CN113364626A - Service placement and bandwidth allocation method for video analysis application facing edge environment - Google Patents

Service placement and bandwidth allocation method for video analysis application facing edge environment Download PDF

Info

Publication number
CN113364626A
CN113364626A CN202110639353.6A CN202110639353A CN113364626A CN 113364626 A CN113364626 A CN 113364626A CN 202110639353 A CN202110639353 A CN 202110639353A CN 113364626 A CN113364626 A CN 113364626A
Authority
CN
China
Prior art keywords
decision
service
user
bandwidth allocation
edge
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
CN202110639353.6A
Other languages
Chinese (zh)
Other versions
CN113364626B (en
Inventor
韦磊
缪巍巍
曾锃
曹拓
钱柱中
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
State Grid Jiangsu Electric Power Co Ltd
Information and Telecommunication Branch of State Grid Jiangsu Electric Power Co Ltd
Original Assignee
Nanjing University
State Grid Jiangsu Electric Power Co Ltd
Information and Telecommunication Branch of State Grid Jiangsu Electric Power Co Ltd
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, State Grid Jiangsu Electric Power Co Ltd, Information and Telecommunication Branch of State Grid Jiangsu Electric Power Co Ltd filed Critical Nanjing University
Priority to CN202110639353.6A priority Critical patent/CN113364626B/en
Publication of CN113364626A publication Critical patent/CN113364626A/en
Application granted granted Critical
Publication of CN113364626B publication Critical patent/CN113364626B/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
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/14—Network analysis or design
    • H04L41/142—Network analysis or design using statistical or mathematical methods
    • 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
    • H04L41/08—Configuration management of networks or network elements
    • H04L41/0896—Bandwidth or capacity management, i.e. automatically increasing or decreasing capacities
    • 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
    • H04L67/50—Network services
    • H04L67/51—Discovery or management thereof, e.g. service location protocol [SLP] or web services

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Physics & Mathematics (AREA)
  • Algebra (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Analysis (AREA)
  • Mathematical Optimization (AREA)
  • Mathematical Physics (AREA)
  • Probability & Statistics with Applications (AREA)
  • Pure & Applied Mathematics (AREA)
  • Data Exchanges In Wide-Area Networks (AREA)

Abstract

The invention discloses a service placement and bandwidth allocation method for video analysis application facing edge environment, which comprises the following steps: B1) collecting the system state information of the edge environment of the current time slot; B2) based on the current state information of the edge environment system, calculating a service placement decision and a bandwidth allocation decision by using a two-layer iterative algorithm; B3) performing a service placement decision and a bandwidth allocation decision; B4) and calculating the service migration cost of each user according to the decision result, and calculating and updating the queue length to be used as the state information of one edge environment system of the next time slot. The invention provides an online two-layer iterative service placement and bandwidth allocation method aiming at video analysis application in an edge computing system, efficiently uses limited computing resources and bandwidth resources of edge nodes, and effectively balances service migration cost and user experience.

Description

Service placement and bandwidth allocation method for video analysis application facing edge environment
Technical Field
The invention relates to the field of edge computing, in particular to a service placement and bandwidth allocation method and system for video analysis application facing to an edge environment.
Background
In recent years, with the rise and development of industrial internet of things, the internet of everything and intelligent energization become development trends, and one important application is to analyze video data acquired by various cameras by using technologies such as machine learning and the like to realize functions such as area monitoring, identity authentication, anomaly detection and the like. For example, in the electric power thing networking, the transmission line patrols and examines personnel and utilize the camera multi-angle on the unmanned aerial vehicle to gather the circuit image, combines image recognition technology to carry out the anomaly detection to the circuit, ensures transmission of electricity safety. With the wide deployment of the video analysis applications, a traditional centralized cloud computing mode is no longer suitable, firstly, the consumption of network resources is high due to the fact that massive video data are uploaded to a cloud data center, and secondly, the real-time requirement is difficult to meet due to long-distance communication between cloud and terminals. Under the background, a new computing mode of edge computing is proposed, computing power is put to the edge of a network, and video analysis services are provided by utilizing the advantage of a position close to a user, so that the traffic is effectively reduced, and the time delay is reduced, and therefore the method is considered to be an effective method for solving the problems.
However, it is not an easy task to provide video analytics services at the edge of the network. Relevant research work shows that the time delay and the frame rate are two important factors influencing the application performance of video analysis, and the lower the time delay is, the higher the frame rate is, the better the performance is, and the better the user experience is. Therefore, the service entity of the user should be deployed on the edge node closest to the user and allocate sufficient bandwidth to the edge node, so that the delay is as small as possible, the frame rate is as large as possible, and the performance is as good as possible. However, the resources of the edge node are very limited, and the requirements of all users nearby cannot be met. In addition, when a user moves from one area to another area, in order to ensure user experience, a service entity should be migrated to a new edge node, and this migration process may bring additional system overhead, and service interruption during the migration process may also affect user experience.
In summary, for video analysis applications in the edge environment, how to provide an efficient and reasonable service placement and bandwidth allocation method to make the overall performance of all video analysis applications better and better is a urgent need to be researched and solved.
Disclosure of Invention
The purpose of the invention is as follows: in order to solve the defects in the prior art, the invention aims to provide a service placement and bandwidth allocation method and system for video analysis application facing to an edge environment.
The technical scheme is as follows: in order to solve the above technical problem, the present invention provides a service placement and bandwidth allocation method for video analysis application in an edge environment, which includes the following steps:
B1) collecting the system state information of the edge environment of the current time slot;
B2) based on the current state information of the edge environment system, calculating a service placement decision and a bandwidth allocation decision by using a two-layer iterative algorithm;
B3) performing a service placement decision and a bandwidth allocation decision;
B4) calculating the service migration cost of each user according to the decision result, thereby calculating and updating the queue length to be used as the state information of one edge environment system of the next time slot;
where queue length represents the extent to which the current actual time-averaged migration cost exceeds the time-averaged service migration cost budget.
Preferably, an initialization step of a0) discretizing the sustained system time into time slots and initializing the queue length to 0 is further included.
Preferably, the system state information of the edge environment in the step B1) includes dynamic information and static information.
Preferably, the static information comprises at least one of the following: the method comprises the steps of providing a user set, video analysis services requested by all users, an edge node set, a video analysis service set provided by an edge environment system, position information of all edge nodes, computing power of all edge nodes, computing resources required by any video analysis service, resolution of any video analysis service, an experience influence function of any video analysis service on time delay and frame rate, a Lyapunov control parameter V and a smoothing parameter beta.
Preferably, the dynamic information includes at least one of the following: the position information of all users, the bandwidth capacity of any edge node, the propagation delay between any nodes including the cloud data center node and all edge nodes, and the service placement decision and queue length of the previous time slot.
Preferably, the two-layer iterative algorithm in step B2) specifically includes:
based on a Markov approximation theory, an inner nested layer iterative algorithm is adopted to calculate an outer layer iterative algorithm of an expected approximately optimal service placement decision; and
and calculating an optimal inner-layer iterative algorithm of the bandwidth allocation decision under a given service placement decision based on the KKT condition in the convex optimization theory.
Preferably, the outer iterative algorithm comprises the following steps:
11) taking the current formed service placement state as an initial service placement decision, calculating an initial bandwidth allocation decision under the initial service placement decision by using an inner layer iterative algorithm, and calculating the initial service placement decision and a target value f under the initial bandwidth allocation decision;
12) randomly selecting a user and a node, and changing the service entity of the user on the node on the basis of the initial service placement decision to form a candidate service placement decision;
13) judging whether the candidate service placement decision is feasible, if not, returning to the step 12), if so, calculating the candidate bandwidth allocation decision under the candidate service placement decision by using an inner-layer iterative algorithm, and countingTarget values under candidate service placement decisions and candidate bandwidth allocation decisions
Figure BDA0003107056870000031
14) Computing candidate decision acceptance probabilities
Figure BDA0003107056870000032
Wherein β is a smoothing parameter; adopting the candidate decision according to the probability of eta, making the candidate decision as a new initial decision, and rejecting the candidate decision according to the probability of 1-eta;
repeatedly executing the steps 12) to 14) until the initial decisions in the W iterations are kept unchanged or the difference of the target values in the W iterations is smaller than a certain preset threshold; wherein W is a predetermined number of times and is a natural number.
Further preferably, in the step 13), it is determined whether the candidate service placement decision is feasible, and whether the following computational resource constraint, placement constraint and indication constraint are simultaneously satisfied is determined as a basis;
wherein the computing resource constraints are: at any node, the total demand for computing resources cannot exceed the total supply of computing resources;
wherein the placement constraint is: the service entity of any user is required to be placed on one node;
wherein the indicated constraint is: the service placement decision variable is an indication variable, and the value can only be 0 or 1.
Even more preferably, the service placement decision xtAnd bandwidth allocation decision ytThe following formula for calculating the target value f is:
Figure BDA0003107056870000033
wherein V is a Lyapunov control parameter for controlling the importance of user experience compared with system cost;
where t represents the current time slot and,
Figure BDA0003107056870000039
representing a set of users;
gkrepresenting the video analytics service requested by user k;
Figure BDA0003107056870000034
and
Figure BDA0003107056870000035
respectively representing user k in-service placement decisions xtAnd bandwidth allocation decision ytTime delay and frame rate;
Figure BDA0003107056870000036
and
Figure BDA0003107056870000037
respectively representing experience influence functions of the video analysis service g on time delay and frame rate, wherein the value ranges are 0-1, and the smaller the value is, the better the value is;
qtindicating the queue length at time slot t;
Figure BDA0003107056870000038
represents the service migration cost generated by the process that the user k realizes the service placement decision from the time slot t-1 to the time slot t, and the calculation formula is
Figure BDA0003107056870000041
Figure BDA0003107056870000042
Wherein
Figure BDA00031070568700000412
Is a node set comprising a cloud data center node and each edge node, wherein
Figure BDA0003107056870000043
Indicating whether the service entity of user k of time slot t is placed at node nIf so, then
Figure BDA0003107056870000044
If not, then
Figure BDA0003107056870000045
Wherein
Figure BDA00031070568700000413
Serving g for video analysis of user k at time slot tkSystem cost resulting from migration from node i to node j;
Cavgrepresenting a time-averaged service migration cost budget.
Further preferably, the inner-layer iterative algorithm for calculating the optimal bandwidth allocation decision under a given service placement decision based on the KKT condition in the convex optimization theory specifically includes: under the given service placement decision, each edge node independently and parallelly calculates the respective bandwidth allocation decision; for the bandwidth allocation decision of any edge node n, obtaining the optimal condition of the bandwidth allocation decision by applying the KKT condition in the convex optimization theory to obtain the optimal bandwidth allocation decision, wherein the optimal condition is as follows: the bandwidth resources of the edge node are used up, and the derivatives of the experience functions of all users within the service range are the same or the maximum difference is smaller than a certain preset threshold.
More preferably, for the bandwidth allocation decision of any edge node, obtaining the optimal condition of the bandwidth allocation decision by applying the KKT condition in the convex optimization theory to obtain the optimal bandwidth allocation decision specifically includes: the edge node takes uniformly distributed bandwidth resources as an initial bandwidth distribution decision, and iteratively updates the bandwidth distribution decision until the experience function gamma of all users in the service range of the edge nodekThe maximum difference in the derivatives with respect to their bandwidths is less than some preset threshold;
the optimal conditions are as follows:
Figure BDA0003107056870000046
wherein
Figure BDA0003107056870000047
Representing the local edge node of user k at time slot t,
Figure BDA0003107056870000048
indicating the bandwidth resources allocated by user k at time slot t,
Figure BDA0003107056870000049
is the bandwidth capacity of the edge node n;
Figure BDA00031070568700000410
representing a set of users; λ is a constant;
wherein gamma isk(y) a video analysis experience function for user k when the allocated bandwidth is y:
Figure BDA00031070568700000411
wherein lkIs to place a decision x at a given servicetPlacing the communication time delay between the node of the service entity of the user k and the local edge node of the user k;
wherein
Figure BDA00031070568700000512
An ith parameter associated with the video analytics service g representing user k,
Figure BDA00031070568700000513
video analytics service g representing user kkPicture resolution of.
Still further preferably, in the inner-layer iterative algorithm, for each edge node, the following steps are performed independently and in parallel:
21) uniformly distributing the bandwidth resources of the edge node
Figure BDA0003107056870000051
And assigning an initial value to a variable gamma for regulating the magnitude of the updated bandwidth allocation decision:
Figure BDA0003107056870000052
wherein
Figure BDA0003107056870000053
A user set in the service range of the edge node is served;
22) finding out user k with minimum derivative of gamma functionmAnd the corresponding derivative m thereof, and finding out the user k with the maximum derivative of the gamma functionMAnd their corresponding derivatives M, wherein
Figure BDA0003107056870000054
23) From the allocation to user kMBandwidth of delta is taken out and allocated to user kmWherein
Figure BDA0003107056870000055
Figure BDA0003107056870000056
Enabling gamma to increase by 1;
24) repeatedly executing the step 22) to the step 23) until the difference between the maximum derivative and the minimum derivative is smaller than a certain preset threshold value, and obtaining the optimal bandwidth allocation decision of the edge node;
and after all the edge nodes finish the steps, obtaining and outputting the overall optimal bandwidth allocation decision.
Preferably, the step B3) performs service placement and bandwidth allocation decisions, including:
B31) for any user, if the current service placement decision of the user is consistent with the service placement decision of the previous time slot, keeping the current service placement decision unchanged, otherwise, according to the current service placement decision, migrating the service entity corresponding to the user to a new node in the form of a virtual machine or a container; the nodes are any nodes including a cloud data center node and each edge node;
B32) for any user, the local edge node allocates bandwidth resources with corresponding size to the user according to the bandwidth allocation decision.
Preferably, the step B4) calculates the service migration cost of each user according to the decision result, and calculates and updates the queue length: represents the service migration cost generated by the process that the user k completes the service placement from the time slot t-1 to the time slot t
Figure BDA0003107056870000057
Wherein
Figure BDA0003107056870000058
Is a computing node set comprising a cloud data center node and edge nodes,
Figure BDA0003107056870000059
indicating whether the service entity of the user k of the time slot t is placed on the node n, if yes, then
Figure BDA00031070568700000510
If not, then
Figure BDA00031070568700000511
Wherein
Figure BDA00031070568700000514
Serving g for video analysis of user k at time slot tkSystem cost resulting from migration from node i to node j;
wherein the queue length
Figure BDA0003107056870000061
Wherein
Figure BDA0003107056870000062
Representing a set of users, CavgThe cost budget is migrated for the time-averaged service.
The invention also provides a service placement and bandwidth allocation system for the video analysis application facing the edge environment, which comprises the following components:
the state collection module is used for collecting the state information of the edge environment system of the current time slot;
the decision calculation module is used for calculating a service placement decision and a bandwidth allocation decision by utilizing a two-layer iterative algorithm based on the current state information of the edge environment system;
a decision execution module for executing a service placement decision and a bandwidth allocation decision;
the state updating module is used for calculating the service migration cost of each user according to the decision result so as to calculate and update the queue length to serve as the state information of one edge environment system of the next time slot;
wherein the queue length is the extent to which the current actual time-averaged migration cost exceeds the time-averaged service migration cost budget.
Preferably, the system further comprises an initialization module for discretizing the continuous system time into time slots and initializing the queue length to 0.
The invention also provides another service placement and bandwidth allocation system for the video analysis application facing the edge environment, which comprises a cloud data center and a plurality of edge computing nodes connected with the cloud data center through a network;
all edge computing nodes are connected through a backhaul network; each edge computing node comprises a wireless access point and an edge server;
the edge computing nodes deploy and execute the service placement decision and the bandwidth allocation decision of each edge computing node by adopting any one of the service placement and bandwidth allocation methods.
Has the advantages that: the invention provides a service placement and bandwidth allocation method and system for video analysis application facing to an edge environment, which firstly provides an online two-layer iteration service placement and bandwidth allocation method based on Lyapunov optimization aiming at video analysis application in an edge computing system, efficiently uses limited computing resources and bandwidth resources on edge nodes, effectively balances service migration cost and user experience, and effectively reduces adverse effects of time delay and frame rate on the user experience under the constraint of meeting the computing resources, the bandwidth resources and long-term service migration cost.
Drawings
Fig. 1 is a schematic system structure diagram of a video analysis application in a marginal environment.
Detailed Description
The present invention will be described in further detail with reference to the following embodiments and the accompanying drawings, wherein the following embodiments are only for the purpose of clearly understanding the technical solutions of the present invention, and the present invention is not limited thereto.
In consideration of different sensitivities of various video analysis applications to time delay and frame rate, the invention provides a more practical and effective method and thought to select some service entities to be placed on a certain non-nearest but proper edge node and reasonably allocate bandwidth resources correspondingly, thereby efficiently utilizing the resources of the edge node and improving the overall experience of all users, which needs to carefully balance performance improvement and extra cost caused by service migration when making a service placement decision.
Edge compute nodes, also referred to herein as edge nodes, are described herein. The time-averaged service migration cost is also referred to herein as a long-term time-averaged service migration cost or a time-averaged migration cost. In this context
Figure BDA0003107056870000078
It is the local edge node of user k, and is also the location of user k, because the location scale in this scheme is the edge node. The edge node n broadly refers to a general edge node.
The video analysis application referred to herein means that a conventional single video analysis application is divided into a service entity (i.e., a video analysis service) and a client entity, where the client entity operates on a mobile device, and the service entity operates as a service on a cloud data center or an edge node, and the service entity and the cloud data center cooperate to complete a video analysis application experience process of a user. The service placement and bandwidth allocation problem described herein refers to determining whether a service entity is deployed in a cloud data center or which edge node, and how much bandwidth resources are allocated to each user by the edge node.
FIG. 1 is a schematic diagram of a system architecture of a video analysis application in an edge environment, as shown in FIG. 1, in the mobile edge computing system, there is a group of edge nodes, which are recorded as a set
Figure BDA0003107056870000071
And a cloud data center. Each edge node is a combination of a wireless access point (such as a base station, a WiFi hotspot, and the like) and an edge server, and has certain network capacity and computing capacity, so that the edge node can serve as a substitute of a cloud data center for nearby users. The cloud data center and the edge nodes are connected with a Wide Area Network (WAN) through a backhaul network, and the edge nodes are connected with each other through the backhaul network and can communicate with the cloud data center through the WAN. All the computing nodes form a set
Figure BDA0003107056870000072
Wherein node 0 represents the cloud data center. Each node
Figure BDA0003107056870000073
All consist of a binary group
Figure BDA0003107056870000074
Is shown in which W isnRepresenting computing power in cycles/s, is fixed, and
Figure BDA0003107056870000075
representing the bandwidth of the wireless access link in bps, is affected by the current environment and thus varies with time t. Considering the particularity of the cloud data center as a computing node, the computing power is nearly infinite, and the bandwidth of the wireless access link is 0, namely W0＝∞,
Figure BDA0003107056870000076
t＝1,2,…。
The system provides G video analysis services (corresponding to the service entities in FIG. 1), which are recorded as a set
Figure BDA0003107056870000077
Figure BDA0003107056870000081
Each video analytics service
Figure BDA0003107056870000082
All consist of a five-membered group
Figure BDA0003107056870000083
To depict/describe, wherein: w is agThe calculation resources required by the running of the video analysis service are represented, and the unit is cycles/s; r isgA picture resolution representing the video analytics service;
Figure BDA0003107056870000084
and
Figure BDA0003107056870000085
the experience influence functions of the video analysis service on the time delay and the frame rate are respectively represented, the value ranges are all 0-1, and the smaller the value is, the better the value is. The experience impact functions described above have the following forms, respectively:
Figure BDA0003107056870000086
Figure BDA0003107056870000087
wherein { pg,iI ═ 1,2, …,6} is a function parameter, which represents the ith parameter related to the video analysis service g, and is obtained from the fitting result, and the function parameters of different video analysis applications are different (three video analysis applications are experienced under different time delays and frame rates, and the results are respectively subjected to fitting analysis to obtain experience influence functions, and the functions have the same form and different parameters). And CgIs a function of the source node i, the destination node j and the time slot t, and represents the result of migrating the video analysis service g from the node i to the node j at the time slot tThe system cost. In practical applications, the same video analysis application may support multiple screen resolutions, but the user generally does not change the resolution during the experience, so the video analysis application supporting multiple resolutions can be regarded as multiple video analysis applications supporting only a single resolution without loss of generality.
System-oriented user group forming user set
Figure BDA0003107056870000088
Any one of the users
Figure BDA0003107056870000089
Requested video analytics service is noted
Figure BDA00031070568700000810
The user is in motion, and the moving mode (such as motion track and speed) of the user is not predictable. To characterize user mobility, system time is discretized into time slots
Figure BDA00031070568700000811
At any time slot
Figure BDA00031070568700000812
At the beginning, any user
Figure BDA00031070568700000813
Will connect to its nearest edge node, denoted
Figure BDA00031070568700000814
So that the location of the user, also called the local edge node of user k, can be determined.
In addition, with indicating variables
Figure BDA00031070568700000815
Indicating a time slot
Figure BDA00031070568700000816
Time user
Figure BDA00031070568700000817
Whether the service entity of (2) is placed in the node
Figure BDA00031070568700000818
If so, then
Figure BDA00031070568700000819
At this point node n is called the serving node for user k, otherwise,
Figure BDA00031070568700000820
using positive real variables
Figure BDA00031070568700000821
Indicating a time slot
Figure BDA00031070568700000822
Is a user
Figure BDA00031070568700000823
The allocated bandwidth. And, a time slot
Figure BDA00031070568700000824
The overall service placement decision is
Figure BDA00031070568700000825
Time slot recording
Figure BDA00031070568700000826
The overall bandwidth allocation decision is
Figure BDA00031070568700000827
When the service placement decision is determined to be xtAnd the bandwidth allocation decision is ytLater, the time delay perceived by any user k
Figure BDA00031070568700000828
Sum frame rate
Figure BDA00031070568700000829
The calculation method is as follows:
Figure BDA0003107056870000091
Figure BDA0003107056870000092
where alpha represents the compression ratio,
Figure BDA0003107056870000097
video analytics service g representing user kkThe resolution of the picture of (a) is,
Figure BDA0003107056870000093
representing the communication delay between the local edge node of user k and node n at time slot t.
Based on the system model of video analysis application under the mobile edge calculation, the service placement and bandwidth allocation method for video analysis application facing to the edge environment provided by the embodiment is an implementation framework of an online two-layer iteration method based on lyapunov optimization, and comprises an initialization step and a service placement and bandwidth allocation step.
Wherein the service placement and bandwidth allocation step comprises, at any time slot t ═ 1,2,3, … (or for any time slot t ═ 1,2,3, …), performing the steps of:
B1) collecting the system state information of the edge environment of the current time slot t;
B2) based on the current state information of the edge environment system, calculating a service placement decision x by using a two-layer iterative algorithmtAnd bandwidth allocation decision yt；
B3) Performing service placement decision xtAnd bandwidth allocation decision yt；
B4) Calculating the service migration cost of each user according to the decision result
Figure BDA0003107056870000098
Thereby calculating and updating the queue length qtAs one of the edge environment system state information of the next slot.
Wherein the initializing step (before the integrally performing the service placement and bandwidth allocation step) comprises: A0) discretizing the continuous system time into time slots, and initializing the queue length to 0; the method specifically comprises the following steps: service placement decision x0All service entities are deployed in a cloud data center, and an inner layer iteration method is utilized to calculate a service placement decision x0Bandwidth allocation decision y0Queue length q00. Where queue length represents the extent to which the current actual time-averaged migration cost exceeds the time-averaged service migration cost budget.
The system state information of the edge environment includes dynamic information, wherein the system state dynamic information of the collection time slot t includes at least one of the following: location information for any user k
Figure BDA0003107056870000094
Bandwidth capacity of any edge node n
Figure BDA0003107056870000095
Propagation delay between any node i and any node j (including the cloud data center node and all edge nodes)
Figure BDA0003107056870000096
And service placement decision x for last slott-1And queue length qt-1。
The system state information of the edge environment further includes static information, including at least one of the following: the above-mentioned user set, the video analysis service requested by each user, the edge node set, the video analysis service set provided by the edge environment system, the location information of all edge nodes, the computing power of all edge nodes, the computing resources required by any video analysis service, the resolution of any video analysis service, the experience impact function of any video analysis service with respect to latency and frame rate, and the lyapunov control parameter V and the smoothing parameter β.
Wherein the step B3) executes a service placement decision xtAnd bandwidth allocation decision ytThe method specifically comprises the following steps:
B31) put decision x according to servicetService entity deployment is carried out, namely: for any user, if the current service placement decision of the user is consistent with the service placement decision of the previous time slot, keeping the current service placement decision unchanged, otherwise, according to the current service placement decision, migrating the service entity corresponding to the user to a new node in the form of a virtual machine or a container; the nodes are any nodes including a cloud data center node and each edge node;
specifically, it can be said that: for any user k, if a node i and a node j exist, i is not equal to j,
Figure BDA0003107056870000101
then the service entity corresponding to the user is migrated from the node i to the node j in the form of a virtual machine or a container; otherwise, the placement position of the service entity of the user is kept unchanged;
B32) decision y based on bandwidth allocationtPerforming bandwidth allocation, namely: for any user k, its local edge node
Figure BDA0003107056870000102
Allocate size for bandwidth allocation decision
Figure BDA0003107056870000103
Bandwidth resources of (a);
wherein, the service migration cost of user k in the step B4) is described above
Figure BDA0003107056870000104
Refers to the cost of user k in completing the process from the service placement of time slot t-1 to the service placement of time slot t (i.e. the service migration cost of user k at time slot t):
Figure BDA0003107056870000105
wherein
Figure BDA0003107056870000109
Serving g for video analysis of user k at time slot tkSystem cost resulting from migration from node i to node j; c abovegIs a function of the service g and,
Figure BDA0003107056870000107
is the service g requested by user kkA function of (a);
wherein the queue length qtExceeding the time-averaged service migration cost budget C for the current actual time-averaged migration costavgDegree of (c):
Figure BDA0003107056870000106
wherein C isavgRepresenting a long-term time-averaged service migration cost budget.
The computational service placement decision x in this method is described in detail belowtAnd bandwidth allocation decision ytAnd the implementation process thereof. Wherein the two-layer iterative algorithm in the step B2) specifically includes:
based on Markov approximate theory and internal nested inner layer iterative algorithm, calculating expected approximate optimal service placement decision xtThe outer iteration algorithm of (1); and
placing a decision x at a given service based on KKT conditions in convex optimization theorytLower computational optimal bandwidth allocation decision ytThe inner iterative algorithm of (1).
Specifically, the outer-layer iterative algorithm includes the following steps:
11) placing state x with currently formed servicet-1Placing decision x for initial servicetComputing an initial service placement decision x using a inlier iterative algorithmtInitial bandwidth allocation decision ytAnd calculating an initial decision xtAnd ytA target value f; where service placement decision xtAnd bandwidth allocation decision ytThe following formula for calculating the target value f is:
Figure BDA0003107056870000111
wherein V is a Lyapunov control parameter for controlling the importance of user experience compared with system cost; wherein
Figure BDA0003107056870000112
And
Figure BDA0003107056870000113
respectively representing user k in-service placement decisions xtAnd bandwidth allocation decision ytThe delay and the frame rate of the lower layer,
Figure BDA0003107056870000114
and
Figure BDA0003107056870000115
the video analysis service g respectively representing the current bandwidth allocation decision of the user kkExperience impact values with respect to latency and frame rate;
12) self-user collection
Figure BDA0003107056870000117
Randomly selecting a user k' from the node set
Figure BDA0003107056870000118
In the method, a certain node n' is randomly selected, and a decision x is placed in the initial servicetOn the basis, the service entity of the user k ' is placed on the node n ', and the positions of other service entities are kept unchanged, so that a candidate service placement decision x ' is formed;
13) judging whether the candidate service placement decision x ' is feasible, if not, returning to the step 12), if so, calculating a candidate bandwidth allocation decision y ' under the candidate service placement decision x ' by using an inner layer iterative algorithm,and calculating target values under the candidate decisions x' and y
Figure BDA0003107056870000119
Wherein, whether the candidate service placement decision x' is feasible is judged, and whether the following computing resource constraint, placement constraint and indication constraint are simultaneously met is judged according to the judgment:
wherein the computing resource constraints are: at any node, the total demand for computing resources cannot exceed the total supply of computing resources; namely:
Figure BDA0003107056870000116
wherein the placement constraint is: the service entity of any user is required to be placed on one node; namely:
Figure BDA0003107056870000121
wherein the indicated constraint is: the service placement decision variable is an indication variable, and the value is only 0 or 1; namely, it is
Figure BDA0003107056870000122
If x ' does not satisfy any of the above constraints (calculation resource constraint, placement constraint and indication constraint), considering that x ' is not feasible, returning to step 12), otherwise, calculating a candidate bandwidth allocation decision y ' under x ' by using an inner-layer iterative algorithm, and calculating target values under the candidate decision x ' and y
Figure BDA0003107056870000123
Target value
Figure BDA0003107056870000124
The calculation method of (1) is the same as that in step 11), and is not described again;
14) computing candidate decision acceptance probabilities
Figure BDA0003107056870000125
Wherein beta is a smoothing parameter used for balancing contradiction between exploring unknown information and utilizing the current existing information; adopting the candidate decision with the probability of eta, i.e. making the candidate decision a new initial decision, xt＝x′，ytY'; rejecting the candidate decision with a probability of 1- η, the initial decision remaining unchanged;
repeating steps 12) to 14) until the initial decision remains unchanged for several iterations or the target value differs for several iterations (i.e. the target value is different for several iterations)
Figure BDA0003107056870000126
) Are all less than a certain preset threshold; in this embodiment, the number of times is W, where W is a predetermined number of times and is a natural number.
In the above steps of the method, in each iteration, if the candidate decision has a smaller target value, the algorithm tends to adopt the candidate decision, and the larger the difference with the target value of the initial decision, the larger the probability that the candidate decision is adopted; if the target value of the candidate decision is larger, the algorithm rejects the candidate decision with a larger probability, and it should be noted that the algorithm has a smaller probability to adopt the candidate decision, so that the algorithm can be prevented from falling into the local optimal trap. As the iteration of the algorithm proceeds, the algorithm eventually converges to a better decision with a greater probability.
In the service placement and bandwidth allocation method provided by the present invention, in step B2), the service placement decision x is determined based on the KKT condition in the convex optimization theory analysistLower computational optimal bandwidth allocation decision ytThe inner-layer iterative algorithm of (1) can independently calculate the bandwidth allocation decisions of each edge node in parallel because the bandwidth allocation decisions of each edge node are independent and do not influence each other, and specifically comprises the following steps: placing decision x at the given servicetThe video analysis experience of user k is determined by the allocated bandwidth, and is a bandwidth-related experience function denoted as ΓkThe bandwidth allocation decisions of the edge nodes are independent of each other, and each edge node is independentComputing respective bandwidth allocation decisions in parallel; for the bandwidth allocation decision of any edge node n, obtaining a bandwidth allocation decision y by applying a KKT condition in a convex optimization theorytTo obtain an optimal bandwidth allocation decision thereof, wherein the optimal conditions are as follows: the bandwidth resources of the edge node are used up, and the derivatives of the experience functions of all users within the service range are the same or the maximum difference is smaller than a certain preset threshold.
For any edge node n, when the service placement decision xtGiven that the video analysis experience of the user is determined only by the allocated bandwidth, let us note that the video analysis experience function (representing the degree of influence on the video analysis experience) of the user k when the allocated bandwidth is y is Γk(y) the calculation formula is:
Figure BDA0003107056870000131
wherein the content of the first and second substances,
Figure BDA0003107056870000132
is to place a decision x at a given servicetPlacing the communication time delay between the node of the service entity of the user k and the local edge node of the user k; wherein
Figure BDA00031070568700001311
An ith parameter associated with the video analytics service g representing user k,
Figure BDA00031070568700001312
video analytics service g representing user kkPicture resolution of.
Obtaining the bandwidth allocation decision y by applying the KKT condition in the convex optimization theorytTo obtain its optimal bandwidth allocation decision ytThe optimum conditions to be satisfied include:
Figure BDA0003107056870000133
that is, the bandwidth resources of the edge node n are used up, and the derivatives of Γ functions of all users within the service range of the edge node n are the same, or, in other words, any edge node n is based on uniformly allocating bandwidth resources (as an initial bandwidth allocation decision), and iteratively updating the bandwidth allocation manner, that is, iteratively updating the bandwidth allocation decision, until the experience functions Γ of all users within the service range of the edge node n are reachedkThe maximum difference in the derivatives with respect to their bandwidths is less than some preset threshold; the λ is a constant and is an unknown constant that does not require a specific value.
According to the above optimal conditions, in the inner-layer iterative algorithm, for each edge node, the following steps are independently and parallelly executed:
21) uniformly distributing the bandwidth resources of the edge node
Figure BDA0003107056870000134
All users within: namely to
Figure BDA0003107056870000135
Figure BDA0003107056870000136
Wherein
Figure BDA0003107056870000137
For the set of users within the service range of the edge node,
Figure BDA0003107056870000138
is bandwidth capacity, wherein
Figure BDA0003107056870000139
Representing the bandwidth resource allocated to the user k at the current time slot t; and initially assigning a variable gamma for regulating and controlling the amplitude of the updated bandwidth allocation decision:
Figure BDA00031070568700001310
22) finding the minimum derivative of the gamma functionFamily kmAnd the corresponding derivative m thereof, and finding out the user k with the maximum derivative of the gamma functionMAnd its corresponding derivative M;
23) from the allocation to user kMBandwidth of delta is taken out and allocated to user kmInstant command
Figure BDA0003107056870000141
Wherein
Figure BDA0003107056870000142
And let γ increase by 1, that is γ ═ γ + 1;
24) repeatedly executing the step 22) to the step 23) until the difference between the maximum derivative and the minimum derivative is smaller than a certain preset threshold value, and obtaining the optimal bandwidth allocation decision of the edge node;
and after all the edge nodes finish the steps, integrating all the results to obtain and output an overall optimal bandwidth allocation decision.
The embodiment also provides a service placement and bandwidth allocation system for a video analysis application facing an edge environment, which includes:
the state collection module is used for collecting the state information of the edge environment system of the current time slot;
the decision calculation module is used for calculating a service placement decision and a bandwidth allocation decision by utilizing a two-layer iterative algorithm based on the current state information of the edge environment system;
a decision execution module for executing a service placement decision and a bandwidth allocation decision;
the state updating module is used for calculating the service migration cost of each user according to the decision result so as to calculate and update the queue length to serve as the state information of one edge environment system of the next time slot;
wherein the queue length is the extent to which the current actual time-averaged migration cost exceeds the time-averaged service migration cost budget.
In this system embodiment, the system further comprises an initialization module for discretizing the continuous system time into time slots and initializing the queue length to 0.
It should be understood that, the system embodiment formed by combining the above modules may implement all technical solutions in the above method embodiment; the functions of each module may be specifically implemented according to the method in the foregoing method embodiment, and the specific implementation process may refer to the relevant description in the foregoing embodiment, which is not described herein again.
The embodiment also provides another service placement and bandwidth allocation system for the video analysis application facing the edge environment, which comprises a cloud data center and a plurality of edge computing nodes connected with the cloud data center through a network;
all edge computing nodes are connected through a backhaul network; each edge computing node comprises a wireless access point and an edge server;
the edge computing nodes deploy and execute the service placement decision and the bandwidth allocation decision of each edge computing node by adopting any one of the service placement and bandwidth allocation methods. The specific implementation process may refer to the related description in the above method embodiments, and is not described herein again.
As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the method and/or system provided by the invention.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the method and/or system provided by the invention.
The above is only a preferred embodiment of the present invention, it should be noted that the above embodiment does not limit the present invention, and various changes and modifications made by workers within the scope of the technical idea of the present invention fall within the protection scope of the present invention.

Claims (10)

1. A service placement and bandwidth allocation method for video analysis application facing edge environment is characterized by comprising the following steps:
B1) collecting the system state information of the edge environment of the current time slot;
B2) based on the current state information of the edge environment system, calculating a service placement decision and a bandwidth allocation decision by using a two-layer iterative algorithm;
B3) performing a service placement decision and a bandwidth allocation decision;
B4) calculating the service migration cost of each user according to the decision result, and calculating and updating the queue length to be used as the state information of one edge environment system of the next time slot;
wherein the queue length represents the extent to which the current actual time-averaged migration cost exceeds the time-averaged service migration cost budget.
2. The method for service placement and bandwidth allocation for video analytics applications oriented to edge environments of claim 1, wherein:
further comprising a0) an initialization step of discretizing the sustained system time into time slots and initializing the queue length to 0;
the system state information of the edge environment in the step B1) comprises dynamic information and static information:
wherein the static information comprises at least one of: the method comprises the steps that a user set, video analysis services requested by all users, an edge node set, a video analysis service set provided by an edge environment system, position information of all edge nodes, computing capacity of all edge nodes, computing resources required by any video analysis service, resolution of any video analysis service, an experience influence function of any video analysis service on time delay and frame rate, a Lyapunov control parameter V and a smoothing parameter beta are obtained;
wherein the dynamic information comprises at least one of: the position information of all users, the bandwidth capacity of any edge node, the propagation delay between any nodes including the cloud data center node and all edge nodes, and the service placement decision and queue length of the previous time slot.
3. The service placement and bandwidth allocation method for video analytics applications facing edge environments of claim 1 or 2, wherein: the two-layer iterative algorithm in the step B2) specifically includes:
based on a Markov approximation theory, an inner nested layer iterative algorithm is adopted to calculate an outer layer iterative algorithm of an expected approximately optimal service placement decision; and
and calculating an optimal inner-layer iterative algorithm of the bandwidth allocation decision under a given service placement decision based on the KKT condition in the convex optimization theory.
4. The method for service placement and bandwidth allocation for video analytics applications oriented to edge environments as claimed in claim 3, wherein said outer layer iterative algorithm comprises the steps of:
11) taking the current formed service placement state as an initial service placement decision, calculating an initial bandwidth allocation decision under the initial service placement decision by using an inner layer iterative algorithm, and calculating the initial service placement decision and a target value f under the initial bandwidth allocation decision;
12) randomly selecting a user and a node, and changing the service entity of the user on the node on the basis of the initial service placement decision to form a candidate service placement decision;
13) judgment ofWhether the candidate service placement decision is feasible or not is judged, if not, the step 12) is returned to, if so, the candidate bandwidth allocation decision under the candidate service placement decision is calculated by using an inner layer iterative algorithm, and the candidate service placement decision and the target value under the candidate bandwidth allocation decision are calculated
Figure FDA0003107056860000021
14) Computing candidate decision acceptance probabilities
Figure FDA0003107056860000022
Wherein β is a smoothing parameter; adopting the candidate decision according to the probability of eta, making the candidate decision as a new initial decision, and rejecting the candidate decision according to the probability of 1-eta;
repeatedly executing the steps 12) to 14) until the initial decisions in the W iterations are kept unchanged or the difference of the target values in the W iterations is smaller than a certain preset threshold; wherein W is a predetermined number of times and is a natural number.
5. The method for service placement and bandwidth allocation of video analytics application oriented to an edge environment of claim 4, wherein in the step 13) it is determined whether a candidate service placement decision is feasible, based on whether the following computational resource constraints, placement constraints and indication constraints are simultaneously satisfied;
wherein the computing resource constraints are: at any node, the total demand for computing resources cannot exceed the total supply of computing resources;
wherein the placement constraint is: the service entity of any user is required to be placed on one node;
wherein the indicated constraint is: the service placement decision variable is an indication variable, and the value can only be 0 or 1.
6. The method of claim 4, wherein the service placement decision x is a service placement decision for the video analytics application facing the edge environmenttAnd bandwidth allocation decision ytThe following formula for calculating the target value f is:
Figure FDA0003107056860000023
wherein V is a Lyapunov control parameter for controlling the importance of user experience compared with system cost;
where t represents the current time slot and,
Figure FDA0003107056860000024
representing a set of users;
gk represents the video analytics service requested by user k;
Figure FDA0003107056860000031
and
Figure FDA0003107056860000032
respectively representing user k in-service placement decisions xtAnd bandwidth allocation decision ytTime delay and frame rate;
Figure FDA0003107056860000033
and
Figure FDA0003107056860000034
respectively representing experience influence functions of the video analysis service g on time delay and frame rate, wherein the value ranges are 0-1, and the smaller the value is, the better the value is;
qtindicating the queue length at time slot t;
Figure FDA0003107056860000035
represents the service migration cost generated by the process that the user k realizes the service placement decision from the time slot t-1 to the time slot t, and the calculation formula is
Figure FDA0003107056860000036
Figure FDA0003107056860000037
Wherein
Figure FDA00031070568600000311
Is a node set comprising a cloud data center node and each edge node, wherein
Figure FDA0003107056860000038
Indicating whether the service entity of the user k of the time slot t is placed on the node n, if yes, then
Figure FDA0003107056860000039
If not, then
Figure FDA00031070568600000310
Wherein
Figure FDA00031070568600000312
Serving g for video analysis of user k at time slot tkSystem cost resulting from migration from node i to node j;
Cavgrepresenting a time-averaged service migration cost budget.
7. The service placement and bandwidth allocation method for video analytics application oriented to an edge environment according to claim 3, wherein the inner-layer iterative algorithm for calculating an optimal bandwidth allocation decision under a given service placement decision based on the KKT condition in the convex optimization theory is specifically: under the given service placement decision, each edge node independently and parallelly calculates the respective bandwidth allocation decision; for the bandwidth allocation decision of any edge node n, obtaining the optimal condition of the bandwidth allocation decision by applying the KKT condition in the convex optimization theory to obtain the optimal bandwidth allocation decision, wherein the optimal condition is as follows: the bandwidth resources of the edge node are used up, and the derivatives of the experience functions of all users within the service range are the same or the maximum difference is smaller than a certain preset threshold.
8. The method for service placement and bandwidth allocation for video analytics applications oriented to edge environments of claim 7, wherein: for the bandwidth allocation decision of any edge node, obtaining the optimal condition of the bandwidth allocation decision by applying the KKT condition in the convex optimization theory to obtain the optimal bandwidth allocation decision, specifically comprising: the edge node takes uniformly distributed bandwidth resources as an initial bandwidth distribution decision, and iteratively updates the bandwidth distribution decision until the maximum difference of the experience functions Γ k of all users in the service range of the edge node with respect to the derivative of the bandwidth is less than a certain preset threshold;
the optimal conditions are as follows:
Figure FDA0003107056860000041
wherein
Figure FDA0003107056860000042
Representing the local edge node of user k at time slot t,
Figure FDA0003107056860000043
indicating the bandwidth resources allocated by user k at time slot t,
Figure FDA0003107056860000044
is the bandwidth capacity of the edge node n;
Figure FDA00031070568600000414
representing a set of users; λ is a constant;
wherein gamma isk(y) a video analysis experience function for user k when the allocated bandwidth is y:
Figure FDA0003107056860000045
wherein lkIs to place a decision x at a given servicetPlacing the communication time delay between the node of the service entity of the user k and the local edge node of the user k;
wherein
Figure FDA0003107056860000046
An ith parameter associated with the video analytics service g representing user k,
Figure FDA0003107056860000047
video analytics service g representing user kkPicture resolution of.
9. The method for service placement and bandwidth allocation for video analytics applications facing edge environments as claimed in claim 7 or 8, wherein: in the inner-layer iterative algorithm, the following steps are independently and parallelly executed for each edge node:
21) uniformly distributing the bandwidth resources of the edge node
Figure FDA0003107056860000048
And assigning an initial value to a variable gamma for regulating the magnitude of the updated bandwidth allocation decision:
Figure FDA0003107056860000049
wherein
Figure FDA00031070568600000410
A user set in the service range of the edge node is served;
22) finding out user k with minimum derivative of gamma functionmAnd the corresponding derivative m thereof, and finding out the user k with the maximum derivative of the gamma functionMAnd their corresponding derivatives M, wherein
Figure FDA00031070568600000411
23) From the allocation to user kMBandwidth of delta is taken out and allocated to user kmWherein
Figure FDA00031070568600000412
Figure FDA00031070568600000413
Enabling gamma to increase by 1;
24) repeatedly executing the step 22) to the step 23) until the difference between the maximum derivative and the minimum derivative is smaller than a certain preset threshold value, and obtaining the optimal bandwidth allocation decision of the edge node;
and after all the edge nodes finish the steps, obtaining and outputting the overall optimal bandwidth allocation decision.
10. The method for service placement and bandwidth allocation for video analytics applications oriented to an edge environment of claim 1,
said step B3) performs service placement and bandwidth allocation decisions including:
B31) for any user, if the current service placement decision of the user is consistent with the service placement decision of the previous time slot, keeping the current service placement decision unchanged, otherwise, according to the current service placement decision, migrating the service entity corresponding to the user to a new node in the form of a virtual machine or a container; the nodes are any nodes including a cloud data center node and each edge node;
B32) for any user, the local edge node allocates bandwidth resources with corresponding size to the user according to the bandwidth allocation decision;
the step B4) calculates the service migration cost of each user according to the decision result, and calculates and updates the queue length: represents the service migration cost generated by the process that the user k completes the service placement from the time slot t-1 to the time slot t
Figure FDA0003107056860000051
Wherein
Figure FDA0003107056860000052
Is a computing node set comprising a cloud data center node and edge nodes,
Figure FDA0003107056860000053
indicating whether the service entity of the user k of the time slot t is placed on the node n, if yes, then
Figure FDA0003107056860000054
If not, then
Figure FDA0003107056860000055
Wherein
Figure FDA0003107056860000056
Serving g for video analysis of user k at time slot tkSystem cost resulting from migration from node i to node j;
wherein the queue length
Figure FDA0003107056860000057
Wherein
Figure FDA0003107056860000058
Representing a set of users, CaυgThe cost budget is migrated for the time-averaged service.
CN202110639353.6A 2021-06-08 2021-06-08 Service placement and bandwidth allocation method for video analysis application facing edge environment Active CN113364626B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110639353.6A CN113364626B (en) 2021-06-08 2021-06-08 Service placement and bandwidth allocation method for video analysis application facing edge environment

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110639353.6A CN113364626B (en) 2021-06-08 2021-06-08 Service placement and bandwidth allocation method for video analysis application facing edge environment

Publications (2)

Publication Number Publication Date
CN113364626A true CN113364626A (en) 2021-09-07
CN113364626B CN113364626B (en) 2022-09-30

Family

ID=77533084

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110639353.6A Active CN113364626B (en) 2021-06-08 2021-06-08 Service placement and bandwidth allocation method for video analysis application facing edge environment

Country Status (1)

Country Link
CN (1) CN113364626B (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114090221A (en) * 2022-01-20 2022-02-25 浙江万雾信息科技有限公司 Dual migration method for tasks in edge computing environment
CN114302233A (en) * 2021-12-10 2022-04-08 网络通信与安全紫金山实验室 Video compression and network service quality joint decision method and device

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111278132A (en) * 2020-01-19 2020-06-12 重庆邮电大学 Resource allocation method for low-delay high-reliability service in mobile edge calculation
CN111459505A (en) * 2020-05-22 2020-07-28 南京大学 Multi-version inference model deployment method, device and system in edge computing environment
WO2020216135A1 (en) * 2019-04-25 2020-10-29 南京邮电大学 Multi-user multi-mec task unloading resource scheduling method based on edge-end collaboration

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2020216135A1 (en) * 2019-04-25 2020-10-29 南京邮电大学 Multi-user multi-mec task unloading resource scheduling method based on edge-end collaboration
CN111278132A (en) * 2020-01-19 2020-06-12 重庆邮电大学 Resource allocation method for low-delay high-reliability service in mobile edge calculation
CN111459505A (en) * 2020-05-22 2020-07-28 南京大学 Multi-version inference model deployment method, device and system in edge computing environment

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
TUO CAO等: "Service Placement and Bandwidth Allocation for MEC-enabled Mobile Cloud Gaming", 《IEEE》 *
ZHAOLONG NING 等: "Distributed and Dynamic Service Placement in Pervasive Edge Computing Networks", 《 IEEE TRANSACTIONS ON PARALLEL AND DISTRIBUTED SYSTEMS》 *

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114302233A (en) * 2021-12-10 2022-04-08 网络通信与安全紫金山实验室 Video compression and network service quality joint decision method and device
CN114302233B (en) * 2021-12-10 2023-10-27 网络通信与安全紫金山实验室 Video compression and network service quality joint decision method and device
CN114090221A (en) * 2022-01-20 2022-02-25 浙江万雾信息科技有限公司 Dual migration method for tasks in edge computing environment
CN114090221B (en) * 2022-01-20 2022-04-29 浙江万雾信息科技有限公司 Dual migration method for tasks in edge computing environment

Also Published As

Publication number Publication date
CN113364626B (en) 2022-09-30

Similar Documents

Publication Publication Date Title
Ghobaei-Arani et al. A cost-efficient IoT service placement approach using whale optimization algorithm in fog computing environment
CN113364626B (en) Service placement and bandwidth allocation method for video analysis application facing edge environment
CN110069341B (en) Method for scheduling tasks with dependency relationship configured according to needs by combining functions in edge computing
US11184794B2 (en) Systems and methods for distribution of application logic in digital networks
CN113115252B (en) Delay sensitive task distributed mobile edge computing resource scheduling method and system
CN113472844B (en) Edge computing server deployment method, device and equipment for Internet of vehicles
CN109947574B (en) Fog network-based vehicle big data calculation unloading method
CN112291335B (en) Optimized task scheduling method in mobile edge calculation
CN112148492A (en) Service deployment and resource allocation method considering multi-user mobility
CN114265631A (en) Mobile edge calculation intelligent unloading method and device based on federal meta-learning
CN114302407A (en) Network decision method and device, electronic equipment and storage medium
CN113315669B (en) Cloud edge cooperation-based throughput optimization machine learning inference task deployment method
CN111131447A (en) Load balancing method based on intermediate node task allocation
Donassolo et al. Load aware provisioning of IoT services on fog computing platform
CN109460301B (en) Method and system for configuring elastic resources of streaming data load
Majeed et al. Modelling fog offloading performance
Chen et al. A game theoretic approach to task offloading for multi-data-source tasks in mobile edge computing
Murti et al. Learning-based orchestration for dynamic functional split and resource allocation in vRANs
CN114205317A (en) Service function chain SFC resource allocation method based on SDN and NFV and electronic equipment
CN112398917A (en) Real-time task scheduling method and device for multi-station fusion architecture
CN116996941A (en) Calculation force unloading method, device and system based on cooperation of cloud edge ends of distribution network
CN114978913B (en) Cross-domain deployment method and system for service function chains based on cut chains
CN116781532A (en) Optimization mapping method of service function chains in converged network architecture and related equipment
Garg et al. Heuristic and reinforcement learning algorithms for dynamic service placement on mobile edge cloud
Malazi et al. Distributed service placement and workload orchestration in a multi-access edge computing environment

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