention has the following advantages:
(1) whether an initial allocation strategy is generated or a new allocation strategy is generated by adjusting on the basis of the initial allocation strategy, corresponding coverage constraint and resource constraint need to be met in the allocation process, the difficulty of problems is increased through the coverage constraint and the resource constraint, and the task completion quality of feasibility of user allocation is ensured through the corresponding coverage constraint and the resource constraint, so that the final allocation strategy must meet constraint conditions, the setting of initial iteration parameters is reasonable, and the larger the initial iteration parameters are, the more the iteration times are, the better the iteration times are.
(2) Under the condition that the resources of the edge server are limited, the invention can meet the requirements of users, greatly reduce the total time cost and greatly improve the resource utilization rate. The distribution strategy is continuously updated in an iteration mode to jump out of the local optimal distribution strategy in a probability mode, the distribution strategy close to the optimal is obtained, the distribution strategy which is finally output is irrelevant to the input initial distribution strategy, meanwhile, the distribution strategy has asymptotic convergence, and the distribution strategy is a global optimization method which converges on the global optimal solution with certain probability.
Drawings
FIG. 1 is a process flow diagram of the present invention;
FIG. 2 is a detailed allocation flow diagram of an embodiment;
FIG. 3 is a schematic diagram illustrating the distribution of servers and users in an embodiment;
FIG. 4 is a schematic diagram of the time cost of various algorithms for the number of fixed servers and the number of users whose resource amount is increasing in the embodiment;
FIG. 5 is a schematic diagram of time cost of each algorithm, with a fixed number of users, an increasing number of edge servers and an increasing amount of resources in the embodiment;
fig. 6 is a time cost diagram of algorithms in which the number of users is proportional to the number of edge servers and the number of users is increased continuously in the embodiment.
Detailed Description
The technical solution of the present invention is described in detail below, but the scope of the present invention is not limited to the embodiments.
As shown in fig. 1, the low-time-cost user allocation method in the edge environment of the present invention can jump out the local optimal allocation policy and find the global optimal allocation policy, and specifically includes the following steps:
(1) the method comprises the steps of obtaining a user resource request, residual resource amount of each edge server and respective longitude and latitude, sequencing users according to the size of required resources, dividing the users into a plurality of parts with equal number, collecting the resource request of the users and the residual resource amount of the edge servers, and representing the resource size in a vector mode. For example, according to the amount of the resource required by the user, the user is divided into 5 parts, each part is represented by i, and the number of the users in each part is n/5 of the total number of the users.
(2) And setting iteration parameters, controlling the time for outputting the optimal distribution strategy through the iteration parameters, and simultaneously outputting the optimal distribution strategy better in a shorter time through reasonable iteration parameters.
(3) And generating an initial allocation strategy according to the processing speed of the edge server, and starting segmentation iteration on the user.
(4) Generating a new allocation strategy, calculating the time cost of the allocation strategy
Figure BDA0003354776480000092
(5) And iteratively updating the allocation decision and the iteration statement for each part of the users according to the change of the time cost.
(6) Judging whether a termination condition is met or not, outputting an optimal distribution strategy, and continuously iterating to generate a better distribution strategy if the generation of a new distribution strategy and the judgment of whether the current distribution strategy is updated are carried out in an inner loop; after each external loop is finished, the optimal allocation strategy at the beginning of the external loop and the allocation strategy generated when the current internal loop is finished are judged to be better, the optimal allocation strategy becomes the optimal allocation strategy of the next external loop, iteration is carried out again, the internal loop is updated until the external loop is finished, and the optimal allocation strategy is finally output. The best allocation strategy after the end of each segmentation iteration is the initial allocation strategy for the next segmentation iteration.
As shown in fig. 2 and fig. 3, the low-time-cost user allocation method in the edge environment of the present embodiment includes the following steps:
step (1), obtaining a user resource request, residual resource amount of each edge server and respective longitude and latitude
(1.1) acquiring resource requests of users and residual resource amount of the edge server in a centralized manner, and representing the resource size in a vector form, wherein the expressions (1) and (2) comprise resource information of the users and the edge server, and the expression (3) represents the relation of computing speeds of the cloud server and the edge server, wherein fcloudWhich represents the computing speed of the cloud server,
Figure BDA0003354776480000091
representing edge servers SjThe calculated speed of (2).
Uiui＝<vcpu，memory，…，weight> (1)
Figure BDA0003354776480000101
Figure BDA0003354776480000102
Acquiring longitude and latitude of each user and the edge server, calculating the geographical position distance between each user and the server through a Haverine formula, and calculating to obtain an alternative edge server set of each user described by the formulas (4) and (5)Close SUiAnd a set of overlay users SU per edge serverjThe number of each set is uncertain, and to prevent the user from not being covered by the signal, there is an intersection between the signal coverage areas of the edge servers, so the user may have multiple alternative servers.
USi＝{s1，s，…} (4)
SUj＝{u1，u2，…}(5)
(1.2) sorting the users according to the size of the resources required by the users, dividing all the users into five segments with the size of n/5, and starting with the value of i being 0.
Step (2) setting iteration parameters
IN this embodiment, the parameters INIT _ TEM is 3000, RATE is 0.96, fine _ TEM is 1E-5, IN _ LOOP is 5000, LIMIT is 3000, and OUT _ LOOP is 1000, and the larger the parameter value is, the more the cycle number is, the more the output distribution policy is, the closer the distribution policy is to the global optimum distribution policy, but the corresponding time overhead is also larger.
Step (3) generating an initial allocation strategy
(3.1) equation (6) assigns policy variables to users, { aijThe size of the server is nxm, the total number of the users is multiplied by the total number of the edge servers, the value is 0, and the user is distributed to the cloud end; a value of 1 indicates that the user is assigned to the corresponding edge server, a, before the initial assignment policy is generatedijAll values of (A) are 0;
Figure BDA0003354776480000103
(3.2) randomly selecting one user UiTo user UiThe allocation is carried out, firstly, the alternative server set of the user is searched, and the calculation speed f of the edge server is calculatededgeThe size of the edge server US is sorted, and the edge server US with the highest calculation speed is judgedi[0]Whether constraint conditions are met or not and whether residual resources of the edge server meet the user u or notiIf the condition is satisfied, the corresponding decision variable value becomes 1, and if the condition is not satisfied, the decision variable value is sequentially judgedHis alternative server, if the condition is not met aijSatisfying equation (7), the user is assigned to the cloud.
Figure BDA0003354776480000111
(3.3) after all the users are allocated, generating an initial allocation strategy, wherein the initial allocation strategy is used as the input of the first stage user iterative algorithm, but the state of the initial allocation strategy does not influence the final output optimal allocation strategy. A user can only be allocated to one user, so the generated initial allocation policy must satisfy the constraint of equation (8).
Figure BDA0003354776480000112
Only the edge server covering the user and having sufficient resources can become the alternative server of the user, so the final allocation policy must satisfy formula (9), that is, the sum of the resource requirements of each type of all users allocated to the edge server is smaller than the size of the remaining resource of the edge server;
Figure BDA0003354776480000113
(3.4) the initial allocation strategy generated at the moment is not the optimal allocation strategy, continuous iteration is needed to generate the optimal low-time-cost user allocation strategy, and the initial allocation strategy is assigned to the variable AbestAs the current best allocation strategy.
Step (4), generating a new distribution strategy and calculating the time cost of the distribution strategy
(4.1) after the iteration parameters and the initial allocation strategy are set, starting to enter iteration for searching a global optimal allocation strategy, adjusting by an optimal decision generated by the previous step of outer loop to generate a new allocation strategy, and if the current loop enters the loop for the first time, generating by the initial allocation strategy, wherein the relation between the current loop and the new allocation strategy is represented by an expression (10);
Figure BDA0003354776480000114
(4.2) at AcurrentBased on the user allocation strategy, randomly selecting one or more user UselectBecause of the complexity of the edge computing environment, this selection process may need to be performed multiple times; random to one UselectThen, the user U is judgedselectUS of alternative server setselectState by YselectMeans of USselectThe collection size, equation (11) is the judgment of the collection state of the user server; if Y isselectReselecting U as 0 or 1select；
Uselect＝Random({ui，i∈[1,n]})
Figure BDA0003354776480000121
(4.3) referring to fig. 2, since there are a large number of overlapping areas in the coverage area of the server, the allocation policy of one user is usually changed, which may affect the allocation decisions of many users; find a satisfied Yselect>1 user UselectThen, the user allocation strategy is adjusted, firstly in the USselectIn selecting an edge server snewAnd U isselectWhen the currently allocated edge servers are different, whether the residual resources on the edge servers are enough to allocate the user U is judged by the formula (12)selectBy computing edge servers snewRemaining amount of kth class resource and UselectResource requirements of class k
Figure BDA0003354776480000129
Handle bar
Figure BDA00033547764800001210
The state of (2) is divided into two states of formula (13);
Figure BDA0003354776480000122
Figure BDA0003354776480000123
(4.4) when SnewAll types of resources satisfy the user UselectThen, the user is directly assigned to SnewConcurrently adjusting the decision variables and the edge server according to equation (14);
Figure BDA0003354776480000124
when in use
Figure BDA0003354776480000125
When, to snewMaking adjustments by first revoking a portion of the assignment to snewUntil the resource on the server satisfies UselectThe revoked user set is Udel＝{u1,u2,…,ueThe change of the revoked user in the resource amount satisfies the formula (15), and simultaneously changes the values of the corresponding allocation strategy variables of all revoked users,
Figure BDA0003354776480000126
in order to revoke the value of the policy currently assigned to the user,
Figure BDA0003354776480000127
the value of the new allocation policy for the revoked user;
Figure BDA0003354776480000128
(4.5) adjusting U according to equation (14) after the revocation user is completedselectWhile determining the set UdelWhether or not the user in the house has an alternative garmentThe server can continue to distribute, the value of the adjustment distribution strategy variable meeting the distribution condition is 1, and the value of the distribution strategy variable not meeting the distribution condition is unchanged;
(4.6) after a new allocation strategy is generated, simultaneously having a current allocation strategy and a new allocation strategy, and calculating the time cost of the current allocation strategy and the time cost of the new allocation strategy, wherein the time cost of task transmission and the time cost of task execution of a user in the cloud end are required to be considered when the user is allocated to the cloud end, and the calculation is carried out by adopting a formula (16);
Figure BDA0003354776480000131
(4.7) the time cost of the user to distribute to the edge server, because the calculation speed of the edge server is different, the distribution strategy of the user is different, and the generated time cost is also different, and the calculation is carried out by using the formula (17);
Figure BDA0003354776480000132
wherein the Capacity in formula (16) and formula (17)iComputing Capacity provided by the amount of resources required to represent the user, computing Capacity using equation (18) while computing the cost of timei；
Figure BDA0003354776480000133
(4.8) finally calculating the total time cost of each allocation policy using equation (19);
Figure BDA0003354776480000134
step (5), according to the change of the time cost, the distribution decision is updated in an iterative way
(5.1) calculating the time cost of the current allocation strategy
Figure BDA0003354776480000135
And time cost of newly generated allocation policies
Figure BDA0003354776480000136
Calculating the difference value delete between the two by using the formula (20), and judging whether the distribution strategy needs to be updated or not;
Figure BDA0003354776480000137
Figure BDA0003354776480000138
(5.2) judging the size of the delete, if the delete is less than or equal to 0 and allocationcur≤allocationnewWith AnewAlternative Acurrent，
LIMIT and OUT _ LOOP return initial values, internal cycle number IN _ LOOP-1; if the delete >0 is present, the data is decoded,
the probability p is calculated using equation (21),
generating a random number r epsilon [0,1], comparing the relation of r and p, if r is less than or equal to p, not replacing the current allocation strategy, such as
Fruit r > p, continue judging condition
delate≤a and allocationnew≥allocationcurB, if the condition is true, using AnewInstead of
Trade AcurrentOtherwise, not replace when
Pre-allocation policy, IN _ LOOP-1, LIMIT-1, regardless of the relationship of r to p;
Figure BDA0003354776480000141
(5.3) judging whether LIMIT is 0, if LIMIT is 0, finishing the inner LOOP, and OUT _ LOOP + 1; LIMIT ≠ 0, judges whether IN _ LOOP is 0, if IN _ LOOP is 0, the inner LOOP is ended; if both LIMIT and IN _ LOOP are not 0, continuously and circularly generating a new distribution strategy to judge whether to replace the current strategy;
(5.4) after completion of the internal circulation, judgment AbestAnd optimum A after completion of internal circulationcurrentTime cost of
Figure BDA0003354776480000142
And
Figure BDA0003354776480000143
is large or small, if
Figure BDA0003354776480000144
Lower, not update the optimal allocation policy, otherwise AcurrentAlternative AbestBecoming a new optimal allocation strategy and becoming the input of the next outer loop; if the value of β is judged by equation (22), and β is false, the outer loop is ended, otherwise, INIT _ TEM is updated by equation (23), and the next outer loop is entered;
β＝OUT_LOOP OR (INIT_TEM≥FINNAL_TEM) (22)
INIT_TEM＝INIT_TEM×RATE (23)
step (6), after the iterative updating distribution strategy is finished, the final AbestNamely the optimal user allocation strategy obtained by the i-th section of users through iteration,
Figure BDA0003354776480000145
is the time cost of the optimal user allocation strategy.
(6.1) output AbestComparison of
Figure BDA0003354776480000146
Time cost of optimal allocation strategy generated by iteration with section i-1 user
(6.2) if the time cost is lower, updating the allocation strategy of the i-1 st segment, otherwise not following the update
And (6.3) if i is less than 5, returning to the step (2), not recalculating the initial strategy, taking the distribution strategy generated in the previous step as the initial distribution strategy, and entering the iteration of the next section of i +1 users until 5 sections of users finish the loop.
Examples
The present invention is further explained below with reference to simulation results.
Through experimental simulation, the final total time cost of each distribution method is counted by comparing the distribution strategies formulated by the random distribution users and the greedy method with the conventional VSVSVPP algorithm.
As shown in fig. 4, in this embodiment, the number of edge servers is set to be 100, the number of users is continuously increased under the condition that the resource amount is not changed, along with the increase of users, the scale of users is larger and larger, the resources of the edge servers are more and more limited relative to the users, and the difficulty of making a good allocation strategy is also larger and larger.
As shown in fig. 6, in this embodiment, the number of users is set to 50, 100, 200, 400, and 800, and then a certain number of edge servers are proportionally allocated to each group of users, the total resources of all the servers are enough to be allocated, but the resources are dispersed to each server, so that the difficulty of the allocation policy of the users is increased.
By integrating the simulation results and analysis, the invention can ensure that a distribution strategy with low user time cost is formulated in the edge computing environment with limited resources, and is also suitable for the situation with rich resources; the invention thoroughly solves the problem of high time cost caused by unreasonable allocation of the user to the edge server in the edge environment; in addition, the invention can be better applied in practice because the invention can be adapted to the actual non-linear user allocation strategy formulation.

Claims (6)

1. A low-time-cost user allocation method in a marginal environment is characterized in that: the method comprises the following steps:
(1) obtaining a user resource request, residual resources of each edge server and longitude and latitude of each user and each edge server to obtain a user set U ═ U1，…，umAnd the set of edge servers S ═ S1，…，snAnd then, ordering the users according to the size of the request resource, dividing the user set into L sections with equal quantity and size according to the size of the resource, wherein each section of users is Ul；
Figure FDA0003354776470000011
User resource request is represented as u by a vectori＝<vcpu，memory，…，weight>；
Edge server sjIs represented by a vector
Figure FDA0003354776470000012
Wherein n is the total number of users requesting to calculate resources, m is the total number of edge servers, and vcpu represents the number of cpu micro-cores; memory represents a memory; weight represents the workload of the user; range represents the effective coverage area of the edge server signal;
Figure FDA0003354776470000013
representing edge servers sjThe cpu performance of the user u, the geographical distance between each user and each edge server is calculated through longitude and latitude informationiAnd edge server sjFor the geographical distance dijIs represented by dijWhen the coverage area is less than or equal to s, the method represents the range of the user in the edge server sWithin the enclosure at sjIn case the remaining resources of (1) are sufficient, user uiCan be distributed to edge servers sj；
(2) Initializing iteration parameters, including an initial temperature INIT _ TEM, a temperature decay RATE RATE, a termination temperature FINNAL _ TEM, an inner circulation time IN _ LOOP, a continuous non-updating current maximum solution time LIMIT and an outer circulation time OUT _ LOOP;
(3) initial allocation strategy a ═ { a ═ aij(ii) a i belongs to {1, …, n }, j belongs to {1, …, m } }, the initialized A is assigned to the initial allocation strategy of the first outer loop, each section of user is subjected to segmentation iteration, namely, before the first section of user is iterated, an initial allocation strategy is generated first, and the initial allocation strategy of each section of user iteration is the optimal allocation strategy generated after the previous section of user iteration;
in the process, a user needs to be within the coverage range of the edge server, the residual resources of the edge server can meet the requirements of the user, then an initial allocation strategy is generated according to the processing speed of the edge server, the alternative edge servers of each user are sorted according to the processing speed, an unallocated user is randomly selected, the user is allocated to the edge server with the highest processing speed in the alternative edge servers, if the processing speed of the edge server still cannot meet the resource constraint, the next alternative edge server is judged until the user is allocated, and if the alternative edge server which does not meet the resource constraint finally, the next alternative edge server is allocated to the cloud end;
(4) generating a new allocation strategy and calculating the total time cost T of the current allocation strategytotle
Randomly selecting a user to re-formulate an allocation strategy on the basis of the current allocation strategy, judging whether the adjusted server can be allocated to the user without the allocated server when the allocation strategy is re-formulated after the operation is finished, and simultaneously judging whether the changed user can be continuously allocated to other edge servers covering the user;
wherein, if the user is assigned to the edge server, the time cost is
Figure FDA0003354776470000021
Namely, it is
Figure FDA0003354776470000022
Denoted as user uiAt edge server sjThe time cost of;
if the user does not have the assignable edge server, the user is assigned to the cloud end to execute the task, and the time cost is
Figure FDA0003354776470000023
Figure FDA0003354776470000024
Figure FDA0003354776470000025
CapacityiRepresenting a user UiThe amount of computing power provided by the resources in demand,
Figure FDA0003354776470000026
Figure FDA0003354776470000027
representative user uiD represents the total number of resource types, k is {1, …, D };
(5) updating allocation decisions based on changes in time cost
Computing a new classification policy AnewTime cost of
Figure FDA0003354776470000028
And with the current allocation policy A generating a new allocation decisioncurrentTotal time cost of
Figure FDA0003354776470000029
Comparing, wherein the time cost is low, the time cost is high when replacing, if the time cost is equal or the time cost of the new distribution strategy is higher, whether the distribution strategy is updated or not is selected according to the corresponding probability P;
Figure FDA00033547764700000210
Figure FDA00033547764700000211
Figure FDA00033547764700000212
where delete is the difference between the time cost of the newly generated allocation policy and the time cost of the current allocation policy,
Figure FDA00033547764700000213
representing the total time cost of the newly generated allocation policy,
Figure FDA00033547764700000214
represents the total time cost of the current allocation policy;
if delete ≦ 0 and allocationcur≤allocationnewThen the current allocation policy is updated with the newly generated allocation policy, wherein allocationcurIs the total number of users allocated by the current allocation policy, allocationnewRepresenting the total number of users allocated by the newly generated allocation policy, otherwise, randomly generating an r e [0,1 ∈]If r is greater than or equal to P, the condition of delete is less than or equal to a and allocation is judgednew≥allocationcur-b, updating the current allocation policy if the condition holds;
a represents the difference value of the total time cost of all users generated by the newly generated allocation strategy and the total time cost of all users generated by the current allocation strategy, and b represents the maximum value that the total number of the users allocated by the newly generated allocation strategy is less than the total number of the users allocated by the current allocation strategy;
the above condition is described as if the newly generated allocation policy is worse than the current policy, then if
Figure FDA0003354776470000031
Not exceeding
Figure FDA0003354776470000032
And allocationnewMaximum ratio allocationcurB less;
(6) judging whether the circulation process meets the termination condition
The end conditions of the cycle are controlled by six parameters, INIT _ TEM, RATE, FINNAL _ TEM, IN _ LOOP, LIMIT and OUT _ LOOP; judging whether to accept the poor distribution decision through the probability P, and finally outputting the optimal distribution strategy;
Figure FDA0003354776470000033
(7) after the iteration of each section of user is finished, judging the advantages and disadvantages of the distribution strategy generated by the iteration of the previous section of user and the distribution strategy generated by the current section of user, and taking the more optimal distribution strategy as the initial distribution strategy of the iteration of the next section of user.
2. The method for low-time-cost customer distribution in a fringe environment of claim 1, further comprising: the step (1) is that the edge server SjIs calculated at a speed of
Figure FDA0003354776470000034
The computing speed of the cloud server is fcloud，
Figure FDA0003354776470000035
3. The method for low-time-cost customer distribution in a fringe environment of claim 1, further comprising: a in the step (3)ijRepresenting user uiThe value of the distribution strategy is 0 or 1, wherein 0 represents that the user is distributed to the cloud end, and 1 represents that the user is distributed to the edge server Sj(ii) a Namely:
Figure FDA0003354776470000036
the initial allocation policy must satisfy the constraint:
Figure FDA0003354776470000037
that is, the edge server and the users are in a one-to-many relationship, one edge server can be allocated with a plurality of users, and one user can be allocated to only one edge server;
corresponding coverage constraint and resource limitation constraint must be satisfied in the process of allocating servers by a user, only an edge server which covers the user and has enough resources can become an alternative server of the user, and a final allocation strategy must satisfy that the sum of the resource requirements of each type of the users allocated to the edge server is smaller than the size of the residual resources of the edge server;
wherein the coverage constraint is:
Figure FDA0003354776470000041
the resource limitation constraint is:
Figure FDA0003354776470000042
dijrepresenting user uiAnd edge server sjThe geographical position distance adopts the longitude and latitude distance between the user and the edge server; rangejRepresenting the signal coverage of an edge server j, wherein n represents the total number of users applying for computing resources, and m represents the total number of edge servers;
Figure FDA0003354776470000043
representing user uiThe requirements of the kth class of resources of (1),
Figure FDA0003354776470000044
representing edge servers sjThe remaining amount of the kth class resource of (1), all assigned to the edge server sjThe sum of the various resource requirements of the user is less than that of the edge server sjThe resource types of the resource are D types.
4. The method for low-time-cost customer distribution in a fringe environment of claim 1, further comprising: the specific process of generating the new allocation strategy in the step (4) is as follows:
randomly selecting a user, and then changing and adjusting the user UselectIf the other alternative edge servers of the user meet the constraint conditions, the user is allocated to a certain alternative server, otherwise, an alternative server is randomly selected, the user allocated on the server is allocated to the other alternative servers until the server meets the constraint conditions for allocating the selected user, the selected user is allocated to the server, and whether the original edge server of the selected user can reallocate other unallocated users is judged;
wherein, Uselect＝Random({ui，i∈[1，n]})；
YselectFor user uiThe number of alternative servers of (a); then:
Figure FDA0003354776470000045
5. the method for low-time-cost customer distribution in a fringe environment of claim 1, further comprising: generating a new distribution strategy and judging whether updating the current distribution strategy is executed through an inner LOOP, wherein the LIMIT and the OUT _ LOOP return to the initial value when the new distribution strategy directly replaces the current strategy, the LIMIT is reduced by 1 when the current distribution strategy is not directly replaced, the OUT _ LOOP is reduced by 1 when the LIMIT is equal to 0, the IN _ LOOP is reduced by 1 when the inner LOOP is performed each time, the initial temperature of the outer LOOP is reduced by INIT _ TEM multiplied by RATE when the INIT _ TEM is not greater than the INIT _ TEM, and the outer LOOP is not ended until the INIT _ TEM is not greater than the FINNA _ TEM or the number of outer LOOPs is used up;
wherein, whether to terminate the circulation is judged by adopting the following formula;
Figure FDA0003354776470000051
6. the method for low-time-cost customer distribution in a fringe environment of claim 1, further comprising: in the step (7), after the outer loop is finished each time, the advantages and disadvantages of the optimal allocation strategy at the beginning of the outer loop and the allocation strategy generated when the inner loop is finished at this time need to be judged, so that the optimal allocation strategy becomes the optimal allocation strategy of the outer loop at the next time better, iteration is performed again, the inner loop is updated until the outer loop is finished, and the optimal allocation strategy is finally output; the best allocation strategy after each segmentation iteration is the initial allocation strategy of the next segmentation iteration; where he decides whether the allocation strategy is good or bad by the overall time cost.
CN202111347569.1A 2021-11-15 2021-11-15 User allocation method with low time cost in edge environment Active CN113923223B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111347569.1A CN113923223B (en) 2021-11-15 2021-11-15 User allocation method with low time cost in edge environment

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111347569.1A CN113923223B (en) 2021-11-15 2021-11-15 User allocation method with low time cost in edge environment

Publications (2)

Publication Number Publication Date
CN113923223A true CN113923223A (en) 2022-01-11
CN113923223B CN113923223B (en) 2024-02-06

Family

ID=79246551

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111347569.1A Active CN113923223B (en) 2021-11-15 2021-11-15 User allocation method with low time cost in edge environment

Country Status (1)

Country Link
CN (1) CN113923223B (en)

Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110740168A (en) * 2019-09-24 2020-01-31 安徽大学 Self-adaptive method for multi-tenant server in cloud
CN110941667A (en) * 2019-11-07 2020-03-31 北京科技大学 Method and system for calculating and unloading in mobile edge calculation network
CN111240701A (en) * 2019-12-31 2020-06-05 重庆大学 Task unloading optimization method for end-edge-cloud collaborative computing
CN111611076A (en) * 2020-05-20 2020-09-01 云南大学 Fair distribution method for mobile edge computing shared resources under task deployment constraint
CN112105062A (en) * 2020-09-16 2020-12-18 云南大学 Mobile edge computing network energy consumption minimization strategy method under time-sensitive condition
CN112689303A (en) * 2020-12-28 2021-04-20 西安电子科技大学 Edge cloud cooperative resource joint allocation method, system and application
US20210176327A1 (en) * 2017-11-06 2021-06-10 Athonet S.R.L. Policy-driven local offload of selected user data traffic at a mobile edge computing platform
CN113032149A (en) * 2021-03-25 2021-06-25 中山大学 Edge computing service placement and request distribution method and system based on evolutionary game
CN113556764A (en) * 2021-07-30 2021-10-26 云南大学 Method and system for determining calculation rate based on mobile edge calculation network

Patent Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20210176327A1 (en) * 2017-11-06 2021-06-10 Athonet S.R.L. Policy-driven local offload of selected user data traffic at a mobile edge computing platform
CN110740168A (en) * 2019-09-24 2020-01-31 安徽大学 Self-adaptive method for multi-tenant server in cloud
CN110941667A (en) * 2019-11-07 2020-03-31 北京科技大学 Method and system for calculating and unloading in mobile edge calculation network
CN111240701A (en) * 2019-12-31 2020-06-05 重庆大学 Task unloading optimization method for end-edge-cloud collaborative computing
CN111611076A (en) * 2020-05-20 2020-09-01 云南大学 Fair distribution method for mobile edge computing shared resources under task deployment constraint
CN112105062A (en) * 2020-09-16 2020-12-18 云南大学 Mobile edge computing network energy consumption minimization strategy method under time-sensitive condition
CN112689303A (en) * 2020-12-28 2021-04-20 西安电子科技大学 Edge cloud cooperative resource joint allocation method, system and application
CN113032149A (en) * 2021-03-25 2021-06-25 中山大学 Edge computing service placement and request distribution method and system based on evolutionary game
CN113556764A (en) * 2021-07-30 2021-10-26 云南大学 Method and system for determining calculation rate based on mobile edge calculation network

Non-Patent Citations (5)

* Cited by examiner, † Cited by third party
Title
FUTIAN WANG; TAO HONG; DECHUN WANG; CHENG ZHANG: "A Probability-Based Fault Tolerance Strategy For Service-Based Systems", 《2021 3RD INTERNATIONAL CONFERENCE ON ADVANCES IN COMPUTER TECHNOLOGY, INFORMATION SCIENCE AND COMMUNICATION (CTISC)》 *
QINGLAN PENG; YUNNI XIA; YAN WANG; CHUNRONG WU; WANBO ZHENG; XIN LUO; SHANCHEN PANZ; YONG MA; CHUNXU JIANG: "A Decentralized Collaborative Approach to Online Edge User Allocation in Edge Computing Environments", 《2020 IEEE INTERNATIONAL CONFERENCE ON WEB SERVICES (ICWS)》 *
徐佳;李学俊;丁瑞苗;刘晓;: "移动边缘计算中能耗优化的多重资源计算卸载策略", 计算机集成制造系统, no. 04 *
李学俊; 徐佳; 王福田; 朱二周; 吴蕾: "云工作流系统中能耗感知的任务调度算法", 《模式识别与人工智能》 *
黄冬艳;付中卫;王波;: "计算资源受限的移动边缘计算服务器收益优化策略", 计算机应用, no. 03 *

Also Published As

Publication number Publication date
CN113923223B (en) 2024-02-06

Similar Documents

Publication Publication Date Title
Ouyang et al. Adaptive user-managed service placement for mobile edge computing: An online learning approach
CN112020103B (en) Content cache deployment method in mobile edge cloud
Tran et al. COSTA: Cost-aware service caching and task offloading assignment in mobile-edge computing
CN108076486A (en) A kind of dynamic radio resource allocation algorithm based on load balancing
US11184794B2 (en) Systems and methods for distribution of application logic in digital networks
US11856246B2 (en) CDN optimization platform
CN105703927A (en) Resource allocation method, network device and network system
CN111614754A (en) Fog-calculation-oriented cost-efficiency optimized dynamic self-adaptive task scheduling method
CN111901145A (en) Power Internet of things heterogeneous shared resource allocation system and method
CN113300861B (en) Network slice configuration method, device and storage medium
CN114205317B (en) SDN and NFV-based service function chain SFC resource allocation method and electronic equipment
Seid et al. Blockchain-empowered resource allocation in Multi-UAV-enabled 5G-RAN: a multi-agent deep reinforcement learning approach
CN113010317B (en) Combined service deployment and task offloading method and device, computer equipment and medium
CN113923223A (en) User allocation method with low time cost in edge environment
CN108833295B (en) SDN-oriented virtual network reconfiguration method based on tabu search
CN113301583B (en) Network slice remapping method, device and storage medium
Krolikowski et al. Fair distributed user-traffic association in cache equipped cellular networks
Li et al. Handoff control and resource allocation for ran slicing in iot based on dtn: An improved algorithm based on actor–critic framework
CN114222371A (en) Flow scheduling method for coexistence of eMBB (enhanced multimedia broadcast/multicast service) and uRLLC (unified radio link control) equipment
Guan et al. Multidimensional resource fragmentation-aware virtual network embedding for IoT applications in MEC networks
CN115051999B (en) Energy consumption optimal task unloading method, device and system based on cloud edge cooperation
Li et al. Adaptive handover based on traffic balancing and multi-dimensional collaborative resource management in MEC environment
CN110380808A (en) Half sub-clustering disturbance coordination method of Microcell centered on user equipment
CN113630792B (en) Traffic load balancing breadth-first search optimization method, system and equipment
JP2020014111A (en) Cooperative virtual network allocation method and apparatus

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