CN112148492A - Service deployment and resource allocation method considering multi-user mobility - Google Patents

Service deployment and resource allocation method considering multi-user mobility Download PDF

Info

Publication number
CN112148492A
CN112148492A CN202011038113.2A CN202011038113A CN112148492A CN 112148492 A CN112148492 A CN 112148492A CN 202011038113 A CN202011038113 A CN 202011038113A CN 112148492 A CN112148492 A CN 112148492A
Authority
CN
China
Prior art keywords
service
user
deployment
computing
overhead
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
CN202011038113.2A
Other languages
Chinese (zh)
Other versions
CN112148492B (en
Inventor
陈智麒
张胜
钱柱中
李文中
陆桑璐
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.)
Nanjing University
Original Assignee
Nanjing 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 Nanjing University filed Critical Nanjing University
Priority to CN202011038113.2A priority Critical patent/CN112148492B/en
Publication of CN112148492A publication Critical patent/CN112148492A/en
Application granted granted Critical
Publication of CN112148492B publication Critical patent/CN112148492B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00—Arrangements for program control, e.g. control units
    • G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/46—Multiprogramming arrangements
    • G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
    • G06F9/5005—Allocation of resources, e.g. of the central processing unit [CPU] to service a request
    • G06F9/5027—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resource being a machine, e.g. CPUs, Servers, Terminals
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2111/00—Details relating to CAD techniques
    • G06F2111/04—Constraint-based CAD
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2111/00—Details relating to CAD techniques
    • G06F2111/10—Numerical modelling
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D30/00—Reducing energy consumption in communication networks
    • Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks

Abstract

The invention provides a service deployment and resource allocation method considering multi-user mobility, which is applied to an edge computing network scene. The method is characterized in that modeling is carried out according to an edge computing scene, the edge computing scene is regarded as an optimization problem which integrates service computing time delay expense, transmission time delay expense and service migration expense, decision constraint is combined, and the optimization problem is solved to obtain a multi-user service deployment and computing resource allocation scheme. The invention fills the blank of the field, supports the multi-user service deployment, considers the mobility of the user, has wide applicability, and improves the task allocation and execution efficiency under the edge computing scene, thereby improving the overall processing performance of the network.

Description

Service deployment and resource allocation method considering multi-user mobility
Technical Field
The invention relates to the field of edge computing, in particular to a service deployment and resource allocation method considering multi-user mobility under an edge computing network environment.
Background
The last decade is a decade of rapid development of cloud computing technology, and cloud computing ensures dynamic network resource pool, virtualization and high availability for users, so that the internet becomes a data and computing center of users, a large amount of services can be performed in the cloud, and the cloud computing becomes an important development direction of current information technology. However, everything has two sides, and as people's knowledge of cloud computing is continuously deepened, its disadvantages are revealed: (1) privacy security issues: since users need to provide their own information on the cloud computing platform to use the service, the platform risks revealing privacy of these users. (2) Data transmission cost problem: the development of the internet of things enables various novel intelligent devices to emerge continuously, and the intelligent devices generate a large amount of data which far exceeds the bearing capacity of a network and a cloud computing center. For example, surveillance cameras that are now in view generate large amounts of video data each day; for the automatic driving car, data of up to 5TB is generated every day, and it is hard to transmit all the data to the cloud for processing under the current technical conditions. (3) The problem of high latency: due to the centralized characteristic of cloud computing, part of nodes are necessarily far away from a computing center, and the real-time performance of the nodes cannot be guaranteed. However, it is known that autodrive and industrial automation have high real-time requirements for data processing, and if network delay caused by data transmission cannot meet the real-time requirements, catastrophic consequences are likely to be caused.
For the existing problems of cloud Computing, Mobile Edge Computing (MEC) comes along. The mobile edge computing overcomes a series of defects of a traditional cloud computing architecture by pushing resources such as computing, storage and the like to the edge of a network, and provides a network computing mode with low time delay, high privacy and safety, high expandability, rapidness and high efficiency. In a typical moving edge computing scenario, there are three computing devices: cloud computing node, edge computing node (if deploy on edge server), user equipment node, and the three have the following advantages:
cloud computing node: generally, computing resources of cloud computing nodes are not limited, and any task can be processed by high-speed computing power after being submitted to a cloud, but the disadvantage is that communication cost for uploading local data to the cloud is high. Representative public Cloud services are AWS, Google Cloud, Microsoft Azure, and the like.
A user equipment node: these user devices may be smart mobile phones, smart car systems, embedded internet of things devices, etc. that are held by people and can be viewed as the origin of data or close to the origin itself, so that they have little data communication cost, but are not usually equipped with high computing performance in consideration of device hardware cost and long-term endurance.
Edge computing node: the above two trade-offs can be considered, there is a certain computational processing power, and the data communication cost is within an acceptable range. These edge computing nodes are often deployed on some Access Points (APs) of a mobile network, and can provide edge computing services for mobile devices within the coverage of the APs. Edge compute nodes may be considered extensions of cloud compute nodes, but are subject to performance and cost considerations of servers, providing limited computing resources. When facing a plurality of services, it is assumed that the edge computing node can allocate the CPU computing resource proportion, and different services occupy different CPU proportions to generate a certain service computing delay.
But in a typical mobile edge computing environment, it is often desirable to take into account the mobility of the user. During a certain period of time, a user may roam from one location to another, and accordingly the data latency of the cloud computing node and the user equipment node may not change due to a change in network location, but the latency to the respective edge computing nodes may change. The edge node providing the computing service needs to consider whether to continue running the service at the edge node or to migrate the service to another edge computing node with lower latency to provide the service. The contradiction here is that the former is only the increase of data communication delay, while the latter can effectively reduce the communication delay, but brings extra migration overhead, where the migration overhead includes the overhead of destroying and recreating the virtual machine, the bandwidth occupation of migrating the virtual machine state, and even the overhead of migration failure. However, according to the work of the current signature inventors, there are many disadvantages in the current research, such as considering only a single-user service decision deployment scheme and only coarse-grained allocation of computing resources, which cannot satisfy the high-performance allocation and execution of tasks in the case of multi-user mobility.
Disclosure of Invention
The purpose of the invention is as follows: aiming at the blank of the prior art, the invention provides a service deployment and resource allocation method considering multi-user mobility in an edge computing network environment, which realizes the comprehensive consideration of how to deploy user services to edge computing nodes and how to allocate CPU occupation rates to the services by the edge computing nodes.
The technical scheme is as follows: a service deployment and resource allocation method considering multi-user mobility comprises the following steps:
(1) establishing a mathematical model for a moving edge calculation scene:
computing cloud nodes under scene according to mobile edges
Figure BDA0002705772530000021
User equipment node
Figure BDA0002705772530000022
And (4) obtaining a computing node of the model by setting the edge computing node set as {1,2, …, E }, wherein
Figure BDA0002705772530000023
The decision to deploy a service is made between a discrete time slice, and the model time is defined as a set of discrete time slices
Figure BDA0002705772530000024
And analyzing a decision space formed by the service deployment and resource allocation problems: when the time slice t comes, useThe service of user u needs to decide on which computing node to deploy; and for each computing node, if more than 1 user service deployment exists in the time slice t, the CPU computing resource distribution ratio among the services needs to be considered;
three overheads contained in the scene are calculated from the moving edges: calculating time delay overhead, transmission time delay overhead and service migration overhead, and determining the optimization target of the model as follows: minimizing the weighted sum of the three overheads, wherein the calculation delay overheads refer to the time delay from the time when the service request of the user reaches the calculation node to the time when the calculation of the request is completed; the transmission delay overhead refers to the delay from the time when the service request of the user is sent by the user to the time when the computing node receives the request; the service migration overhead refers to the overhead generated by the corresponding migration of the service after the user moves;
(2) and solving the established mathematical model to obtain a multi-user service deployment and resource allocation scheme.
Further, the optimization objective of the model is represented as:
Figure BDA0002705772530000031
wherein the content of the first and second substances,
Figure BDA0002705772530000032
is a two-dimensional indicator variable that indicates whether, at time slice t, a user service j is deployed to compute node i,
Figure BDA0002705772530000033
represents deployment, and
Figure BDA0002705772530000034
on the contrary;
Figure BDA0002705772530000035
indicating the CPU computing resource allocation ratio,
Figure BDA0002705772530000036
is a number from 0 to 1An inter-rate variable representing the rate at which user service j is allocated computing resources at computing node i at time slice t;
Figure BDA0002705772530000037
represents the computation delay overhead of user service j at time slice t:
Figure BDA0002705772530000038
representing the transmission delay overhead of the user service j in the time slice t;
Figure BDA0002705772530000039
represents the migration overhead of user service j in time slice t; w is a1,w2,w3Respectively, the weights of the three overheads.
Further, the step (2) of solving the established mathematical model includes:
note the book
Figure BDA00027057725300000310
Is X, any one X' is belonged to X, and has a unique corresponding optimal solution y*Defining a service deployment and computing resource allocation state as s ═ x', y*) E S, the total cost of the state is Cs'→sThe added overhead is needed to migrate from the state S' of the last time slice to the state S of the current time slice, where S refers to a feasible solution set of the service deployment and the computing resource allocation scheme;
the construction graph G is (V, L), the vertex set V represents a set of states, the edge set L represents a set of total expenses between two states, and particularly two adjacent states
Figure BDA00027057725300000311
And
Figure BDA00027057725300000312
the weight of the edge between represents the service migration cost involved from the state i to the state j, and the service calculation delay and transmission delay cost in the time slice t;
adding artificial nodes S and D on two sides of the graph G, and solving the shortest path between S and D through the following algorithm:
A. receiving input and initializing a dynamic programming state table phi, phi recording a deployment scenario s from a certain deployment scenariotA mapping to corresponding overhead;
B. continuously cycling from T to T;
C. obtaining all (E +2) by Cartesian product for each cycleURecording a possible deployment scheme as a set H, wherein E is the number of edge computing nodes, and U is the number of user services;
D. traversing the deployment schemes in each set H, and updating phi into one item with the minimum sum of the current overhead and the overhead of all the schemes in the last time slice t-1;
E. and returning one deployment scheme with the lowest cost in all the deployment schemes of phi.
As a more preferred embodiment, the step (2) of solving the established mathematical model includes:
note the book
Figure BDA0002705772530000041
Is X, any one X' is belonged to X, and has a unique corresponding optimal solution y*Defining a service deployment and computing resource allocation state as s ═ x', y*) E S, the total cost of the state is Cs'→sThe added overhead is needed to migrate from the state S' of the last time slice to the state S of the current time slice, where S refers to a feasible solution set of the service deployment and the computing resource allocation scheme;
the construction graph G is (V, L), the vertex set V represents a set of states, the edge set L represents a set of total expenses between two states, and particularly two adjacent states
Figure BDA0002705772530000042
And
Figure BDA0002705772530000043
the weight of the edge between represents the service migration overhead involved from state i to state j, such thatCalculating the time delay and the transmission time delay overhead in the service of the time slice t;
adding artificial nodes S and D on two sides of the graph G, and solving the shortest path between S and D through the following algorithm:
a. receiving input and initializing Φ recording a deployment scenario s fromtA mapping to corresponding overhead;
b. continuously cycling from T to T;
c. all (E +2) are obtained in each cycleURecording a possible deployment scheme as a set H, wherein E is the number of edge computing nodes, and U is the number of user services;
d. obtaining kappa schemes from the deployment schemes in each set H through uniform sampling, and updating phi into one item with the minimum sum of the overhead of all the sampling schemes in the last time slice t-1 and the overhead of the current sampling scheme;
e. and returning one deployment scheme with the lowest cost in all the deployment schemes of phi.
As a more preferred embodiment, the step (2) of solving the established mathematical model includes:
recording the service set deployed on the edge computing node i as lambdai＝SUBSET({λ1,λ2,…,λn})，λnDenotes the nth service by wijRepresents the cost of the transmission delay and the service migration for deploying the user service j to the edge computing node i, and uses vij＝λj/ciTo represent the computational latency overhead of deploying a user service j onto an edge compute node i, ciComputing the computing power of node i for the edge;
deployment in arbitrary order { λ1,λ2,…,λnH, will bejDeploying to all edge computing nodes, aiming at the ith computing nodeiDefinition of LOAD: (i) Meaning the sum of the costs made for the user services already carried by the current node, defining a LOAD (a), (b), (c), (di,λj) Is represented iniThe total expense generated by deploying one user service j on the basis of the existing service;
and executing the following updating algorithm, and selecting the deployment with the minimum final cost increase:
1) computing nodes for each edgeiSetting LOAD: (i)＝0；
2) When the time slice t comes, the following steps are executed:
i. for each customer service j, calculate LOAD: (i,λj) And selecting the edge computing node with the minimum cost;
deploy customer service j on the edge compute node and then use the LOAD (, λ)j) LOAD () is updated.
Has the advantages that: aiming at the problem of multi-user service deployment and computing resource allocation under the edge computing environment, the invention carries out modeling according to an edge computing scene, regards the edge computing scene as an optimization problem integrating service computing time delay overhead, transmission time delay overhead and service migration overhead, combines decision constraint, and solves the optimization problem to obtain a multi-user service deployment and computing resource allocation scheme. The invention fills the blank of the field, supports the multi-user service deployment, considers the mobility of the user, has wide applicability, and improves the task allocation and execution efficiency under the edge computing scene, thereby improving the overall processing performance of the network.
Drawings
FIG. 1 is a schematic diagram of a service deployment and computing resource allocation scenario provided by an embodiment of the present invention;
fig. 2 is an exemplary diagram of a dynamic planning algorithm of an offline algorithm FDSP according to an embodiment of the present invention;
fig. 3 is an exemplary diagram of an embodiment of the present invention providing an online algorithm OSP.
Detailed Description
The technical solution of the present invention is further explained below with reference to the accompanying drawings.
Referring to the edge computing scenario of fig. 1, in one embodiment, a service deployment and resource allocation method considering multi-user mobility includes the following steps:
step (1), establishing a mathematical model for the service deployment and resource allocation problem according to the mobile edge computing scene.
In a mobile edge computing scenario, related devices including cloud computing nodes are also called cloud servers
Figure BDA0002705772530000061
User equipment node
Figure BDA0002705772530000062
The edge compute node is also referred to as the set of edge servers {1,2, …, E }. All three computing devices can be regarded as so-called computing nodes
Figure BDA0002705772530000063
To facilitate modeling user movement, and service deployment decisions are made between a discrete time slice, time is thus defined as a collection of discrete time slices
Figure BDA0002705772530000064
Two points need to be considered for analyzing a decision space formed by service deployment and resource allocation problems: first, when a time slice t comes, the service of a user u needs to decide on which computing node to deploy; and a second point, for each compute node, if there are more than 1 user service deployment at time slice t, then the CPU computation resource allocation ratio between these services needs to be considered.
For the first point, firstly, the user equipment cannot be too far away from the edge computing node running the service, otherwise, too high network transmission delay is generated; then, the user service is not deployed on the same edge computing node as much as possible, otherwise, due to the preemption of the same server resource by a plurality of services, the service computing time delay is obviously increased; in addition, because the edge computing nodes have heterogeneity, different nodes can provide different computing capabilities when facing the same user service according to the current machine state, such as the configuration of hardware and software of a CPU main frequency, a memory frequency, an instruction system and the like. The invention uses decision variables
Figure BDA0002705772530000065
To characterize a service deployment scenario in which,
Figure BDA0002705772530000066
is an 0/1 indicator variable that indicates whether, at time slice t, user service j was deployed to compute node i,
Figure BDA0002705772530000067
represents deployment, and
Figure BDA0002705772530000068
the opposite is true. Note that in time slice t, one service can and can only be deployed on one compute node, so there are the following decision constraints:
Figure BDA0002705772530000069
Figure BDA00027057725300000610
for the second point, a simplest allocation manner is to allocate CPU computing resources uniformly, and assuming that a certain node provides a computing power with a size of c (e.g. using cycles per minute of CPU to characterize), and n user services have been deployed on the node, according to the principle of uniform allocation, each service obtains a computing power of c
Figure BDA00027057725300000611
For the ith service, the service calculation delay can be used
Figure BDA00027057725300000612
To indicate. It is easy to find that this distribution method following the uniform principle is not an optimal distribution method, a better method is according to the data processing amount λ of the serviceiTo dynamically decide the CPU computing resource allocation rate of the service. The invention uses decision variables
Figure BDA00027057725300000613
To indicate that the user is not in a normal position,
Figure BDA00027057725300000614
is a ratio variable from 0 to 1 onwards that represents the ratio of computing resources that user service j has been allocated on compute node i during time slice t. Note that at time slice t, a compute node can provide a compute resource allocation ratio that sums up to 100%, so there are the following decision constraints:
Figure BDA0002705772530000071
Figure BDA0002705772530000072
Figure BDA0002705772530000073
wherein the last inequality is expressed as if
Figure BDA0002705772530000074
Indicating that in time slice t, user service j will not be deployed on compute node i, then there is naturally
Figure BDA0002705772530000075
If it is
Figure BDA0002705772530000076
User service j is deployed on computing node i, then
Figure BDA0002705772530000077
Any value between 0 and 1 may be taken.
The method mainly comprises the following steps of analyzing the processes from generation to execution of the service of a user in a mobile edge computing scene, wherein the processes mainly comprise three expenses: calculating time delay overhead, transmission time delay overhead and service migration overhead, wherein the calculated time delay overhead refers to the time delay from the time when a service request of a user arrives at a calculation node to the time when the calculation of the request is completed; the transmission delay overhead refers to the delay from the time when the service request of the user is sent by the user to the time when the computing node receives the request; the service migration overhead refers to the overhead generated by the corresponding migration of the service after the user moves. Then our goal is to minimize the weighted sum of these three costs. The optimization objectives and constraints of the mathematical model are given according to the above analysis of the decision space as follows:
in time slice t, the user has a service to calculate the demand amount and use according to the application demand of the user
Figure BDA0002705772530000078
Meaning specifically the computational demand of user service j at time slice t. It can be noted that the service calculation demand of the user changes with the time slice t, which is the result of the combination of factors of user mobility, service timeliness and user demand change, and it is assumed that the demand is directly known. Taking the video analysis service as an example,
Figure BDA0002705772530000079
the method can be obtained according to the size and frame rate of the input video, the precision requirement of the analysis task and the like. In addition, the present invention uses the symbol ciTo represent the computing power (e.g., CPU clock frequency) provided by the computing node i, it is noted here that for cloud computing nodes
Figure BDA00027057725300000710
And edge computing nodes, which provide no difference in computing power to different users, but the local computing power of each user device may be considered different because internet-of-things devices, such as embedded computing devices, themselves often provide different computing power due to hardware cost and power endurance, among other considerations. Given service deployment and computing resource allocation decisions in accordance with the above definitions
Figure BDA00027057725300000711
And
Figure BDA00027057725300000712
the calculation delay cost of the user service j in the time slice t can be obtained as follows:
Figure BDA00027057725300000713
the transmission delay is usually composed of an access delay and a propagation delay. The access delay refers to a delay from the user equipment to the nearby access point, which is generally determined by the wireless environment and the terminal equipment together. For propagation delay, if the user service is deployed locally, the propagation delay is almost negligible and can be regarded as 0; if the user service is deployed at the cloud end, the propagation delay from the user equipment to the cloud end can be considered as a constant, and the specific numerical value is determined by a cloud service provider; if the user service is deployed on an edge compute node, the delay is typically related to the edge-to-edge delay between the access point to the edge compute node where the service is located. In the most ideal case, the service is deployed directly on the access point, and the edge-to-edge propagation delay is negligible. Aiming at the transmission time delay, the invention is uniformly used
Figure BDA0002705772530000081
Meaning that the time delay from the service request of user j to compute node i at time slice t. Given a service deployment decision, according to the above definition
Figure BDA0002705772530000082
The transmission delay overhead of the user service j in the time slice t can be obtained as follows:
Figure BDA0002705772530000083
because the network environment is changed due to the movement of the user, the edge computing node closest to the user can also be usedCorrespondingly, when the user generates a new movement, the deployment location of the user service needs to be reconsidered: one way is to always run the service in the original location no matter how the user moves; another way is to migrate services as the user moves, which incurs service migration overhead. The service migration overhead includes the overhead of destroying and recreating the virtual machine, the bandwidth occupation overhead of migrating the virtual machine state, and even the failure overhead of migration failure. The invention uses symbols
Figure BDA0002705772530000084
And representing service migration overhead which represents migration overhead for migrating the user service j from the computing node i to i' in the time slice t, wherein the overhead is related to a specific user service type and a user service running state, can be obtained through measurement in practical application, and can also be assumed to be a constant for convenience. Note that when i is equal to i', i.e. no migration of the service has occurred, the service migration overhead can be considered to be 0. According to the above definition, it is easy to derive the service migration cost of the user service j at the time slice t:
Figure BDA0002705772530000085
wherein
Figure BDA0002705772530000086
Indicating whether, at time slice t-1, user service j is deployed onto compute node i,
Figure BDA0002705772530000087
indicating whether, at time slice t, user service j is deployed onto compute node i',
Figure BDA0002705772530000088
and
Figure BDA0002705772530000089
are all [000 … 1 … 000 ]]Such a vector, with only one node being 1, is the currently deployed node,so as to pass through
Figure BDA00027057725300000810
Service migration overhead from the previous node to the next node can be calculated. Note that i 'and i are irrelevant and are both used to traverse the set of compute nodes, with the difference that i is the compute node under the t-1 time slice, and i' is the compute node under the t time slice.
The delay, transmission delay overhead, and service migration overhead are calculated from the services discussed above, and the overall goal is to optimize these three overheads to minimize their overall overhead. One way is to give different weights to the three overheads and then sum them, and respectively record the weights of the three overheads as w1,w2,w3. Correspondingly, in time slice t, the total weighted sum cost of user service j is:
Figure BDA0002705772530000091
when given a time sequence
Figure BDA0002705772530000092
We wish to minimize the overall overhead for all users over time, and the overall problem can be formally expressed as:
Figure BDA0002705772530000093
the following table shows the meaning of the above symbols:
TABLE 1 symbolic meanings used in the mathematical model
Figure BDA0002705772530000094
Figure BDA0002705772530000101
And (2) solving the established mathematical model to obtain a multi-user service deployment and resource allocation scheme.
Aiming at the optimization problem established in the step (1), when the scale of the problem is small, the invention provides an FDSP (full Dynamic Service plan) algorithm based on Dynamic programming, which can obtain an accurate solution; when the problem scale becomes larger, the invention provides an improved SDSP (sampling Dynamic Service plan) algorithm, which can effectively relieve the problem of FDSP algorithm state number combination explosion based on the idea of sampling. Considering the unknown of the system future information in Service decision, the invention also provides an OSP (Online Service plan) online algorithm based on the greedy idea. Each is described in detail below.
FDSP is an accurate algorithm based on dynamic programming, and because an optimal calculation resource allocation scheme is uniquely determined by a service deployment scheme and a problem has a time series property, the FDSP can be used for dividing a plurality of stages for iterative calculation. The problem is converted into the problem of solving the shortest path between two artificial nodes by adding the two artificial nodes, and a state transition equation is given. In particular, a service deployment and computing resource allocation scheme may be considered as a state, and then by combining computations, all states total (E +2)UAnd (4) respectively. Suppose that
Figure BDA0002705772530000102
Is X, then arbitrarily take one X' epsilon X from the space, and have the only corresponding optimal solution y*Therefore, a service deployment and computing resource allocation state can be defined as s ═ x', y*) E S, where S refers to the set of feasible solutions for service deployment and computing resource allocation schemes. From the foregoing definitions of computation delay, transmission delay overhead, and service migration overhead for a service, the overall cost of this state can be defined as Cs'→sI.e. the increased overhead required to transition from the state s' of the last time slice to the state s of the current time slice. In fact, a graph G ═ V, L can be constructed where V is the set of vertices representing states, L is the set of edges between vertices representing the overhead between two states, and for a time slice t there is (E +2)UOne state represents all (E +2)UKind of service deployment scenario, E+2 is the number of all computing nodes, 2 is composed of cloud server and user equipment, and the state of two adjacent computing nodes
Figure BDA0002705772530000103
And
Figure BDA0002705772530000104
the weights of the edges in between represent the service migration overhead involved from state i to state j, as well as the service computation and transmission delay overhead at time slice t. By adding artificial nodes S and D on both sides of graph G as starting and ending points of the shortest path, the problem translates into finding the shortest path between S and D. In fact, the problem can be solved using dynamic programming, the heart of which is the state transition equation
Figure BDA0002705772530000105
Wherein
Figure BDA0002705772530000106
Which represents the cumulative minimum cost of picking state s in the last step in time slice t, and can be regarded as a sub-problem of the original problem.
The algorithm is based on a dynamic planning idea, and comprises the following specific steps:
A. receiving input includes computing power c of the computing nodeiServicing computing needs
Figure BDA0002705772530000111
Transmission time delay
Figure BDA0002705772530000112
Service migration overhead
Figure BDA0002705772530000113
Time slice number T, edge server number E and user service number U. Initializing a dynamic programming state table Φ, Φ records a deployment scenario s from a certain deployment scenariotA mapping to corresponding overhead;
B. continuously cycling from T to 1 to T, and performing C, D for two steps;
C. obtaining all (E +2) by Cartesian product for each cycleUA possible deployment scheme is marked as a set H;
D. traversing the deployment schemes in each set H, and updating phi into one item with the minimum sum of the current overhead and the overhead of all the schemes in the last time slice t-1;
E. and returning one deployment scheme with the lowest cost in all the deployment schemes of phi.
Taking fig. 2 as an example, it can be seen that T is 4, E is 2, and U is 2, and when T is 1, one can find
Figure BDA0002705772530000114
All 16 possible deployment scenarios to
Figure BDA0002705772530000115
For example, it represents a deployment scheme in which the user service 1 is deployed to the edge server 1 while the user service 2 is also deployed to the edge server 1. The optimal path can be found between S and D in the figure
Figure BDA0002705772530000116
And obtaining an optimal deployment scheme.
FDSP can obtain an accurate solution, but because the mode based on dynamic programming faces the problem of state number combination explosion, an off-line algorithm SDSP based on state sampling is provided. The algorithm is based on a dynamic programming idea with state sampling, and comprises the following specific steps:
a. receiving input includes computing power c of the computing nodeiServicing computing needs
Figure BDA0002705772530000117
Transmission time delay
Figure BDA0002705772530000118
Service migration overhead
Figure BDA0002705772530000119
Number of time slices T, number of edge servers E, user serviceThe number U. Initializing a dynamic programming state table Φ, Φ records a deployment scenario s from a certain deployment scenariotA mapping to corresponding overhead;
b. continuously cycling from T to 1 to T, and executing two steps of c and d;
c. all possible deployment schemes are obtained in each circulation and are marked as a set H;
d. from each set H (E +2)UAcquiring kappa schemes through uniform sampling in the deployment scheme, and updating phi into one item with the minimum sum of overhead of all sampling schemes in the last time slice t-1 and overhead of the current sampling scheme;
e. and returning one deployment scheme with the lowest cost in all the deployment schemes of phi.
Considering that the offline algorithm needs to know a period of time in advance
Figure BDA00027057725300001110
The invention converts the original continuous time optimization problem into a plurality of single step optimization problems and provides an online algorithm OSP to solve the problems.
For online problems, the last-time service deployment scenario x is knownt-1To obtain the service deployment scheme x under the current time slice tt. In the original modeling, the calculation delay overhead, the transmission delay overhead and the service migration overhead are calculated from each user, but in fact, the calculation delay overhead, the transmission delay overhead and the service migration overhead can be considered from a calculation node. For the edge computing node i, if the service set deployed by the edge computing node i is lambdai＝SUBSET({λ1,λ2,…,λn}) then its transmission delay overhead and service migration overhead are respectively
Figure BDA0002705772530000121
And
Figure BDA0002705772530000122
these two overheads are denoted here by wijShow to dress the userAnd the service j is deployed to the transmission delay overhead and the service migration overhead of the edge computing node i. Likewise, with vij＝λj/ciTo represent the computational latency overhead of deploying a user service j onto an edge compute node i. To this problem, the following greedy strategy may be used: deploy λ in arbitrary order1,λ2,…,λnCan try to get λjAnd deploying to all edge servers, and selecting deployment with minimum final cost increase. For each edge serveriDefinition of LOAD: (i) Meaning the sum of the costs made for the user services already carried by the current server, and the token LOAD: (i,λj) Is represented iniThe total cost generated by deploying one user service j on the basis of the existing service is increased.
There are the following algorithms:
1) for each edge serveriSetting LOAD: (i)＝0；
2) When the time slice t comes, the following steps are executed:
i. for each customer service j, calculate LOAD: (i,λj) And selecting the least costly;
deploy user service j on the edge server and then use LOAD (, λ)j) LOAD () is updated.
Taking FIG. 3 as an example, edge compute nodes1Has deployed { lambda1,λ2Two user services, then its existing overhead LOAD: (1)＝2(v1,1+v1,2)+w1,1+w1,2(ii) a Similarly, there is LOAD (2)＝v2,3+w2,3And LOAD (3)＝3(v3,4+v3,5+v3,6)+w3,4+w3,5+w3,6. Now consider λ7Is deployed to1,2,3On one of them, then first try to get λ7Deployed onto each, there are:
LOAD(1∪λ7)＝3(v1,1+v1,2+v1,7)+w1,1+w1,2+w1,7
LOAD(2∪λ7)＝2(v2,3+v2,7)+w2,3+w2,7
LOAD(3∪λ7)＝4(v3,4+v3,5+v3,6+v3,7)+w3,4+w3,5+w3,6+w3,7
and selecting the deployment with the minimum cost from the three schemes.
While the present invention has been described in detail with reference to the specific embodiments thereof, it will be apparent to those skilled in the art that various changes and modifications can be made therein without departing from the spirit and scope of the invention as defined by the appended claims.

Claims (6)

1. A service deployment and resource allocation method considering multi-user mobility, applied to an edge computing network environment, is characterized in that the method comprises the following steps:
(1) establishing a mathematical model for a moving edge calculation scene:
computing cloud nodes under scene according to mobile edges
Figure FDA00027057725200000110
User equipment node
Figure FDA00027057725200000111
And (4) obtaining a computing node of the model by setting the edge computing node set as {1,2, …, E }, wherein
Figure FDA00027057725200000112
The decision to deploy a service is made between a discrete time slice, and the model time is defined as a set of discrete time slices
Figure FDA00027057725200000113
And divides the decision space formed by the service deployment and resource allocation problemsAnd (3) analysis: when the time slice t comes, the service of the user u needs to decide on which computing node to deploy; and for each computing node, if more than 1 user service deployment exists in the time slice t, the CPU computing resource distribution ratio among the services needs to be considered;
three overheads contained in the scene are calculated from the moving edges: calculating time delay overhead, transmission time delay overhead and service migration overhead, and determining the optimization target of the model as follows: minimizing the weighted sum of the three overheads, wherein the calculation delay overheads refer to the time delay from the time when the service request of the user reaches the calculation node to the time when the calculation of the request is completed; the transmission delay overhead refers to the delay from the time when the service request of the user is sent by the user to the time when the computing node receives the request; the service migration overhead refers to the overhead generated by the corresponding migration of the service after the user moves;
(2) and solving the established mathematical model to obtain a multi-user service deployment and resource allocation scheme.
2. The method for multi-user mobility-aware service deployment and resource allocation according to claim 1, wherein the optimization objective of the model is expressed as:
Figure FDA0002705772520000011
Figure FDA0002705772520000012
Figure FDA0002705772520000013
Figure FDA0002705772520000014
wherein the content of the first and second substances,
Figure FDA0002705772520000015
is a two-dimensional indicator variable that indicates whether, at time slice t, a user service j is deployed to compute node i,
Figure FDA0002705772520000016
represents deployment, and
Figure FDA0002705772520000017
on the contrary;
Figure FDA0002705772520000018
indicating the CPU computing resource allocation ratio,
Figure FDA0002705772520000019
is a ratio variable between 0 and 1, which represents the ratio of the computing resources allocated by the user service j on the computing node i in the time slice t;
Figure FDA0002705772520000021
represents the computation delay overhead of user service j at time slice t:
Figure FDA0002705772520000022
representing the transmission delay overhead of the user service j in the time slice t;
Figure FDA0002705772520000023
represents the migration overhead of user service j in time slice t; w is a1,w2,w3Weights for three overheads respectively; t is the number of time slices.
3. The method for multi-user mobility-aware service deployment and resource allocation according to claim 2, wherein the method comprises
Figure FDA0002705772520000024
In a manner such asThe following:
Figure FDA0002705772520000025
wherein
Figure FDA0002705772520000026
Representing the computational demand of user service j, c, at time slice tiRepresenting the computing power provided by computing node i;
the above-mentioned
Figure FDA0002705772520000027
The calculation method is as follows:
Figure FDA0002705772520000028
wherein
Figure FDA0002705772520000029
Representing the time delay of the service request of the user j to the computing node i in the time slice t;
the above-mentioned
Figure FDA00027057725200000210
The calculation method is as follows:
Figure FDA00027057725200000211
wherein
Figure FDA00027057725200000212
Represents the migration overhead for migrating user service j from compute node i to i' at time slice t.
4. The method for service deployment and resource allocation considering multi-user mobility according to claim 2, wherein the step (2) of solving the established mathematical model comprises:
note the book
Figure FDA00027057725200000213
Is X, any one X' is belonged to X, and has a unique corresponding optimal solution y*Defining a service deployment and computing resource allocation state as s ═ x', y*) E S, the total cost of the state is Cs'→sThe added overhead is needed to migrate from the state S' of the last time slice to the state S of the current time slice, where S refers to a feasible solution set of the service deployment and the computing resource allocation scheme;
the construction graph G is (V, L), the vertex set V represents a set of states, the edge set L represents a set of total expenses between two states, and particularly two adjacent states
Figure FDA00027057725200000214
And
Figure FDA00027057725200000215
the weight of the edge between represents the service migration cost involved from the state i to the state j, and the service calculation delay and transmission delay cost in the time slice t;
adding artificial nodes S and D on two sides of the graph G, and solving the shortest path between S and D through the following algorithm:
A. receiving input and initializing a dynamic programming state table phi, phi recording a deployment scenario s from a certain deployment scenariotA mapping to corresponding overhead;
B. continuously cycling from T to 1 to T, and performing C, D for two steps;
C. obtaining all (E +2) by Cartesian product for each cycleURecording a possible deployment scheme as a set H, wherein E is the number of edge computing nodes, and U is the number of user services;
D. traversing the deployment schemes in each set H, and updating phi into one item with the minimum sum of the current overhead and the overhead of all the schemes in the last time slice t-1;
E. and returning one deployment scheme with the lowest cost in all the deployment schemes of phi.
5. The method for service deployment and resource allocation considering multi-user mobility according to claim 2, wherein the step (2) of solving the established mathematical model comprises:
note the book
Figure FDA0002705772520000031
Is X, any one X' is belonged to X, and has a unique corresponding optimal solution y*Defining a service deployment and computing resource allocation state as s ═ x', y*) E S, the total cost of the state is Cs'→sThe added overhead is needed to migrate from the state S' of the last time slice to the state S of the current time slice, where S refers to a feasible solution set of the service deployment and the computing resource allocation scheme;
the construction graph G is (V, L), the vertex set V represents a set of states, the edge set L represents a set of total expenses between two states, and particularly two adjacent states
Figure FDA0002705772520000032
And
Figure FDA0002705772520000033
the weight of the edge between represents the service migration cost involved from the state i to the state j, and the service calculation delay and transmission delay cost in the time slice t;
adding artificial nodes S and D on two sides of the graph G, and solving the shortest path between S and D through the following algorithm:
a. receiving input and initializing Φ recording a deployment scenario s fromtA mapping to corresponding overhead;
b. continuously cycling from T to 1 to T, and executing two steps of c and d;
c. all (E +2) are obtained in each cycleUAnd recording a possible deployment scheme as a set H, wherein E is the number of the edge computing nodes,U is the number of user services;
d. obtaining kappa schemes from the deployment schemes in each set H through uniform sampling, and updating phi into one item with the minimum sum of the overhead of all the sampling schemes in the last time slice t-1 and the overhead of the current sampling scheme;
e. and returning one deployment scheme with the lowest cost in all the deployment schemes of phi.
6. The method for service deployment and resource allocation considering multi-user mobility according to claim 2, wherein the step (2) of solving the established mathematical model comprises:
recording the service set deployed on the edge computing node i as lambdai＝SUBSET({λ1,λ2,…,λn})，λnDenotes the nth service by wijRepresents the cost of the transmission delay and the service migration for deploying the user service j to the edge computing node i, and uses vij＝λj/ciTo represent the computational latency overhead of deploying a user service j onto an edge compute node i, ciComputing the computing power of node i for the edge;
deployment in arbitrary order { λ1,λ2,…,λnH, will bejDeploying to all edge computing nodes, aiming at the ith edge computing nodeiDefinition of LOAD: (i) Meaning the sum of the costs made for the user services already carried by the current node, defining a LOAD (a), (b), (c), (di,λj) Is represented iniThe total expense generated by deploying one user service j on the basis of the existing service;
and executing the following updating algorithm, and selecting the deployment with the minimum final cost increase:
1) computing nodes for each edgeiSetting LOAD: (i)＝0；
2) When the time slice t comes, the following steps are executed:
i. for each customer service j, calculate LOAD: (i,λj) And selecting the edge computing node with the minimum cost;
deploy customer service j on the edge compute node and then use the LOAD (, λ)j) LOAD () is updated.
CN202011038113.2A 2020-09-28 2020-09-28 Service deployment and resource allocation method considering multi-user mobility Active CN112148492B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011038113.2A CN112148492B (en) 2020-09-28 2020-09-28 Service deployment and resource allocation method considering multi-user mobility

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011038113.2A CN112148492B (en) 2020-09-28 2020-09-28 Service deployment and resource allocation method considering multi-user mobility

Publications (2)

Publication Number Publication Date
CN112148492A true CN112148492A (en) 2020-12-29
CN112148492B CN112148492B (en) 2023-07-28

Family

ID=73895122

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011038113.2A Active CN112148492B (en) 2020-09-28 2020-09-28 Service deployment and resource allocation method considering multi-user mobility

Country Status (1)

Country Link
CN (1) CN112148492B (en)

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112511652A (en) * 2021-02-03 2021-03-16 电子科技大学 Cooperative computing task allocation method under edge computing
CN112822701A (en) * 2020-12-31 2021-05-18 中山大学 Multi-user deep neural network model segmentation and resource allocation optimization method in edge computing scene
CN113259472A (en) * 2021-06-08 2021-08-13 江苏电力信息技术有限公司 Edge node resource allocation method for video analysis task
CN113301151A (en) * 2021-05-24 2021-08-24 南京大学 Low-delay containerized task deployment method and device based on cloud edge cooperation
CN113595801A (en) * 2021-08-09 2021-11-02 湘潭大学 Deployment method of edge cloud network server based on task flow and timeliness
CN114139730A (en) * 2021-06-30 2022-03-04 武汉大学 Dynamic pricing and deployment method for machine learning task in edge cloud network
CN114139730B (en) * 2021-06-30 2024-04-19 武汉大学 Dynamic pricing and deployment method for machine learning tasks in edge cloud network

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107846704A (en) * 2017-10-26 2018-03-27 北京邮电大学 A kind of resource allocation and base station service arrangement method based on mobile edge calculations
CN109151864A (en) * 2018-09-18 2019-01-04 贵州电网有限责任公司 A kind of migration decision and resource optimal distribution method towards mobile edge calculations super-intensive network
US20190098474A1 (en) * 2016-05-28 2019-03-28 Huawei Technologies Co., Ltd. Application migration method in a mobile edge system, related device, and system
CN109862592A (en) * 2018-12-06 2019-06-07 北京邮电大学 Resource management and dispatching method under a kind of mobile edge calculations environment based on multi-base station cooperative
CN111090522A (en) * 2019-12-13 2020-05-01 南京邮电大学 Scheduling system and decision method for service deployment and migration in mobile edge computing environment
WO2020119648A1 (en) * 2018-12-14 2020-06-18 深圳先进技术研究院 Computing task unloading algorithm based on cost optimization
CN111585916A (en) * 2019-12-26 2020-08-25 国网辽宁省电力有限公司电力科学研究院 LTE electric power wireless private network task unloading and resource allocation method based on cloud edge cooperation

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20190098474A1 (en) * 2016-05-28 2019-03-28 Huawei Technologies Co., Ltd. Application migration method in a mobile edge system, related device, and system
CN107846704A (en) * 2017-10-26 2018-03-27 北京邮电大学 A kind of resource allocation and base station service arrangement method based on mobile edge calculations
CN109151864A (en) * 2018-09-18 2019-01-04 贵州电网有限责任公司 A kind of migration decision and resource optimal distribution method towards mobile edge calculations super-intensive network
CN109862592A (en) * 2018-12-06 2019-06-07 北京邮电大学 Resource management and dispatching method under a kind of mobile edge calculations environment based on multi-base station cooperative
WO2020119648A1 (en) * 2018-12-14 2020-06-18 深圳先进技术研究院 Computing task unloading algorithm based on cost optimization
CN111090522A (en) * 2019-12-13 2020-05-01 南京邮电大学 Scheduling system and decision method for service deployment and migration in mobile edge computing environment
CN111585916A (en) * 2019-12-26 2020-08-25 国网辽宁省电力有限公司电力科学研究院 LTE electric power wireless private network task unloading and resource allocation method based on cloud edge cooperation

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
CAN WANG: "Edge Cloud Capacity Allocation for Low Delay Computing on Mobile Devices", 2017 IEEE INTERNATIONAL SYMPOSIUM ON PARALLEL AND DISTRIBUTED PROCESSING WITH APPLICATIONS AND 2017 IEEE INTERNATIONAL CONFERENCE ON UBIQUITOUS COMPUTING AND COMMUNICATIONS (ISPA/IUCC) *
蔡政: "移动边缘计算中的任务迁移与任务部署", 中国优秀硕士学位论文全文数据库-信息科技辑, pages 3 *

Cited By (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112822701A (en) * 2020-12-31 2021-05-18 中山大学 Multi-user deep neural network model segmentation and resource allocation optimization method in edge computing scene
CN112511652A (en) * 2021-02-03 2021-03-16 电子科技大学 Cooperative computing task allocation method under edge computing
CN112511652B (en) * 2021-02-03 2021-04-30 电子科技大学 Cooperative computing task allocation method under edge computing
CN113301151A (en) * 2021-05-24 2021-08-24 南京大学 Low-delay containerized task deployment method and device based on cloud edge cooperation
CN113301151B (en) * 2021-05-24 2023-01-06 南京大学 Low-delay containerized task deployment method and device based on cloud edge cooperation
CN113259472A (en) * 2021-06-08 2021-08-13 江苏电力信息技术有限公司 Edge node resource allocation method for video analysis task
CN114139730A (en) * 2021-06-30 2022-03-04 武汉大学 Dynamic pricing and deployment method for machine learning task in edge cloud network
CN114139730B (en) * 2021-06-30 2024-04-19 武汉大学 Dynamic pricing and deployment method for machine learning tasks in edge cloud network
CN113595801A (en) * 2021-08-09 2021-11-02 湘潭大学 Deployment method of edge cloud network server based on task flow and timeliness

Also Published As

Publication number Publication date
CN112148492B (en) 2023-07-28

Similar Documents

Publication Publication Date Title
CN112148492A (en) Service deployment and resource allocation method considering multi-user mobility
Cicconetti et al. A decentralized framework for serverless edge computing in the internet of things
Sefati et al. Load balancing in cloud computing environment using the Grey wolf optimization algorithm based on the reliability: performance evaluation
Maleki et al. Mobility-aware computation offloading in edge computing using machine learning
Abd Elaziz et al. IoT workflow scheduling using intelligent arithmetic optimization algorithm in fog computing
CN112199189A (en) Adaptation of deep learning models to resource constrained edge devices
CN114915630A (en) Task allocation method based on Internet of things equipment, network training method and device
Lee et al. Adaptive fault-tolerant scheduling strategies for mobile cloud computing
WO2021087639A1 (en) Cdn optimization platform
CN114595049A (en) Cloud-edge cooperative task scheduling method and device
Abreu et al. A rank scheduling mechanism for fog environments
Gkatzikis et al. Mobiles on cloud nine: Efficient task migration policies for cloud computing systems
Ziafat et al. A hierarchical structure for optimal resource allocation in geographically distributed clouds
CN115134371A (en) Scheduling method, system, equipment and medium containing edge network computing resources
Badri et al. A sample average approximation-based parallel algorithm for application placement in edge computing systems
Kumar et al. C-FDLA: Crow search with integrated fractional dragonfly algorithm for load balancing in cloud computing environments
Filelis-Papadopoulos et al. Towards simulation and optimization of cache placement on large virtual content distribution networks
Cao et al. Distributed workflow mapping algorithm for maximized reliability under end-to-end delay constraint
Premalatha et al. Optimal Energy-efficient Resource Allocation and Fault Tolerance scheme for task offloading in IoT-FoG Computing Networks
Chen et al. A game theoretic approach to task offloading for multi-data-source tasks in mobile edge computing
Patel et al. An improved load-balancing mechanism based on deadline failure recovery on GridSim
CN113190342A (en) Method and system architecture for multi-application fine-grained unloading of cloud-edge cooperative network
Khanh et al. Fuzzy-based mobile edge orchestrators in heterogeneous IoT environments: An online workload balancing approach
Asghari et al. Server placement in mobile cloud computing: a comprehensive survey for edge computing, fog computing and cloudlet
EP3096227A1 (en) Resource allocation method in distributed clouds

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