Multi-user slice resource allocation method based on competitive game

1. A multi-user network slice resource allocation method based on competitive game, comprising:

S1: modeling a network system as a two-tier architecture of virtual infrastructure network service providers (VInPs) and users, wherein a VInP layer comprises a plurality of VInPs and a user layer comprises a plurality of users;

S2: building a VInP utility model and a user utility model;

S3: dividing a network slice resource into nodes and links for allocation, and building a node power consumption model and a link power consumption model;

S4: determining a revenue of each VInP according to the VInP utility model, the node power consumption model, and the link power consumption model, and determining a revenue of each user according to the user utility model, the node power consumption model, and the link power consumption model;

S5: calculating a total revenue of a slice according to the revenue of the VInP and the revenue of the user, and using the total revenue of the slice as a network model;

S6: solving the network model, wherein the VInP is used as a seller, the user is used as a buyer, the seller determines an initial price according to a total quantity of slice resources, and the buyer bids on the slice; and allocating the slice resources by using a competitive game mechanism; and

S7: implementing the network model on a network.

2. The allocation method according to claim 1 , wherein the VInP utility model in S2 is built as:

G ( p,q )= pq−cq,   (1)

wherein p represents an initial unit price of the slice resources given by the VInP, q represents a quantity of slice resources allocated by the VInP, and c represents a cost unit price of the slice resources;

the user utility model is:

F ( p,q )= u ( q )− l ( p,q )+ v ( q ),   (2)

wherein u(q) represents a utility generated from the slice resources acquired by user, l(p,q) represents a cost expended by the user for the resource, v(q) represents the user satisfaction, u(q)=wln(1+q), l(p,q)=pq,

v

⁡

(

q

)

=

ln

⁡

(

m

+

q

m

)

,

wherein w is a constant greater than 0 and represents a user weight, ln( ) represents a logarithmic function with the mathematical constant e as its base, m represents a quantity of resources requested by the user.

3. The allocation method according to claim 1 , wherein the building a node power consumption model in S3 specifically comprises:

calculating power consumption of a single node:

P i =P i SE +P i RE ,   (3)

wherein P i represents link power consumption of a accessed node i in a slice, P i SE represents transmission power consumption of the accessed node i, and P i RE represents reception power consumption of the accessed node i;

calculating node power consumption of the slice according to power consumption of the single node:

p i s =Σh i,l s p i ,  (4)

wherein h i,l s represents whether the node I is used in a path l, the path represents a complete link from a source node to a destination node, and s represents a label of the slice; and

determining a node price ρ i (p i s ) according to the node power consumption of the slice, wherein the node price is a function of the node power consumption, and ρ i (p i s ) is used as the node power consumption model.

4. The allocation method according to claim 3 , wherein the building a link power consumption model in S3 specifically comprises:

calculating a bandwidth of a link e:

x

e

s

=

∑

l

∈

Ψ

y

l

s

=

∑

l

∈

Θ

g

e

,

l

s

⁢

y

l

s

,

(

5

)

wherein a network controller calculates L s candidate paths from the source node to the destination node that meet user demands, paths from the source node to the destination node are denoted by Ψ and amount to O paths in total, the candidate paths, denoted by Θ, are comprised in all paths from the source node to the destination node, wherein Θ⊆Ψ, Ψ={l 1 , l 2 , . . . , l L s , . . . , l O f }, y l s represents bandwidth allocation on a path l, and g e,l s represents whether a link e is used in the path l of a slice s, l 1 is a first candidate path from the source node to the destination node, l 2 is a second candidate path, l L s is an L s th candidate path, and l O f is an O f th path from the source node to the destination node, wherein a method for calculating the candidate paths comprises:

starting, through using a primal-dual algorithm, from any feasible flow in a network with a flow value x≤v, increasing the flow values of links in the network and modify potentials of nodes; and

iterating the links and the nodes in the network until a flow that meets a predetermined constraint condition is obtained, to obtain a target candidate path, wherein v represents a flow value requested by the user, the flow value being a requested data transmission rate,

wherein if an initial flow value is greater than v, the target candidate path is directly obtained; and a link price ρ e (x e s ) is calculated according to the bandwidth of the link e, wherein the link price is a function of the link bandwidth, ρ e (x e s ) is used as the link power consumption model, and x e s represents the bandwidth of the link e.

5. The allocation method according to claim 1 , wherein the determining a revenue of each VInP according to the VInP utility model, the node power consumption model, and the link power consumption model in S4 comprises:

determining utility obtained by the VInP from a slice s:

Q

p

s

(

x

s

,

p

s

)

=

ϕ

s

(

x

s

,

p

s

,

ρ

)

⁢

r

-

(

∑

l

∈

Θ

φ

i

(

p

i

s

)

⁢

h

i

,

l

s

+

∑

l

∈

Θ

φ

e

(

x

e

s

)

⁢

g

e

,

l

s

)

⁢

r

,

(

6

)

wherein ϕ s (x s ,p s ,ρ)r represents a charge for the slice s provided by the VInP, x s is a bandwidth of the slice s, p s s is power consumption of the slice s, ρ is a price,

(

∑

l

∈

Θ

φ

i

(

p

i

s

)

⁢

h

i

,

l

s

+

∑

l

∈

Θ

φ

e

(

x

e

s

)

⁢

g

e

,

l

s

)

⁢

r

represents a cost for providing a service, φ i (p i s ) represents a price of a node i, φ e (x e s ) represents a price of a link e, h i,l s represents whether the node i is used in a path l, g e,l s represents whether the link e is used in a path l, p i s is node power consumption of the slice s, and r represents a data transmission rate; and determining the revenue of the VInP according to the utility obtained by the VInP for the slice s:

Q

p

=

∑

s

∈

S

Q

p

s

(

x

s

,

p

s

)

,

(

7

)

wherein S represents a set of slices, and a server has a largest profit when the following conditions are met:

max Q p s ( x s ,p s ),

s.t.:x e s ≤c e pro ,

p i s ≤v i pro ,

wherein x e s represents a bandwidth of the link e, c e pro represents a remaining maximum bandwidth available for allocation that is provided by the link e, p i s represents power consumption of the node i in the slice s, and v i pro represents a remaining maximum data transmission rate that is provided and is able to be supported by the node i.

6. The allocation method according to claim 1 , wherein the determining a revenue of each user according to the user utility model, the node power consumption model, and the link power consumption model in S4 comprises:

determining utility of the user:

U s ( r )= w s log( l+r ),   (8)

wherein w s represents a service quality request level of the user, and r represents a data transmission rate;

determining a cost of the user for building the slice according to the node power consumption model and the link power consumption model:

ϕ

s

(

x

s

,

p

s

,

ρ

)

=

∑

e

∈

ξ

s

ρ

e

(

x

e

s

)

+

∑

i

∈

N

s

ρ

i

(

x

i

s

)

,

(

9

)

wherein x e s represents link allocation, p l s represents a node allocation status, ρe(·) represents a functional relationship between a link unit price and a link allocation bandwidth, ρ i (·) represents a relationship between a node unit price and node power consumption,

∑

e

∈

ξ

s

ρ

e

(

x

e

s

)

represents a cost paid by the user for purchasing the link,

∑

i

∈

N

s

ρ

i

(

x

i

s

)

represents a cost paid by the user for purchasing the node; ξ s is a set of links of a slice s, and N s is a set of nodes in the slice s;

determining the revenue of the user according to the utility of the user and the cost of the user for building the slice, wherein the revenue of the user is a total revenue of the user for purchasing all slices:

Q c =Σ s∈S Q c s ,   (10)

wherein a following equation is maximized:

max

⁢

Q

c

s

=

∑

f

∈

K

s

U

s

(

r

f

)

-

ϕ

s

(

x

s

,

p

s

,

ρ

)

⁢

r

,

s

.

t

.

:

⁢

x

e

s

≥

c

e

req

,

p

i

s

≥

v

i

r

⁢

e

⁢

q

,

Q c s represents a revenue from the slice s purchased by the user, S represents a set of slices, r f represents a data transmission rate requested by a user f, a expression following s.t. represents a constraint condition, x e s represents a bandwidth of a link e, C e req represents a bandwidth requested by the user, p i s represents power consumption of a node i in the slice s, v i req represents a node data transmission rate requested by the user, and U s (r f ) is utility of the user f.

7. The allocation method according to claim 5 , wherein S5 specifically comprises:

calculating the total revenue of the slice according to the revenue of the VInP and the revenue of the user:

max

⁢

Q

s

=

max

⁡

(

Q

c

s

+

Q

p

s

)

=

max

⁢

∑

f

∈

K

s

w

s

⁢

log

⁡

(

1

+

∑

l

∈

Θ

q

f

,

l

s

⁢

y

l

s

)

-

(

∑

l

∈

Θ

ϕ

i

(

p

i

s

)

⁢

h

i

,

j

s

+

∑

l

∈

Θ

ϕ

e

(

x

e

s

)

⁢

g

e

,

l

s

)

⁢

r

)

,

(

11

)

s

.

t

.

:

⁢

x

e

s

≥

c

e

req

,

p

i

s

≥

v

i

req

,

x

e

s

≤

c

e

pro

,

p

i

s

≤

v

i

pro

,

wherein Q c s represents the revenue from the slice s purchased by the user, Q p s is the utility obtained by the VInP for the slice s, f is a user, K s is a set of users, w s represents the service quality request level of the user, q f,l s represents whether a request of the user f is carried in the path l, y l s is bandwidth allocation on the path l, Θ is the candidate paths comprised in all the paths from the source node to the destination node, l is a path, φ i (p i s ) represents the price of the node i, h i,l s represents whether the node i is used in the path, φ e (x e s ) represents the price of the link e, g e,l s represents whether the link e is used in the path l, c e req represents the bandwidth requested by the user, v i req represents the node data transmission rate requested by the user, c e pro represents the remaining maximum bandwidth available for allocation that is provided by the link, v i pro represents the remaining maximum data transmission rate that is able to be supported by the node, and the constraint condition comprises a constraint on the user requesting a node and link, and a constraint on the VInP responding to a node and bandwidth constraint.

8. The allocation method according to claim 1 , wherein when the competitive game mechanism is used to allocate the slice resources in S6, a mutual information-based competitive game strategy is adopted, and the revenue of the user is used as a estimated value of the user for the slice, the revenue of the VInP is used as a estimated value for the slice, a estimated value a of the user for the slice is used as a random variable, a value of the slice is also a random variable p, a specific relationship exist between a and p, and I(a,p) represents a degree of correlation between the estimated value for the slice and the value of the slice.

9. The allocation method according to claim 6 , wherein S5 specifically comprises:

calculating the total revenue of the slice according to the revenue of the VInP and the revenue of the user:

max

⁢

Q

s

=

max

⁡

(

Q

c

s

+

Q

p

s

)

=

max

⁢

∑

f

∈

K

s

w

s

⁢

log

⁡

(

1

+

∑

l

∈

Θ

q

f

,

l

s

⁢

y

l

s

)

-

(

∑

l

∈

Θ

ϕ

i

(

p

i

s

)

⁢

h

i

,

j

s

+

∑

l

∈

Θ

ϕ

e

(

x

e

s

)

⁢

g

e

,

l

s

)

⁢

r

)

,

(

11

)

s

.

t

.

:

⁢

x

e

s

≥

c

e

req

,

p

i

s

≥

v

i

req

,

x

e

s

≤

c

e

pro

,

p

i

s

≤

v

i

pro

,

wherein Q c s represents the revenue from the slice s purchased by the user, Q p s is the utility obtained by the VInP for the slice s, f is a user, K s is a set of users, w s represents the service quality request level of the user, q f,l s represents whether a request of the user f is carried in the path l, y l s is bandwidth allocation on the path l, Θ is the candidate paths comprised in all the paths from the source node to the destination node, l is a path, φ i (p i s ) represents the price of the node i, h i,l s represents whether the node i is used in the path, φ e (x e s ) represents the price of the link e, g e,l s represents whether the link e is used in the path l, c e req represents the bandwidth requested by the user, v i req represents the node data transmission rate requested by the user, c e pro represents the remaining maximum bandwidth available for allocation that is provided by the link, v i pro represents the remaining maximum data transmission rate that is able to be supported by the node, and the constraint condition comprises a constraint on the user requesting a node and link, and a constraint on the VInP responding to a node and bandwidth constraint.