Dynamic network resource allocation method based on network slicing

Patent metadata

Field	Value
Publication number	US-11792137-B2
Application number	US-202217874313-A
Country	US
Kind code	B2
Filing date	Jul 27, 2022
Priority date	Aug 30, 2021
Publication date	Oct 17, 2023
Grant date	Oct 17, 2023

Abstract

A dynamic network resource allocation method based on network slicing is provided. A historical resource demand dataset of an accessed network slice is inputted into a first neural network for training. Based on a trained first neural network and the historical resource demand of the accessed network slice, a resource demand prediction information corresponding to the accessed network slice in a first prediction time period is determined. Resources are pre-allocated to the accessed network slice based on the resource demand prediction information, and resources are allocated to the accessed network slice when the first prediction time period arrives. In this way, the service provider can reasonably allocate network resources for network slices without violating the SLA, thus avoiding the waste of network resources.

First claim

What is claimed is:

1 . A dynamic network resource allocation method based on network slicing, comprising:
S1: inputting a historical resource demand dataset of an accessed network slice into a first neural network for training; S2: determining, based on a trained first neural network and the historical resource demand of the accessed network slice, resource demand prediction information corresponding to the accessed network slice in a first prediction time period; and S3: pre-allocating resources to the accessed network slice based on the resource demand prediction information, and allocating resources to the accessed network slice when the first prediction time period arrives; wherein the resource demand prediction information comprises a predicted node resource quantity and a predicted link resource quantity; wherein the step of pre-allocating resources to the accessed network slice in the step S3 comprises determining a pre-allocated node resource quantity and a pre-allocated link resource quantity of the accessed network slice in the first prediction time period according to the following formulas:
d ˙ i t = γ d d ^ i t , l ˙ l ˙ t = γ l l ^ l ˙ t , ∀ i ∈ 1 , n t ,
wherein if
∑ i = 1 n t γ ′ d d ^ i t < D , then   γ d = 1 + γ ′ d ,   otherwise,   γ d = D ∑ i = 1 n t d ^ i t ; and
if ∑ i = 1 n t γ ′ l l ^ i t < L ,   then   γ l = 1 + γ ′ l , otherwise,     γ l = L ∑ i = 1 n t l ^ i t ;
wherein
d ˙ i t
is a pre-allocated node resource quantity of an accessed network slice i in a first prediction time period t,
l ˙ l ˙ t
is a pre-allocated link resource quantity of the accessed network slice i in the first prediction time period t,
d ^ i t
is a predicted node resource quantity of the accessed network slice i in the first prediction time period t,
l ^ i t
is a predicted link resource quantity of the accessed network slice i in the first prediction time period t, n
t is a quantity of accessed network slices in a system in the first prediction time period t, γ d ≥ 1 represents node resource redundancy, γ l ≥ 1 represents link resource redundancy,
γ ′ d ≥ 0
and
γ ′ l ≥ 0
are offsets of corresponding predicted values, D is a total node resource quantity of the system, and L is a total link resource quantity of the system.

2 . The dynamic network resource allocation method based on network slicing according to claim 1 , further comprising:
A1: receiving access requests of all to-be-accessed network slices in real-time and storing all the access requests in a request queue; A2: determining a total node resource quantity and a total link resource quantity that are occupied by all accessed network slices in a second prediction time period; and A3: when a preset access time point arrives, sequentially deciding whether to allow the access requests in the request queue in the second prediction time period; if yes, allowing a corresponding access request when the second prediction time period arrives; and if not, continuing to decide whether to allow a next access request until all the access requests in the request queue are completely decided.

3 . The dynamic network resource allocation method based on network slicing according to claim 2 , wherein the step of determining the total node resource quantity and the total link resource quantity in the step A2 is performed by using the following formulas:
d ^ s y s T = ∑ i = 1 n T d ˙ i T , l ^ s y s T = ∑ i = 1 n T l ˙ i T , ∀ T ∈ T a c + 1 , T a c + T s y s , wherein
d ^ s y s T
is the total node resource quantity,
l ^ s y s T
is the total link resource quantity, T
sys is a length of a system resource demand prediction window,
d ˙ i T   and   l ˙ i T
are, respectively, a node resource quantity and a link resource quantity to be occupied by an accessed network slice i at a moment T in the second prediction time period
T a c + 1 , T a c + T s y s , T a c
is the preset access time point, and
n T
is a quantity of accessed network slices in the system at the moment T in the second prediction time period.

4 . The dynamic network resource allocation method based on network slicing according to claim 2 , wherein all the network slices are classified into high-quality network slices and low-quality network slices.

5 . The dynamic network resource allocation method based on network slicing according to claim 4 , wherein the step of deciding whether to allow the access requests in the second prediction time period in the step A3 is performed by using the following formulas:
∃ t s t a r t ∈ T a c + 1 , T a c + T a d m , ∀ T ∈ t s t a r t , t s t a r t + T d u r , d ^ s y s T + d m a x ≤ D , l ^ s y s T + l m a x ≤ L , wherein t start is an instantiation time point of a to-be-accessed network slice corresponding to the access request, T ac is a decision-making time point,
d ^ s y s T
and
l ^ s y s T
are, respectively, a total node resource quantity and a total link resource quantity that are occupied by all the network slices at the moment T in the second prediction time period, T
adm is an access window dimension, T dur is survival duration of the to-be-accessed network slice corresponding to the access request, d max and l max are, respectively, an upper limit of allocable node resources and an upper limit of allocable link resources; if a user requests a high-quality network slice, then
d m a x , l m a x = d m a x H , l m a x H ;
if the user requests a standard quality network slice, then
d m a x , l m a x = d m a x S , l m a x S ;
D is a total node resource quantity of the system, L is a total link resource quantity of the system,
d m a x H
is an upper limit of node resources for the high-quality network slices,
l m a x H
is an upper limit of link resources for the high-qua

Assignees

• Univ Electronic Sci & Tech China

Inventors

• Sun Gang
• Li Qing
• Wang Yuhui
• Yu Hongfang
• Sun Jian
• Ren Jing

Classifications

• H04L47/83 Primary

  based on usage prediction · CPC title

• H04L41/147 Primary

  for predicting network behaviour · CPC title

• H04L41/5019 Primary

  Ensuring fulfilment of SLA · CPC title

• H04L41/5025

  by proactively reacting to service quality change, e.g. by reconfiguration after service quality degradation or upgrade · CPC title

• H04L41/0893

  Assignment of logical groups to network elements · CPC title