CN115511585A - Space-time request resource transaction method and system for edge cloud market - Google Patents

Space-time request resource transaction method and system for edge cloud market Download PDF

Info

Publication number
CN115511585A
CN115511585A CN202211179057.3A CN202211179057A CN115511585A CN 115511585 A CN115511585 A CN 115511585A CN 202211179057 A CN202211179057 A CN 202211179057A CN 115511585 A CN115511585 A CN 115511585A
Authority
CN
China
Prior art keywords
user
resource
cpp
resource pool
computational
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.)
Pending
Application number
CN202211179057.3A
Other languages
Chinese (zh)
Inventor
王晓飞
任晓旭
仇超
陈哲远
边高阳
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.)
Tianjin University
Original Assignee
Tianjin 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 Tianjin University filed Critical Tianjin University
Priority to CN202211179057.3A priority Critical patent/CN115511585A/en
Publication of CN115511585A publication Critical patent/CN115511585A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/06—Buying, selling or leasing transactions
    • G06Q30/08—Auctions
    • 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]
    • G06F9/5061—Partitioning or combining of resources
    • G06F9/5072—Grid computing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/02—Marketing; Price estimation or determination; Fundraising
    • G06Q30/0201—Market modelling; Market analysis; Collecting market data
    • G06Q30/0206—Price or cost determination based on market factors
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/06—Buying, selling or leasing transactions
    • G06Q30/0601—Electronic shopping [e-shopping]
    • G06Q30/0611—Request for offers or quotes
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/06—Buying, selling or leasing transactions
    • G06Q30/0601—Electronic shopping [e-shopping]
    • G06Q30/0639—Item locations
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/50—Network service management, e.g. ensuring proper service fulfilment according to agreements
    • H04L41/5003—Managing SLA; Interaction between SLA and QoS
    • H04L41/5019—Ensuring fulfilment of SLA
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2209/00—Indexing scheme relating to G06F9/00
    • G06F2209/50—Indexing scheme relating to G06F9/50
    • G06F2209/5011—Pool
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2209/00—Indexing scheme relating to G06F9/00
    • G06F2209/50—Indexing scheme relating to G06F9/50
    • G06F2209/502—Proximity

Landscapes

  • Business, Economics & Management (AREA)
  • Engineering & Computer Science (AREA)
  • Accounting & Taxation (AREA)
  • Finance (AREA)
  • Theoretical Computer Science (AREA)
  • Development Economics (AREA)
  • Strategic Management (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • General Business, Economics & Management (AREA)
  • Economics (AREA)
  • Marketing (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Software Systems (AREA)
  • Game Theory and Decision Science (AREA)
  • Data Mining & Analysis (AREA)
  • Mathematical Physics (AREA)
  • General Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention discloses a space-time request resource transaction method and a system facing to an edge cloud market, wherein the method comprises the following steps: dividing all computing resources into a plurality of regional computing resource pools according to regions; aiming at maximizing the revenue functions of the user and the CPP, solving by using an incomplete information game model to obtain the optimal bid price of the user and the CPP for the calculation unit, and calculating the expected transaction price of the calculation unit; converting the resource leasing cost minimization problem into a minimization Lyapunov drift plus penalty term problem based on a Lyapunov optimization method, decomposing the minimization problem into a request assignment sub-problem and a resource allocation sub-problem, and respectively solving by using a linear optimization theory and a mixed integer nonlinear programming; based on a discriminative double auction principle and a maximized overall gain function as targets, obtaining CPPs by using a Lagrange relaxation method; the user's guide payment price is calculated based on the melarson optimum auction theory. The invention can better reduce the lease cost of resources and improve the SLA satisfaction.

Description

Space-time request resource transaction method and system for edge cloud market
Technical Field
The invention belongs to the technical field of edge computing, and particularly relates to a space-time request resource transaction method and system for an edge cloud market.
Background
The ubiquitous resource has attracted attention as a base for Artificial Intelligence (AI). With this trend, edge cloud markets are emerging that focus on trading and scheduling computational resources. With the increasing diversification and publicity of computing participants, compelling economic phenomena and recycling mechanisms have come into play, which bring inherent challenges such as market fluctuations, pricing rigor and inefficiencies. Different from other storable commodities such as crops, nonferrous metals, coal, electric power and the like, the computing resources cannot be reserved, and computing trading has the characteristic of instantaneity, but the existing market lacks a unified and coordinated resource trading mechanism, so that the feedback of supply-demand relations between users and Computing Power Providers (CPPs) is delayed, and the market fluctuation is further promoted; because the resource transaction price is generally determined by the platform, resource pricing decision monopoly exists in the computing power transaction, and factors such as space-time difference, supply and demand variation and the like are not considered; the lack of prior knowledge of the resource capability of the computing power provider for the user results in isolation between the user request and the resource supply provided by the computing power provider, and the user's request intent needs to be converted, i.e., the user's service request is converted into corresponding computing resources.
Disclosure of Invention
Aiming at the technical problems, the invention provides a space-time request resource transaction method and system for an edge cloud market, which disclose supply-demand relations through a large cycle and promote computing power transaction through a small cycle. In order to solve the technical problems, the technical scheme adopted by the invention is as follows:
a space-time request resource transaction method facing an edge cloud market comprises the following steps:
s1, dividing computing resources of all resource providers into a plurality of regional computing resource pools according to regions;
s2, aiming at maximizing the revenue function of the user and the revenue function of the CPP, solving a Bayesian Nash equilibrium solution by using an incomplete information game model to obtain the optimal bid price of the user and the CPP counter-calculation unit, and calculating the expected transaction price of the counter-calculation unit according to the optimal bid price;
s3, calculating resource lease cost according to the expected transaction price in the step S2, converting a resource lease cost minimization problem into a minimization Lyapunov drift plus penalty term problem based on a Lyapunov optimization method, decomposing the minimization Lyapunov drift plus penalty term problem into a request assignment sub-problem and a resource allocation sub-problem, and respectively solving the request assignment sub-problem and the resource allocation sub-problem by utilizing a linear optimization theory and mixed integer nonlinear programming;
s4, based on a discrimination double auction principle and a maximized overall revenue function as a target, acquiring a CPP (Power Point) of a power unit for a user in a regional power resource pool corresponding to a resource allocation decision by using a Lagrange relaxation method;
and S5, calculating the guided payment price of the user according to the CPP which provides the calculation force unit for the user and is obtained according to the step S4 on the basis of the Meilsen optimum auction theory.
The step S2 includes the steps of:
s2.1, respectively calculating resource evaluation values of the user to the force calculating unit and resource evaluation values of the CPP to the force calculating unit according to the average SLA requirement of the user and the resource utilization rate of the force calculating unit in the CPP area force calculating resource pool;
the user's calculation formula for the resource estimation of the computational power unit is:
Figure BDA0003858666630000021
in the formula (I), the compound is shown in the specification,
Figure BDA0003858666630000022
representing t-slot user pair regionsResource estimation, P, of class k computational units in computational resource pool j min Minimum trade unit price, P, representing a predetermined calculation force unit max The highest trade unit price representing a preset calculation power unit, ξ represents a time factor, and ξ ∈ [0,1 ]]D (t) represents the average SLA requirement for all users of the t time slot, and δ represents a time scaling factor;
the CPP is used for calculating the resource estimation of the force unit according to the following formula:
Figure BDA0003858666630000023
in the formula (I), the compound is shown in the specification,
Figure BDA0003858666630000024
representing the resource estimate of the t-slot CPP for the class k computational power unit in the region computational power resource pool j,
Figure BDA0003858666630000025
the resource utilization rate of the k-type computational power unit in the region computational power resource pool j of the t-1 time slot CPP is represented;
s2.2, according to the resource estimation in the step S2.1, aiming at maximizing the profit function of the user and the profit function of the CPP, solving a Bayesian Nash equilibrium solution by using an incomplete information game model to obtain the optimal bid price of the user and the CPP for the power calculating unit;
and S2.3, calculating the expected transaction price of the force calculating unit according to the optimal bid price of the user and the CPP.
In step S2.2, the calculation formula of the revenue function of the user is:
Figure BDA0003858666630000026
in the formula (I), the compound is shown in the specification,
Figure BDA0003858666630000027
represents the transaction price of the k-type computational power unit in the computational power resource pool j of the t time slot region,
Figure BDA0003858666630000028
representing the probability of the user expecting the maximum profit;
the calculation formula of the CPP revenue function is as follows:
Figure BDA0003858666630000029
in the formula (I), the compound is shown in the specification,
Figure BDA00038586666300000210
representing the probability of the CPP when the expected revenue is greatest.
In step S2.2, the formula for calculating the optimal bid price for the calculation unit by the user is:
Figure BDA00038586666300000211
the CPP is used for calculating the optimal bid price of the calculation force unit:
Figure BDA0003858666630000031
in the formula, alpha is a constant,
Figure BDA0003858666630000032
represents the optimal bid price of the user to the k-type force computing unit in the region force computing resource pool j at the time slot t,
Figure BDA0003858666630000033
and (4) representing the optimal bid price of the CPP on the k-type computational power unit in the region computational power resource pool j at the time slot t.
The step S3 includes the steps of:
s3.1, calculating the total request delay according to the transmission path requested by the user;
s3.2, obtaining the resource leasing cost which is requested to be processed in the regional computing resource pool by the user according to the expected transaction price, and constructing a resource leasing cost minimization problem P1;
s3.3, converting the long-term resource lease cost minimization problem P1 into a minimization Lyapunov drift plus penalty item problem P2 by a Lyapunov optimization method;
and S3.4, decomposing the minimized Lyapunov drift plus penalty term problem P2 into a short-term request assignment subproblem and a long-term resource allocation subproblem by extracting a relaxation drift-plus penalty term, and solving a request assignment decision and a resource allocation decision by respectively utilizing a linear optimization theory and a mixed integer nonlinear programming.
In step S3.2, the resource lease cost minimization problem P1 is:
P1：
Figure BDA0003858666630000034
s.t.C1：
Figure BDA0003858666630000035
C2：
Figure BDA0003858666630000036
C3：
Figure BDA0003858666630000037
C4：
Figure BDA0003858666630000038
C5：
Figure BDA0003858666630000039
in the formula (I), the compound is shown in the specification,
Figure BDA00038586666300000310
indicating that a user i's request at time slot t is assigned to the regional computing resource pool j,
Figure BDA00038586666300000311
representing the number of k-type computational units of the regional computational resource pool j leased by the user i at t time slot,
Figure BDA00038586666300000312
denotes the CPU frequency of k-type force computing units of the regional force computing resource pool j leased by the user i at T time slot, O (T) denotes the resource lease cost required for T time slot to lease the force computing units of the regional force computing resource pool, T denotes the transaction time,
Figure BDA00038586666300000313
represents the maximum number of k-type computational units in the t-slot region computational resource pool j,
Figure BDA00038586666300000314
indicating the total delay of requests made by user i at time slot t being allocated to regional computing power resource pool j,
Figure BDA00038586666300000315
a maximum delay constraint is indicated that represents the maximum delay,
Figure BDA00038586666300000316
indicating the computational cost of a user i's request in t time slot being allocated for processing in the regional computational resource pool j,
Figure BDA00038586666300000317
a constraint representing the maximum computational cost is represented,
Figure BDA00038586666300000318
a set of regional computing power resource pools is represented,
Figure BDA00038586666300000319
representing a collection of users in the edge cloud market.
In step S3.3, the problem P2 of minimizing lyapunov drift plus penalty term is:
P2：
Figure BDA0003858666630000041
s.t.C1：
Figure BDA0003858666630000042
C2：
Figure BDA0003858666630000043
C3：
Figure BDA0003858666630000044
C4：
Figure BDA0003858666630000045
C5：
Figure BDA0003858666630000046
in the formula (I), the compound is shown in the specification,
Figure BDA0003858666630000047
Figure BDA0003858666630000048
representing the data queues of user i in the regional computational resource pool j at time slot t,
Figure BDA0003858666630000049
the SLA latency virtual queues for user i in regional computing resource pool j representing the t time slot,
Figure BDA00038586666300000410
virtual queue representing the computational cost of user i in the regional computational resource pool j at time slot t, A θ L (Z (t)) represents a Lyapunov drift, L (Z (t)) represents a Lyapunov function at a time slot t, L (Z (t + theta)) represents a Lyapunov function at a time slot t + theta, O (tau) represents a long-term resource lease cost required for an computing unit of a computing resource pool in a lease region of the tau slot, and V is an unadvantaged weight,
Figure BDA00038586666300000411
indicating that a user i's request at time slot t is assigned to the regional computing resource pool j,
Figure BDA00038586666300000412
the number of k-type computing power units of the regional computing power resource pool j leased by the user i in t time slot is represented,
Figure BDA00038586666300000413
the CPU frequency of the k-type computing power unit of the regional computing power resource pool j leased by the user i in the t time slot,
Figure BDA00038586666300000414
represents the maximum number of k-type computational units in the t-slot region computational resource pool j,
Figure BDA00038586666300000415
representing the total delay of requests made by user i in time slot t being allocated to the regional computing power resource pool j,
Figure BDA00038586666300000416
a maximum delay constraint is indicated that represents the maximum delay,
Figure BDA00038586666300000417
indicating the computational cost of a user i's request in t time slot being allocated for processing in the regional computational resource pool j,
Figure BDA00038586666300000418
a constraint representing the maximum computational cost is represented,
Figure BDA00038586666300000419
a set of regional computing power resource pools is represented,
Figure BDA00038586666300000420
representing a collection of users in the edge cloud market.
In step S4, the formula of the discriminative double auction principle is:
Figure BDA00038586666300000421
in the formula, APD i Represents the average price density of user i,
Figure BDA00038586666300000422
represents the optimal bid price of the user to the k-type computational power unit in the regional computational power resource pool j at the time slot t,
Figure BDA00038586666300000423
the region given by CPP z at time slot t represents the unit price of the k-class computational power units available in the power resource pool j,
Figure BDA00038586666300000424
representing the optimal bid price, P, of the CPP on the k-type force computing unit in the regional force computing resource pool j at the time slot t max Representing the highest trading unit price of a preset calculation force unit;
the formula for maximizing the overall revenue function is:
Figure BDA00038586666300000425
s.t.D1：
Figure BDA00038586666300000426
D2：
Figure BDA00038586666300000427
Figure BDA0003858666630000051
in the formula (I), the compound is shown in the specification,
Figure BDA0003858666630000052
representing a set of users in the edge cloud market, B i (t) represents a bid service price of the user i at the t time slot,
Figure BDA0003858666630000053
represents the set of CPPs in the edge cloud market,
Figure BDA0003858666630000054
the number of k-type computing power units of the regional computing power resource pool j leased by the user i in t time slot is represented,
Figure BDA0003858666630000055
representing a set of regional computing power resource pools when
Figure BDA0003858666630000056
Leaf, representing the k-class computing power unit of CPP z in the user i leased region computing power resource pool j when t time slot
Figure BDA0003858666630000057
Then, the user i does not rent the k-type force computing unit of CPP z in the region force computing resource pool j when the time slot t is represented,
Figure BDA00038586666300000515
represents the maximum number of k-type computational power units allocated to the user by CPP z in the regional computational power resource pool j.
In step S5, the calculation formula of the user' S instruction to pay the price is:
Figure BDA0003858666630000058
in the formula, p i (t) indicates that t-slot user i pays the instructional price to be charged for processing the request,
Figure BDA0003858666630000059
a set of regional computational resource pools is represented,
Figure BDA00038586666300000510
a set of computational force unit types is represented,
Figure BDA00038586666300000511
the number of k-type computing power units of the regional computing power resource pool j leased by the user i in t time slot is represented,
Figure BDA00038586666300000512
the unit price of k types of computational units available in the region computational resource pool j given by CPP z at the time slot t is represented,
Figure BDA00038586666300000513
a weight of the winner CPP z' representing the alternative CPP in step S4 to be reselected by the critically paying user i,
Figure BDA00038586666300000514
representing the critical weight.
An edge cloud market-oriented space-time request resource trading system, comprising:
a resource pool division module: the system is used for dividing the computing resources of all CPPs into a plurality of regional computing resource pools according to regions;
an expected transaction price calculation module: the method comprises the steps of calculating an expected transaction price of a force calculating unit in a force calculating resource pool according to the optimal bid price of a user and the optimal bid price of the CPP, wherein the optimal bid price of the user and the optimal bid price of the CPP are obtained by utilizing an incomplete information game model and solving by taking the profit function of the user and the profit function of the CPP as targets;
a decision solving module: the system comprises a calculation module, a linear optimization theory and a mixed integer nonlinear programming, wherein the calculation module is used for calculating the resource lease cost based on the expected transaction price obtained by the expected transaction price calculation module, converting a resource lease cost minimization problem into a minimization lyapunov drift plus penalty problem by a lyapunov optimization method, decomposing the minimization lyapunov drift plus penalty problem into a request assignment subproblem and a resource allocation subproblem, and then respectively solving the request assignment subproblem and the resource allocation subproblem by utilizing the linear optimization theory and the mixed integer nonlinear programming to obtain a request assignment decision and a resource allocation decision;
a CPP confirmation module: the CPP is used for obtaining a calculation capacity unit for the user in a region calculation capacity resource pool corresponding to the resource allocation decision by utilizing a Lagrange relaxation method based on a discriminative double auction principle and a maximized overall profit function as a target;
a guiding payment price calculation module: for solving a user's instructional payment price based on the meierson optimum auction theory.
The invention has the beneficial effects that:
considering the space-time dynamic pricing and the space-time difference of the force calculation unit provider, a trading method based on futures-spot circulation is designed, and further, the expected trading price of the futures in the medium and long term is preliminarily determined; compared with the traditional method, the architecture has obvious advantages in terms of queue stability and social welfare, and has shorter data queues, smaller time queues and lower energy consumption queues.
In addition, the spatio-temporal request matching can convert the SLA requirements of the users into required resources, thereby meeting the business requirements of the users and ensuring the minimum renting cost. In the spot circulation, the calculation resource distribution result can be determined efficiently and objectively by adopting the discrimination double-shooting seller method. The transaction mechanism of the edge cloud market space-time request avoids abnormal bidding behaviors of the user and the CPP, and improves SLA satisfaction of the user, resource matching efficiency and social welfare of a transaction market.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.
FIG. 1 is a schematic diagram of a transaction framework of the present invention.
Fig. 2 is a schematic diagram comparing the present application with other algorithms in terms of calculation cost queue, delay queue and data queue.
Fig. 3 is a diagram showing comparison of performances of the present application and other algorithms affected by a user.
Fig. 4 is a comparison between the present application and other algorithms in terms of market efficiency.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be obtained by a person skilled in the art without inventive effort based on the embodiments of the present invention, are within the scope of the present invention.
The edge cloud is a cloud computing platform constructed on an edge infrastructure based on the core of the cloud computing technology and the capability of edge computing. An elastic cloud platform with comprehensive computing, network, storage, safety and other capabilities at the edge position is formed, an end-to-end technical framework of 'cloud edge end three-body cooperation' is formed with a central cloud and an internet of things terminal, and by putting network forwarding, storage, computing, intelligent data analysis and other works on the edge for processing, response delay is reduced, cloud pressure is relieved, bandwidth cost is reduced, and cloud services such as whole network scheduling and computing power distribution are provided.
A spatio-temporal request resource transaction method facing an edge cloud market is shown in figure 1 and comprises the following steps:
s1, dividing computing resources of all resource providers into a plurality of regional computing resource pools according to regions;
in the changing edge cloud market, when a user and all resource providers (CPPs) trade, the user does not directly exchange force resources, but the user request is converted into space-time request scheduling and resource allocation of a resource pool through edge cloud equipment according to space-time resource price, namely trade price. The request scheduling refers to determining which area is allocated to the computing resource pool when the user request reaches the edge cloud. The resource allocation means that the number of the computing power units and the CPU frequency rented to the user by the edge cloud are determined, and meanwhile, in order to guarantee multiple requirements of the user, the requirements of the user can be met by adjusting the CPU frequency of the computing power units.
Each CPP hask types of force calculation units are used as force calculation resources, each type of force calculation unit has different CPU frequencies, internal memories and storage configurations, and k belongs to [0]K represents the number of force calculating unit types in the edge cloud market, and the force calculating unit types are collected
Figure BDA0003858666630000077
And (4) showing. Because different CPPs may be located in the same or different geographical areas, the computing resources owned by all CPPs are divided into several regional computing resource pools according to the geographical areas, and therefore, each regional computing resource pool contains CPPs of different numbers and types. Now the user will be represented by i, and i ∈ [0]N represents the total number of users in the edge cloud market, and the collection of users in the edge cloud market is adopted
Figure BDA0003858666630000071
Denotes that CPP is represented by z, and z ∈ [0]Z represents the total number of CPPs in the edge cloud market, the set of CPPs in the edge cloud market is represented by Z, the regional computing resource pool is represented by j, and j belongs to [0, J ]]J represents the total number of the regional computing power resource pools in the edge cloud market, and the set of the regional computing power resource pools is adopted
Figure BDA0003858666630000078
And (4) showing.
Tagging user i's request as<h i ，sx，B i >Wherein h is i Representing the task deadline of user i, i.e. the Service Level Agreement (SLA) requirement, s i Represents the task load, i.e. the processor resources occupied by the CPU when executing the requested task, and s i ∈[0，s max ]，s max Represents the maximum value of all task loads, B i Indicating the bid service price of user i in the requested time slot by adjusting B i The user can manipulate bids to optimize their profit. The computing power characteristic information provided by resource provider z in the regional computing power resource pool j is adopted
Figure BDA0003858666630000072
It is shown that the process of the present invention,
Figure BDA0003858666630000073
representing the number of k-class computational units available in the regional computational resource pool J for CPP z,
Figure BDA0003858666630000074
representing the unit price of k types of computational force units available in the regional computational force resource pool j given by CPP z, CPP can optimize its profit by manipulating the unit price of computational force units.
S2, aiming at maximizing the revenue function of the user and the revenue function of the CPP, solving a Bayesian Nash equilibrium solution by using an incomplete information game model to obtain the optimal bid price of the user and the CPP counter-calculation unit, and calculating the expected transaction price of the counter-calculation unit according to the optimal bid price;
in the resource trading market, the resource users and the resource providers bid according to their valuation of the resource, but do not know the quotation of the other party, so that the two have a game relationship, and the step S2 comprises the following steps:
s2.1, respectively calculating resource evaluation values of the user to the force calculating unit and resource evaluation values of the CPP to the force calculating unit according to the average SLA requirement of the user and the resource utilization rate of the force calculating unit in the CPP area force calculating resource pool;
the evaluation of the user is mainly the SLA requirement of the user, while the evaluation of the CPPs is mainly the calculation resource utilization rate, the transaction in the edge cloud market is set to run according to the time slot t, namely, each time slot is subjected to resource allocation, and the t is a positive integer.
The user's calculation formula for the resource estimation of the computational power unit is:
Figure BDA0003858666630000075
in the formula (I), the compound is shown in the specification,
Figure BDA0003858666630000076
resource estimation, P, representing k-class computational power units in the regional computational power resource pool j by t-slot users min Minimum trade unit price, P, representing a predetermined calculation force unit max The highest trade unit price representing a preset calculation power unit, ξ represents a time factor, and ξ ∈ [0,1 ]]D (t) represents the average SLA requirement for all users of the t time slot, and δ represents a time scaling factor. In the present embodiment, ξ = δ =1/2.
The calculation formula of the average SLA requirement d (t) of all users in the t time slot is as follows:
Figure BDA0003858666630000081
the CPP is used for calculating the resource estimation of the force unit according to the following formula:
Figure BDA0003858666630000082
in the formula (I), the compound is shown in the specification,
Figure BDA0003858666630000083
representing the resource estimate of the t-slot CPP for the class k computational power unit in the region computational power resource pool j,
Figure BDA0003858666630000084
and the resource utilization rate of the k-class force computing units in the regional force computing resource pool j of the t-1 time slot CPP is represented, and is obtained by calculating the ratio of the number of the distributed k-class force computing units in the t-1 time slot to the number of all the k-class force computing units in the regional force computing resource pool j. In this embodiment, to avoid malicious contention settings
Figure BDA0003858666630000085
S2.2, according to the resource estimation in the step S2.1, aiming at maximizing the user gain function and the CPP gain function, solving a Bayesian Nash equilibrium solution by using an incomplete information game model to obtain the optimal bid price of the user and the CPP for the calculation force unit;
the calculation formula of the revenue function of the user is as follows:
Figure BDA0003858666630000086
in the formula (I), the compound is shown in the specification,
Figure BDA0003858666630000087
represents the transaction price of the k-type computational power unit in the computational power resource pool j of the t time slot region,
Figure BDA0003858666630000088
represents the optimal bid price of the t time slot user to the k-type computational power unit in the regional computational power resource pool j, and
Figure BDA0003858666630000089
Figure BDA00038586666300000810
a bidding strategy for a user in a time slot to bid a k-type computing power unit in a regional computing power resource pool j is characterized in that the user wants to buy resources at a low price, the bidding strategy is different bidding prices determined by the user, and P {. Is used for representing the probability when the condition is met.
Figure BDA00038586666300000811
Figure BDA00038586666300000812
Representing the probability at which the user expects the maximum benefit.
The calculation formula of the yield function of the CPP is as follows:
Figure BDA00038586666300000813
in the formula (I), the compound is shown in the specification,
Figure BDA00038586666300000814
bidding strategy for expressing k-type force units in t-slot CPP (resource provider-oriented Power) to regional force resource pool j, namely different tender prices determined by the resource provider due to the hope of selling resources at high price by the resource providerThe number of the grids is equal to or less than the number of the grids,
Figure BDA00038586666300000815
represents the optimal bid price of the t time slot CPP to the k-type computational power unit in the regional computational power resource pool j, and
Figure BDA00038586666300000816
Figure BDA00038586666300000817
Figure BDA00038586666300000818
representing the probability of the CPP when the expected revenue is greatest.
The optimal bid price of the user to the computing power unit
Figure BDA00038586666300000819
The calculation formula of (c):
Figure BDA00038586666300000820
in the formula, alpha is a constant,
Figure BDA0003858666630000091
the optimal bid price of the CPP counter calculation power unit
Figure BDA0003858666630000092
The calculation formula of (c):
Figure BDA0003858666630000093
s2.3, calculating the expected transaction price of the force calculating unit according to the optimal bid price of the user and the CPP;
the calculation formula of the expected transaction price of the calculation force unit is as follows:
Figure BDA0003858666630000094
the relationship between the user and the CPP is represented above as an incomplete information game, which exists under bayesian nash equilibrium. Assuming no malicious bids, the user and the CPP each submit a bid price with equal probability within the respective bid policy interval according to their resource valuation, i.e.
Figure BDA0003858666630000095
And
Figure BDA0003858666630000096
s3, calculating resource lease cost according to the expected transaction price in the step S2, converting a resource lease cost minimization problem into a minimized Lyapunov drift plus penalty term problem based on a Lyapunov optimization method, decomposing the minimized Lyapunov drift plus penalty term problem into a request assignment sub-problem and a resource allocation sub-problem, and respectively solving the request assignment sub-problem and the resource allocation sub-problem by utilizing a linear optimization theory and mixed integer nonlinear programming;
s3.1, calculating the total request delay according to the transmission path requested by the user;
the calculation formula of the total request delay is as follows:
Figure BDA0003858666630000097
in the formula (I), the compound is shown in the specification,
Figure BDA0003858666630000098
means the total delay of requests, D, resulting from the allocation of a user i's request in time slot t to the regional computing resource pool j tra (i, j) represents network transmission delay, i.e. transmission delay generated by transmitting request of user i to regional computing resource pool j in t time slot, D que (i, j) represents the queue wait delay in the regional computation resource pool j for a user i request at time slot t.
The network transmission delay D tra (i, j) meterThe calculation formula is as follows:
Figure BDA0003858666630000099
where, mu and v both represent scaling parameters,
Figure BDA00038586666300000910
i.e., the scheduling decision of the user request, when
Figure BDA00038586666300000911
When a request indicating user i at time slot t is not assigned to the regional computing resource pool j, when
Figure BDA00038586666300000912
Indicating that user i's request at time slot t is assigned to the regional computing resource pool j, d i，j Representing the distance of user i from the regional computer resource pool j.
Queue wait delay D assuming one unit of task load needs to consume one slot que The calculation formula of (i, j) is:
Figure BDA00038586666300000913
in the formula (I), the compound is shown in the specification,
Figure BDA00038586666300000914
indicating that t-slot user i is not processing task load in the regional computing power resource pool j,
Figure BDA00038586666300000915
the number of k-type computing power units of the regional computing power resource pool j leased by the user i in t time slot is represented,
Figure BDA00038586666300000916
the CPU frequency of the k-class computational power unit of the regional computational power resource pool j leased by the user i in the t time slot can be represented, and specifically, the frequency can be adjusted through a frequency modulation technology, and a phi tableIndicating the size of a task that can be processed by a unit frequency.
Ignoring the transmission costs, the computational cost of the user's request is mainly generated by the force unit he rents, the formula of which is:
Figure BDA0003858666630000101
Figure BDA0003858666630000102
indicating the computational cost of the user i's request in t time slot assigned to be processed in the regional computational resource pool j.
S3.2, according to expected transaction price
Figure BDA0003858666630000103
Obtaining the resource leasing cost processed by a user request in the regional computing resource pool, and constructing a resource leasing cost minimization problem P1;
the calculation formula of the resource renting cost is as follows:
Figure BDA0003858666630000104
in the formula, O (t) represents a resource lease cost required for leasing the force units of the force resource pool of the region for t slots.
The problem P1 of minimizing the resource lease cost is:
P1：
Figure BDA0003858666630000105
s.t.C1：
Figure BDA0003858666630000106
C2：
Figure BDA0003858666630000107
C3：
Figure BDA0003858666630000108
C4：
Figure BDA0003858666630000109
C5：
Figure BDA00038586666300001010
in the formula (I), the compound is shown in the specification,
Figure BDA00038586666300001011
represents the maximum number of k-type computational units in the t-slot region computational resource pool j,
Figure BDA00038586666300001012
Figure BDA00038586666300001013
Figure BDA00038586666300001014
representing the number of k-type computational power units available in the region computational power resource pool j for the t-slot cpppz,
Figure BDA00038586666300001015
representing a maximum computational cost constraint (higher computational frequency would increase the processing speed of the task and would also bring higher computational cost),
Figure BDA00038586666300001016
representing the maximum delay constraint and T representing the transaction time. Constraint C1 may ensure that a user's request is handled by at most one regional computational resource pool within a time slot, constraints C2 and C3 indicate that the number of computational units required is within the capacity of the regional computational resource pool, and constraints C4 and C5 indicate that the average latency SLA and computational cost should meet threshold requirements. Since the requests of different users arrive randomly, the price of the computing power unit in each region is time-varying. Therefore, it is difficult to fill up the resource allocation per slotSufficient for long-term restraint.
S3.3, converting the long-term resource leasing cost minimization problem P1 into a minimization Lyapunov drift plus penalty term problem P2 by a Lyapunov (Lyapunov) optimization method;
the problem P2 of minimizing Lyapunov drift and adding penalty term is as follows:
P2：
Figure BDA0003858666630000111
s.t.C1：
Figure BDA0003858666630000112
C2：
Figure BDA0003858666630000113
C3：
Figure BDA0003858666630000114
C4：
Figure BDA0003858666630000115
C5：
Figure BDA0003858666630000116
in the formula (I), the compound is shown in the specification,
Figure BDA0003858666630000117
Figure BDA0003858666630000118
represents the data queue of user i in the local computing resource pool j at time slot t,
Figure BDA0003858666630000119
the SLA delay virtual queue for user i in the regional computing resource pool j representing t slots,
Figure BDA00038586666300001110
virtual queue representing the computation cost of user i in the local computation resource pool j at time slot t, Λ θ L (Z (t)) represents Lyapunov drift, and
Figure BDA00038586666300001111
Figure BDA00038586666300001112
l (Z (t)) represents a Lyapunov function at a time slot t, L (Z (t + theta)) represents a Lyapunov function at a time slot t + theta, and
Figure BDA00038586666300001113
o (tau) represents the long-term resource leasing cost required by the computing unit of the computing resource pool of the tau time slot leasing area, and V is a non-negative weight value. The data queue is used for representing the data backlog size of users in the area, the SLA delay virtual queue is used for representing the delay virtual backlog size of the users in the area, and the cost calculation virtual queue is used for representing the cost virtual backlog size of the users in the area.
Data queue of user i in the time slot time domain computing resource pool j
Figure BDA00038586666300001114
The calculation formula of (2) is as follows:
Figure BDA00038586666300001115
in the formula (I), the compound is shown in the specification,
Figure BDA00038586666300001116
representing the data queue of user i in the regional computing resource pool j at time slot t-1,
Figure BDA00038586666300001117
represents the number of k-type computing power units of the regional computing power resource pool j leased by the user i in the t-1 time slot,
Figure BDA00038586666300001118
to representThe CPU frequency of the k-type computational power units of the regional computational power resource pool j leased by the user i in the t-1 time slot,
Figure BDA00038586666300001119
indicating whether a request from user i in t-1 time slot is assigned to the regional computing power resource pool j, s i (t-1) represents the task load of user i during the t-1 time slot, where s i (t-1)≤S max ，
Figure BDA00038586666300001120
Specifically, the update may be performed according to the number of tasks leaving each time slot, that is, the number of requests and the number of newly arriving tasks.
SLA deferred virtual queues
Figure BDA00038586666300001121
And computational cost virtual queues
Figure BDA00038586666300001122
Is constructed to satisfy constraints C5 and C6, wherein,
Figure BDA00038586666300001123
the corresponding calculation formula is:
Figure BDA00038586666300001124
Figure BDA00038586666300001125
in the formula (f) d And f c Is a positive scale factor that is a function of,
Figure BDA00038586666300001126
the SLA latency virtual queue for user i in the regional computing resource pool j representing the t-1 slot time,
Figure BDA00038586666300001127
indicating a t-1 time slot time zoneThe computation costs of users i in the domain computing power resource pool j are virtually queued,
Figure BDA0003858666630000121
indicating the computational cost of the process in which a user i's request at t-1 time slot is allocated to the regional computational resource pool j.
S3.4, decomposing the minimized Lyapunov drift plus penalty term problem P2 into a short-term request assignment subproblem and a long-term resource allocation subproblem by extracting a relaxation drift-plus penalty term, and respectively solving a request assignment decision and a resource allocation decision by utilizing a linear optimization theory and mixed integer nonlinear programming (MINLP);
the request assignment decision is to determine which area is allocated to which the user's request, and the expression of the short-term request assignment problem is:
Figure BDA0003858666630000122
s.t.C1：
Figure BDA0003858666630000123
in the formula (I), the compound is shown in the specification,
Figure BDA0003858666630000124
indicating whether a request for user i at time slot τ is assigned to the regional computing resource pool j, s i (τ) represents the task load of user i at the time slot of τ,
Figure BDA0003858666630000125
the distance between user i and the area computation resource pool j in the time slot of tau is shown. The request assignment problem is a linear programming problem, and the optimal solution of each time slot request assignment can be obtained by utilizing a linear optimization theory, namely
Figure BDA0003858666630000126
The optimal solution of (a).
The resource allocation decision refers to determining the number of computing power units and CPU frequency in the area computing power resource pool allocated by the user, and the expression of the long-term resource allocation problem is as follows:
Figure BDA0003858666630000127
s.t.C2：
Figure BDA0003858666630000128
C3：
Figure BDA0003858666630000129
in the formula (I), the compound is shown in the specification,
Figure BDA00038586666300001210
Figure BDA00038586666300001211
and (4) representing the virtual queue of the computation cost of the user i in the region computation resource pool j in the time slot tau. The resource allocation problem is an MINLP problem, and a learning-based method can be adopted to obtain more accurate and more stable decision evaluation so as to obtain the optimal decision evaluation allocated to the user in each time slot
Figure BDA00038586666300001212
And
Figure BDA00038586666300001213
specifically, independent variables can be extracted through relaxation drift-penalty term, and then P2 is decomposed into two sub-problems, which is the prior art and is not described in detail in this embodiment.
S4, acquiring a CPP (performance-sharing program) providing a performance unit for the user in a regional performance resource pool corresponding to the resource allocation decision by using a Lagrange relaxation method based on a discriminative double auction principle and a maximized overall profit function as a target;
since the user's bid and the charge of the CPP are different, the condition of the discriminative double auction is set by the average price density, and only the user and the CPP satisfying the condition can participate in the transaction. The formula of the discriminative double auction principle is as follows:
Figure BDA0003858666630000131
in the formula, APD i Represents the average price density of user i;
average price density APD of the user i i The calculation formula of (2) is as follows:
Figure BDA0003858666630000132
in the formula, B i (t) represents the bid service price for user i in the t time slot.
The formula for maximizing the overall revenue function is:
Figure BDA0003858666630000133
s.t.D1：
Figure BDA0003858666630000134
D2：
Figure BDA0003858666630000135
D3：
Figure BDA0003858666630000136
in the formula (II)
Figure BDA0003858666630000137
When representing t time slot, user i rents CPP z k type force calculating unit in the region force calculating resource pool j, when
Figure BDA0003858666630000138
When, represents the time slot t, user i does not rent the zoneThe k-class computation power unit of CPP z in the domain computation power resource pool j,
Figure BDA0003858666630000139
and the maximum number of k-type force computing units which are allocated to the user i by the CPP z in the region force computing resource pool j is shown. Constraints D1 and D3 ensure that each user selects only one CPP, while constraint D2 ensures that a CPP can provide resources to multiple users without exceeding its resource capacity.
To maximize the overall welfare, i.e., the total bid revenue minus the total cost, of the user and all CPPs, it is necessary to determine how to select the best match between the user and the CPPs. That is, according to the request assignment decision and the resource allocation decision of the user obtained in step S3, it is further determined which CPP in the local computing resource pool can provide the resource lease service corresponding to the winning user, and the problem of determining the CPP is actually a bipartite graph matching problem with a 0-1 knapsack constraint, and thus can be solved by a lagrange relaxation method.
S5, obtaining a CPP (contact Perkin 'S optimal auction theory) for providing a force calculating unit for the user according to the step S4 and calculating a guide payment price of the user based on the Myerson' S optimal auction theory;
the calculation formula of the user for guiding the payment price is as follows:
Figure BDA00038586666300001310
in the formula (I), the compound is shown in the specification,
Figure BDA00038586666300001311
expressed as a weight of the winner CPP z' of the CPP in the alternative step S4 reselected by the user i to obtain the critical payment,
Figure BDA00038586666300001312
represents the critical weight, i.e. the minimum weight of other users, pi (t) represents the directive payment price that t-slot user i needs to bear for processing the request, the directive payment price is given through the edge cloud market, and the payment price can be paid according to the directiveTo maximize social welfare.
Critical weight
Figure BDA0003858666630000141
The calculation formula of (2) is as follows:
Figure BDA0003858666630000142
in the formula (I), the compound is shown in the specification,
Figure BDA0003858666630000143
represents the weight of other CPPs except the CPP determined in step S3, and
Figure BDA0003858666630000144
weight of
Figure BDA0003858666630000145
The calculation formula of (c) is:
Figure BDA0003858666630000146
the payment price determination problem is to determine the payment amount for each winning user while ensuring authenticity and personal reasonableness, which ensures that each user actually submits an actual valuation. The method comprises the steps of analyzing a trading method in a marginal cloud market based on a futures-spot double-cycle trading mode, firstly, calculating evaluation values of users and CPPs on computing power unit resources respectively, constructing a futures-spot trading framework, and determining an optimal bidding strategy of the users and CPPs (common public service provider) agents and expected trading pricing of each type of computing power unit in a regional computing power resource pool by using an incomplete information game model; then, converting the request of the user into request assignment and resource allocation, and solving by using a Lyapunov optimization method to obtain a winner user, and the corresponding resource allocation quantity and CPU frequency; and setting discrimination bidirectional auction conditions, eliminating unsatisfied CPPs which do not meet the conditions, solving the overall gain function by using a Lagrange relaxation method to obtain the CPP which can provide a calculation unit for the user, and solving a guide payment price. The results show that the architecture of the present application enables more stable queues and less resource lease cost.
An edge cloud market-oriented space-time request resource trading system, comprising:
a resource pool division module: the system is used for dividing computing resources of all resource providers, namely CPPs, into a plurality of regional computing resource pools according to regions;
an expected transaction price calculation module: the method comprises the steps that expected transaction prices of the effort units in an effort resource pool are calculated according to the optimal bid price of a user and the optimal bid price of the CPP, and the optimal bid price of the user and the optimal bid price of the CPP are obtained by solving a target based on an incomplete information game model and by maximizing the profit function of the user and the profit function of the CPP;
a decision solving module: the system comprises a calculation module, a linear optimization theory and a mixed integer nonlinear programming, wherein the calculation module is used for calculating the resource lease cost based on the expected transaction price obtained by the expected transaction price calculation module, converting a resource lease cost minimization problem into a minimization lyapunov drift plus penalty problem by a lyapunov optimization method, decomposing the minimization lyapunov drift plus penalty problem into a request assignment subproblem and a resource allocation subproblem, and then respectively solving the request assignment subproblem and the resource allocation subproblem by utilizing the linear optimization theory and the mixed integer nonlinear programming to obtain a request assignment decision and a resource allocation decision;
a CPP validation module: the CPP is used for obtaining a calculation power unit for the user in a region calculation power resource pool corresponding to the resource allocation decision by utilizing a Lagrange relaxation method based on a discriminative double auction principle and a maximized overall revenue function as a target;
the guiding payment price calculation module: for solving the user's instructional payment price based on Myerson theory.
The present application is verified below using the edge cloud services company a real data set, with an original data set size of about 3.51GB. The data covers more than 1000 million user requests from 10/1/2021 to 10/29/2021. All experiments were performed in Intel (R) Xeon: (R) ((R))R) E5-2690 v4@2.60GHz CPU and NVIDIATesla P100-PCIE-16GB GPU and 64GB RAM on-server simulation. Default N =10,k =4,j =4,z =20,t =500, θ =8,h i ＝0.2，y i ＝0.08，f d ＝0.001，f c =0.01, v =20, phi =100, u = v =1. The algorithm employed in the prior art is compared with the present application as a baseline algorithm based on three aspects: (i) DYRECEIVE: it only considers the quality of the delay experience; (ii) Price Preferred: the adopted fixed calculation unit price is adopted, and meanwhile, the user request is dispatched to the calculation resource pool with the lowest price; (iii) Random, randomly distributes user requests to a computational resource pool. The baseline algorithm for spot transactions includes: (i) The Kuhn-Munkres (KM) algorithm, which aims to find the maximum resource matching between users and CPPs and maximize the sum of all weights; (ii) The discrimination KM algorithm sets discrimination conditions of auction matching; (iii) AVA, greedily selects the matching winner with the largest weight under the capacity constraint.
Fig. 2 shows queue convergence of different algorithms when N = 5. The mechanism considers the current queue information, and can allocate more user requests to the optimal resource pool, so that the data volume and the delay queue length are reduced. In addition, the method and the device have the best performance in terms of calculation overhead while realizing stable queues. As shown in fig. 3, as the number of users increases, the average queue backlog of all three algorithms increases as expected, while the queue backlog of the present application is the lowest. DYRECEIVE performs the worst in terms of computational cost because it only considers SLA latency. As can be seen from (a) and (b) in fig. 3, when N =30, due to the flexible request scheduling, the present application is 74.71%, 24.47%, and 16.95% higher than the DYRECEIVE, price Preferred, and Random algorithms, respectively, on the average data queue, and 79.17%, 27.71%, and 25.37% higher than the DYRECEIVE, price Preferred, and Random algorithms, respectively, on the average latency. Fig. 3 (d) and (e) show satisfaction ratios of the delay SLAs to the calculation cost defined as the average ratio of the number of requests satisfying the demand to the total number of requests. Experiments show that the method and the device have the advantages of optimal performance on user satisfaction and stronger robustness on the number of users. Furthermore, as shown in fig. 3 (f), the performance of the present application on average cost is close to Price Preferred, and in some cases even better. This is because the present application takes into account the spatiotemporal dynamics of prices, which can determine future prices based on the current state of the market.
As shown in fig. 4, the marginal cloud market, i.e., the market efficiency of the present application, is evaluated in terms of both social welfare and individuality. As can be seen from fig. 4 (a), the performance of the present application is 85.71%, 80.51%, and 9.09% higher than those of KM, discriminative KM, and AVA algorithms, respectively. The user who does not accord with the market law is eliminated, and the CPP with more resources and low price can serve more users instead of one user. Further, in (b) in fig. 4, the individual's reasonableness is verified by comparing the true bid with the corresponding payment.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.

Claims (10)

1. The space-time request resource transaction method facing the edge cloud market is characterized by comprising the following steps of:
s1, dividing computing resources of all resource providers into a plurality of regional computing resource pools according to regions;
s2, aiming at maximizing the revenue function of the user and the revenue function of the CPP, solving a Bayesian Nash equilibrium solution by using an incomplete information game model to obtain the optimal bid price of the user and the CPP counter-calculation unit, and calculating the expected transaction price of the counter-calculation unit according to the optimal bid price;
s3, calculating resource lease cost according to the expected transaction price in the step S2, converting a resource lease cost minimization problem into a minimization Lyapunov drift plus penalty term problem based on a Lyapunov optimization method, decomposing the minimization Lyapunov drift plus penalty term problem into a request assignment sub-problem and a resource allocation sub-problem, and respectively solving the request assignment sub-problem and the resource allocation sub-problem by utilizing a linear optimization theory and mixed integer nonlinear programming;
s4, acquiring a CPP (performance-sharing program) providing a performance-calculating unit for the user in a regional performance-calculating resource pool corresponding to the resource allocation decision by using a Lagrange relaxation method based on a discriminative double auction principle and with the aim of maximizing an overall gain function;
and S5, calculating the guided payment price of the user according to the CPP which provides the calculation force unit for the user and is obtained according to the step S4 on the basis of the Meilsen optimum auction theory.
2. The marginal cloud market-oriented spatio-temporal request resource transaction method according to claim 1, wherein the step S2 comprises the steps of:
s2.1, respectively calculating resource evaluation values of the user to the force calculating unit and resource evaluation values of the CPP to the force calculating unit according to the average SLA requirement of the user and the resource utilization rate of the force calculating unit in the CPP regional force calculating resource pool;
the user's calculation formula for the resource estimation of the computational power unit is:
Figure FDA0003858666620000011
in the formula (I), the compound is shown in the specification,
Figure FDA0003858666620000012
resource estimation, P, representing k-class computational power units in the regional computational power resource pool j by t-slot users min Minimum transaction unit price, P, representing a predetermined force unit max The highest trade unit price representing a preset calculation power unit, ξ represents a time factor, and ξ ∈ [0,1 ]]D (t) represents the average SLA requirement for all users of the t time slot, and δ represents a time scaling factor;
the CPP is used for calculating the resource estimation of the force unit according to the following formula:
Figure FDA0003858666620000013
in the formula，
Figure FDA0003858666620000014
Representing the resource estimate of the t-slot CPP for the class k computational power unit in the region computational power resource pool j,
Figure FDA0003858666620000015
representing the resource utilization rate of a k-type force computing unit in a region force computing resource pool j of a t-1 time slot CPP;
s2.2, according to the resource estimation in the step S2.1, aiming at maximizing the user gain function and the CPP gain function, solving a Bayesian Nash equilibrium solution by using an incomplete information game model to obtain the optimal bid price of the user and the CPP for the calculation force unit;
and S2.3, calculating the expected transaction price of the force calculating unit according to the optimal bid price of the user and the CPP.
3. The marginal cloud market-oriented spatio-temporal request resource trading method according to claim 2, characterized in that in step S2.2, the calculation formula of the user' S revenue function is:
Figure FDA0003858666620000021
in the formula (I), the compound is shown in the specification,
Figure FDA0003858666620000022
represents the transaction price of the k-type computational power unit in the computational power resource pool j of the t time slot region,
Figure FDA0003858666620000023
representing the probability of the user expecting the maximum profit;
the calculation formula of the CPP revenue function is as follows:
Figure FDA0003858666620000024
in the formula (I), the compound is shown in the specification,
Figure FDA0003858666620000025
representing the probability of when the CPP is expected to yield the greatest.
4. The marginal cloud market-oriented space-time request resource trading method according to claim 2, wherein in step S2.2, the user' S calculation formula for the optimal bid price of the computing power unit:
Figure FDA0003858666620000026
the CPP is used for calculating the optimal bid price of the calculation force unit:
Figure FDA0003858666620000027
wherein, alpha is a constant, and alpha is a linear,
Figure FDA0003858666620000028
represents the optimal bid price of the user to the k-type force computing unit in the region force computing resource pool j at the time slot t,
Figure FDA0003858666620000029
and (3) expressing the optimal bid price of the CPP on the k-type force computing unit in the region force computing resource pool j at the time slot t.
5. The marginal cloud market-oriented spatio-temporal request resource transaction method according to claim 1, wherein the step S3 comprises the steps of:
s3.1, calculating the total request delay according to the transmission path requested by the user;
s3.2, obtaining the resource leasing cost which is requested to be processed in the regional computing resource pool by the user according to the expected transaction price, and constructing a resource leasing cost minimization problem P1;
s3.3, converting the long-term resource lease cost minimization problem P1 into a minimization Lyapunov drift plus penalty item problem P2 by a Lyapunov optimization method;
and S3.4, decomposing the minimized Lyapunov drift plus penalty term problem P2 into a short-term request assignment subproblem and a long-term resource allocation subproblem by extracting a relaxation drift plus penalty term, and solving a request assignment decision and a resource allocation decision by respectively utilizing a linear optimization theory and a mixed integer nonlinear programming.
6. The marginal cloud market-oriented spatio-temporal request resource transaction method according to claim 5, characterized in that in step S3.2, the resource lease cost minimization problem P1 is:
Figure FDA00038586666200000210
Figure FDA0003858666620000031
Figure FDA0003858666620000032
Figure FDA0003858666620000033
Figure FDA0003858666620000034
Figure FDA0003858666620000035
in the formula (I), the compound is shown in the specification,
Figure FDA0003858666620000036
indicating that a user i's request at time slot t is assigned to the regional computing resource pool j,
Figure FDA0003858666620000037
representing the number of k-type computational units of the regional computational resource pool j leased by the user i at t time slot,
Figure FDA0003858666620000038
represents the CPU frequency of the k-class computational units of the regional computational resource pool j leased by the user i at T time slot, O (T) represents the resource lease cost required for the computational units of the regional computational resource pool leased at T time slot, T represents the transaction time,
Figure FDA0003858666620000039
represents the maximum number of k-type computational units in the t-slot region computational resource pool j,
Figure FDA00038586666200000310
indicating the total delay of requests made by user i at time slot t being allocated to regional computing power resource pool j,
Figure FDA00038586666200000311
a maximum delay constraint is indicated that represents the maximum delay,
Figure FDA00038586666200000312
indicating the computational cost of a user i request at t time slot being allocated for processing in the regional computational resource pool j,
Figure FDA00038586666200000313
a constraint representing the maximum computational cost is represented,
Figure FDA00038586666200000314
a set of regional computing power resource pools is represented,
Figure FDA00038586666200000315
representing a collection of users in the edge cloud market.
7. The edge cloud market-oriented space-time request resource transaction method according to claim 5, wherein in step S3.3, the minimizing Lyapunov drift plus penalty term problem P2 is:
Figure FDA00038586666200000316
Figure FDA00038586666200000317
Figure FDA00038586666200000318
Figure FDA00038586666200000319
Figure FDA00038586666200000320
Figure FDA00038586666200000321
in the formula (I), the compound is shown in the specification,
Figure FDA00038586666200000322
Figure FDA00038586666200000323
use in resource pool j for representing regional computational power in time slot tThe data queue of the user i is set,
Figure FDA00038586666200000324
SLA-deferred virtual queue, Y, representing user i in regional computing resource pool j at time slot t j i (t) represents a virtual queue of the computation cost of users i in the regional computation resource pool j at time slot t, Λ θ L (Z (t)) represents a Lyapunov drift, L (Z (t)) represents a Lyapunov function at a time slot t, L (Z (t + theta)) represents a Lyapunov function at a time slot t + theta, O (tau) represents a long-term resource lease cost required for an computing unit of a computing resource pool in a lease region of the tau slot, and V is an unadvantaged weight,
Figure FDA0003858666620000041
indicating that a user i's request at time slot t is assigned to the regional computing power resource pool j,
Figure FDA0003858666620000042
the number of k-type computing power units of the regional computing power resource pool j leased by the user i in t time slot is represented,
Figure FDA0003858666620000043
the CPU frequency of the k-type computing power unit of the regional computing power resource pool j leased by the user i in the t time slot,
Figure FDA0003858666620000044
represents the maximum number of k-type computational units in the t-slot region computational resource pool j,
Figure FDA0003858666620000045
representing the total delay of requests made by user i in time slot t being allocated to the regional computing power resource pool j,
Figure FDA0003858666620000046
a maximum delay constraint is indicated that represents the maximum delay,
Figure FDA0003858666620000047
indicating the computational cost of a user i request at t time slot being allocated for processing in the regional computational resource pool j,
Figure FDA0003858666620000048
a constraint representing the maximum computational cost is represented,
Figure FDA0003858666620000049
a set of regional computing power resource pools is represented,
Figure FDA00038586666200000410
representing a collection of users in the edge cloud market.
8. The marginal cloud market-oriented spatio-temporal request resource trading method according to claim 1, wherein in step S4, the formula of the discriminative double auction principle is:
Figure FDA00038586666200000411
in the form of APD i Represents the average price density of the user i,
Figure FDA00038586666200000412
represents the optimal bid price of the user to the k-type computational power unit in the regional computational power resource pool j at the time slot t,
Figure FDA00038586666200000413
the unit price of k types of computational units available in the region computational resource pool j given by CPP z at the time slot t is represented,
Figure FDA00038586666200000414
representing the optimal bid price, P, of the CPP on the k-type force computing unit in the regional force computing resource pool j at the time slot t max Representing the highest trading unit price of a preset calculation force unit;
the formula for maximizing the overall revenue function is:
Figure FDA00038586666200000415
Figure FDA00038586666200000416
Figure FDA00038586666200000417
Figure FDA00038586666200000418
in the formula (I), the compound is shown in the specification,
Figure FDA00038586666200000419
representing a set of users in the edge cloud market, B i (t) represents a bid service price for user i in the t slot,
Figure FDA00038586666200000420
represents the set of CPPs in the edge cloud market,
Figure FDA00038586666200000421
the number of k-type computing power units of the regional computing power resource pool j leased by the user i in t time slot is represented,
Figure FDA00038586666200000422
represents a collection of regional computational resource pools when
Figure FDA00038586666200000423
When the time slot is represented, the user i rents a k-type force calculation unit of the CPP z in the region force calculation resource pool j, when
Figure FDA00038586666200000424
Then, the user i does not rent the k-type force computing unit of CPP z in the region force computing resource pool j when the time slot t is represented,
Figure FDA00038586666200000425
and the maximum number of k-type force computing units allocated to the user by CPP z in the region force computing resource pool j is shown.
9. The marginal cloud market-oriented spatio-temporal request resource transaction method according to claim 1, wherein in step S5, the user' S calculation formula for guiding payment of price is:
Figure FDA0003858666620000051
in the formula, p i (t) indicates that t-slot user i pays the price for the guideline to be charged for processing the request,
Figure FDA0003858666620000052
a set of regional computational resource pools is represented,
Figure FDA0003858666620000053
a set of computational force unit types is represented,
Figure FDA0003858666620000054
representing the number of k-type computational units of the regional computational resource pool j leased by the user i at t time slot,
Figure FDA0003858666620000055
the unit price of k types of computational units available in the region computational resource pool j given by CPP z at the time slot t is represented,
Figure FDA0003858666620000056
representing an alternative to re-selection of user i for obtaining a critical paymentA weight of the winner CPP z' of the CPP in the step S4,
Figure FDA0003858666620000057
representing the critical weight.
10. An edge cloud market-oriented space-time request resource trading system, comprising:
a resource pool division module: the system is used for dividing the computing resources of all CPPs into a plurality of regional computing resource pools according to regions;
an expected transaction price calculation module: the method comprises the steps of calculating an expected transaction price of a force calculating unit in a force calculating resource pool according to the optimal bid price of a user and the optimal bid price of the CPP, wherein the optimal bid price of the user and the optimal bid price of the CPP are obtained by utilizing an incomplete information game model and solving by taking the profit function of the user and the profit function of the CPP as targets;
a decision solving module: the system comprises a calculation module, a linear optimization module and a mixed integer nonlinear programming module, wherein the calculation module is used for calculating resource lease cost based on expected transaction price obtained by an expected transaction price calculation module, converting a resource lease cost minimization problem into a minimized Lyapunov drift plus penalty problem by a Lyapunov optimization method, decomposing the minimized Lyapunov drift plus penalty problem into a request assignment sub-problem and a resource allocation sub-problem, and then respectively solving the request assignment sub-problem and the resource allocation sub-problem by utilizing a linear optimization theory and the mixed integer nonlinear programming to obtain a request assignment decision and a resource allocation decision;
a CPP confirmation module: the CPP is used for obtaining a calculation capacity unit for the user in a region calculation capacity resource pool corresponding to the resource allocation decision by utilizing a Lagrange relaxation method based on a discriminative double auction principle and a maximized overall profit function as a target;
the guiding payment price calculation module: for solving a user's instructional payment price based on the meierson optimum auction theory.
CN202211179057.3A 2022-09-22 2022-09-22 Space-time request resource transaction method and system for edge cloud market Pending CN115511585A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202211179057.3A CN115511585A (en) 2022-09-22 2022-09-22 Space-time request resource transaction method and system for edge cloud market

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202211179057.3A CN115511585A (en) 2022-09-22 2022-09-22 Space-time request resource transaction method and system for edge cloud market

Publications (1)

Publication Number Publication Date
CN115511585A true CN115511585A (en) 2022-12-23

Family

ID=84505648

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202211179057.3A Pending CN115511585A (en) 2022-09-22 2022-09-22 Space-time request resource transaction method and system for edge cloud market

Country Status (1)

Country Link
CN (1) CN115511585A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116362850A (en) * 2023-04-06 2023-06-30 天津大学 Resource allocation scheme for meta-universe service
CN117909086A (en) * 2024-03-19 2024-04-19 珠海全志科技股份有限公司 Intelligent control method and system for performing frequency modulation on CPU

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116362850A (en) * 2023-04-06 2023-06-30 天津大学 Resource allocation scheme for meta-universe service
CN117909086A (en) * 2024-03-19 2024-04-19 珠海全志科技股份有限公司 Intelligent control method and system for performing frequency modulation on CPU

Similar Documents

Publication Publication Date Title
Kong et al. Virtual machine resource scheduling algorithm for cloud computing based on auction mechanism
CN115511585A (en) Space-time request resource transaction method and system for edge cloud market
CN104737132B (en) For the message queue in on-demand service environment based on the resource-sharing bidded
Sheikholeslami et al. Auction‐based resource allocation mechanisms in the cloud environments: a review of the literature and reflection on future challenges
Bahreini et al. An envy-free auction mechanism for resource allocation in edge computing systems
CN106817401B (en) Resource allocation method in cloud environment
Wang et al. A reverse auction based allocation mechanism in the cloud computing environment
Yang et al. Envy-free auction mechanism for VM pricing and allocation in clouds
Mashayekhy et al. A two-sided market mechanism for trading big data computing commodities
Liu et al. Game theory-based optimization of distributed idle computing resources in cloud environments
Sharghivand et al. QoS-aware matching of edge computing services to internet of things
Karaca et al. Core-selecting mechanisms in electricity markets
Baranwal et al. A negotiation based dynamic pricing heuristic in cloud computing
Ren et al. Compcube: A space-time-request resource trading framework for edge-cloud service market
CN113781211A (en) Mobile edge computing resource allocation and pricing method based on futures market and spot market
Mohammadi et al. Optimization of virtual machines placement based on microeconomics theory in cloud network
Alsarhan et al. Resource trading in cloud environments for profit maximisation using an auction model
Borissov et al. Automated bidding in computational markets: an application in market-based allocation of computing services
CN114727319A (en) 5G MEC calculation task unloading method based on VCG auction mechanism
CN113535408A (en) Auction-type optimization method for edge side computing resources
Xue et al. All-pay Auction Based Profit Maximization in End-to-End Computation Offloading System
Farajian et al. Parallel Continuous Double Auction for Service Allocation in Cloud Computing
Di et al. Social-optimized win-win resource allocation for self-organizing cloud
Yan et al. A Game‐Theoretic Based Resource Allocation Strategy for Cloud Computing Services
CN109410014A (en) Virtual resource under satellite network based on pricing mechanism shares distribution method and application

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