Fair and efficient online network slice multi-resource allocation method

Abstract

The problems that an existing network slicing method is low in resource utilization rate and tenant satisfaction degree, and tenant privacy is leaked are solved. The invention provides a fair and efficient online slice multi-resource allocation method. Firstly, when a slice request of a tenant arrives, a basic demand vector of the tenant and an income vector obtained by requesting the slice of the type are determined according to the slice type of the request, and the number of the slice requests of each type in an online environment is an unknown quantity; then calculating the capacity of each current resource; and finally, calculating an unknown quantity according to the probability of future tenant demand types to approximately maximize distribution of Nash social welfare. The method can be suitable for the condition of large-scale online requests, and balance between fairness and efficiency is achieved.

Description

Technical Field

[0001] The present invention mainly relates to the field of computer networks, and in particular to the field of virtual network resource allocation. Background Art

[0002] Ubiquitous computing resource pools and intelligent terminal devices provide massive distributed computing resources for various network computing tasks. Computing power network (CPN) is a new type of information infrastructure that includes computing power, storage, and network. The core idea of ​​CPN is to connect multiple scattered computing nodes through network technology, perceive the status of computing resources in real time, uniformly allocate and manage network resources, and realize efficient sharing and intelligent scheduling of computing resources. CPN provides users with high-performance, low-cost, and responsive computing services to meet the growing demand for computing power and network performance in various industries.

[0003] With the continuous development of artificial intelligence technology, computing services such as autonomous driving, large-scale language model training and reasoning have brought unprecedented challenges to network performance, and put forward new requirements for SLA guarantees, resource isolation, network flexibility and scalability of computing networks. Network slicing is a key virtualization technology for computing networks to achieve flexible, secure and efficient computing services.

[0004] However, existing network slicing algorithms are designed for offline environments, and have problems such as low resource utilization, low tenant satisfaction, and tenant privacy leakage. How to fairly and efficiently allocate multiple resources to tenants with different request types in an online environment is an important issue. Based on this, a fair and efficient online multi-resource slicing method is designed. Summary of the invention

[0005] The present invention discloses a fair and efficient online slice multi-resource allocation method, which mainly calculates unknown quantities according to the probability of future tenant demand types to approximate the allocation that maximizes Nash social welfare.

[0006] According to the application background of the present invention, a fair and efficient online slice multi-resource allocation method is provided, comprising the following steps:

[0007] Step 1: Observe the tenant's demand type t and determine its basic demand vector d t and the return vector V t ;

[0008] 1) After observing tenant i’s demand type t, determine its basic demand vector d i =d t ;

[0009] 2) Get the revenue vector V corresponding to the demand vector of tenant i i =V t ;

[0010] Step 2: Calculate the current capacity of each resource

[0011] Step 3: Calculate the number of requests N for each slice type in the online environment t ;

[0012] 1) For tenant i arriving online, if j = i, then

[0013] 2) If j>i, then

[0014] Step 4: Calculate the approximate Nash social welfare maximization allocation; at least the following steps are included:

[0015] 1) First, establish the offline Nash social welfare maximization problem:

[0016]

[0017] 2) Problem P1 is equivalent to:

[0018]

[0019] 3) Use of N in claim 3 t The calculation formula is used to approximately solve problem P2 and obtain the approximate Nash social welfare maximization allocation.

[0020] Compared with the prior art, the advantages of this method are:

[0021] 1. While ensuring the fairness and efficiency of allocation, it can also protect the privacy information of tenants from being leaked to other tenants. Compared with other methods, the algorithm complexity is lower.

[0022] 2. A trade-off between fairness and efficiency is achieved in an online environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 is a flow chart of the present invention;

[0024] Figure 2 It is a schematic diagram of slice division of the present invention;

[0025] Figure 3 It is a system model diagram of the present invention; DETAILED DESCRIPTION

[0026] like Figure 1 As shown, the specific steps of the technical solution of the present invention are:

[0027] Step 1: Observe the tenant's demand type t and determine its basic demand vector d t and the return vector V t;

[0028] 1) After observing tenant i’s demand type t, determine its basic demand vector d i =d t ;

[0029] 2) Get the revenue vector V corresponding to the demand vector of tenant i i =V t ;

[0030] Step 2: Calculate the current capacity of each resource

[0031] Step 3: Calculate the number of requests N for each slice type in the online environment t ;

[0032] 1) For tenant i arriving online, if j = i, then

[0033] 2) If j>i, then

[0034] Step 4: Calculate the approximate Nash social welfare maximization allocation; at least the following steps are included:

[0035] 1) First, establish the offline Nash social welfare maximization problem:

[0036]

[0037] 2) Problem P1 is equivalent to:

[0038]

[0039] 3) Use of N in claim 3 t The calculation formula is used to approximately solve problem P2 and obtain the approximate Nash social welfare maximization allocation.

[0040] In order to verify the effectiveness of the present invention, this method is tested on a python simulation platform. Considering 50 tenants arriving online, the number of resources is set to 3. Multiple online slice requirements are randomly generated. Tenants' requests arrive online. The platform determines its basic demand vector according to the requested slice type, and then immediately makes an irrevocable allocation. The experimental results show that when the number of tenants increases, our method FEoS achieves a trade-off between fairness and efficiency compared to other methods.

Claims

1. Fair and efficient online slicing multi-resource allocation method, characterized by The method comprises at least the following steps: Step 1: Observe the tenant's demand type t and determine its basic demand vector d t and the payoff vector V t ; Step 2: Calculate the current capacity of each resource Step 3: Calculate the number of requests N for each slice type in the online environment t ; Step 4: Calculate the approximate Nash social welfare maximization distribution.

2. The fair and efficient online slice multi-resource allocation method according to claim 1 is characterized in that After observing the tenant type, the following steps are also included: 1) After observing tenant i’s demand type t, determine its basic demand vector d i =d t ; 2) Obtain the revenue vector V corresponding to the demand vector of tenant i i =V t ; 3) Calculate the current capacity of each resource 3. The fair and efficient online slicing multi-resource allocation method according to claim 1 is characterized in that Calculating the number of requests for each slice type in an online environment also includes at least the following steps: 1) For tenant i who arrives online, if j = i, then 2) If j>i, then 4. The fair and efficient online slice multi-resource allocation method according to claim 1 is characterized in that Computing an approximate Nash social welfare-maximizing allocation involves at least the following steps: 1) First, establish the offline Nash social welfare maximization problem: 2) Problem P1 is equivalent to: 3) Use N in claim 3 t The calculation formula is used to approximately solve problem P2 and obtain the approximate Nash social welfare maximization distribution.

Images