Resource configuration method and apparatus for heterogeneous cloud services

Patent metadata

Field	Value
Publication number	US-11449774-B2
Application number	US-201916700651-A
Country	US
Kind code	B2
Filing date	Dec 2, 2019
Priority date	Jun 19, 2019
Publication date	Sep 20, 2022
Grant date	Sep 20, 2022

Abstract

A resource configuration method and apparatus for heterogeneous cloud services are provided. The method may include: establishing a basic model with a general structure for at least two heterogeneous cloud services, where the basic model comprises a trend model and a periodic model; determining a cloud service in the at least two cloud services as a target cloud service, and acquiring a target historical data set of the target cloud service; training the trend model and the periodic model using the target historical data set; generating a target prediction model corresponding to the target cloud service based on the trained trend model and the trained periodic model; and generating, based on the target prediction model, a resource amount demanded by the target cloud service in a future time period, and configuring resources for the target cloud service according to the demanded resource amount.

First claim

What is claimed is:

1. A resource configuration method for heterogeneous cloud services, comprising:
establishing a basic model with a general structure for at least two heterogeneous cloud services, wherein the basic model comprises a trend model and a periodic model;
determining a cloud service in the at least two cloud services as a target cloud service, and acquiring a target historical data set of the target cloud service;
training the trend model and the periodic model using the target historical data set, to generate a trained trend model and a trained periodic model;
generating a target prediction model corresponding to the target cloud service based on the trained trend model and the trained periodic model; and
generating, based on the target prediction model, a resource amount demanded by the target cloud service in a future time period, and configuring resources for the target cloud service according to the demanded resource amount.

2. The method according to claim 1 , wherein the generating a target prediction model corresponding to the target cloud service based on the trained trend model and the trained periodic model comprises:
acquiring an initial trend weight and an initial periodic weight, wherein the initial trend weight is used to weight the first output result of the trained trend model, and the initial periodic weight is used to weight the second output result of the trained periodic model; and
adjusting, with the target prediction model converges as a goal, the initial trend weight and the initial periodic weight using the target historical data set of the target cloud service, to generate a trend weight and a periodic weight.

3. The method according to claim 2 , wherein the adjusting, with the target prediction model converges as a goal, the initial trend weight and the initial periodic weight using the target historical data set of the target cloud service, to generate a trend weight and a periodic weight comprises:
performing the following first adjustment step based on the initial trend weight: increasing the initial trend weight to obtain a new initial trend weight, and determining whether the target prediction model converges based on the new initial trend weight; determining, in response to the target prediction model converging based on the new initial trend weight, whether a first precision of the target prediction model is smaller than a preset first precision threshold; ending the adjustment in response to the first precision of the target prediction model being smaller than the preset first precision threshold; and
performing, in response to the first precision of the target prediction model being not smaller than the preset first precision threshold, the first adjustment step with the new trend weight as an initial trend weight.

4. The method according to claim 3 , wherein the adjusting, with the target prediction model converges as a goal, the initial trend weight and the initial periodic weight using the target historical data set of the target cloud service, to generate a trend weight and a periodic weight comprises:
performing, in response to the target prediction model not converging based on the new initial trend weight, the following second adjustment step based on the initial periodic weight: increasing the initial periodic weight to obtain a new initial periodic weight, and determining whether the target prediction model converges based on the new initial periodic weight; determining, in response to the target prediction model converging based on the new initial periodic weight, whether a second precision of the target prediction model is smaller than the preset first precision threshold; ending the adjustment in response to the second precision of the target prediction model being smaller than the preset first precision threshold; and
performing, in response to the second precision of the target prediction model being not smaller than the first precision threshold, the second adjustment step with the new periodic weight as an initial periodic weight.

5. The method according to claim 4 , wherein the adjusting, with the target prediction model converges as a goal, the initial trend weight and the initial periodic weight using the target historical data set of the target cloud service, to generate a trend weight and a periodic weight also comprises:
adjusting the preset first precision threshold in response to the target prediction model not converging based on the new initial periodic weight, and continuing to perform the first adjustment step based on the adjusted preset first precision threshold.

6. The method according to claim 2 , wherein the adjusting, with the target prediction model converges as a goal, the initial trend weight and the initial periodic weight using the target historical data set of the target cloud service, to generate a trend weight and a periodic weight comprises:
performing the following third adjustment step based on the initial periodic weight: increasing the initial periodic weight to obtain a new initial periodic weight, and determining whether the target prediction model converges based on the new initial periodic weight; determining, in response to the target prediction model converging based on the new initial periodic weight, whether a third precision of the target prediction model is smaller than a preset second precision threshold; ending the adjustment in response to the third precision of the target prediction model being smaller than the preset second precision threshold; and
performing, in response to the third precision of the target prediction model being not smaller than the second precision threshold, the third adjustment step with the new periodic weight as an initial periodic weight.

7. The method according to claim 6 , wherein the adjusting, with the target prediction model converges as a goal, the initial trend weight and the initial periodic weight using the target historical data set of the target cloud service, to generate a trend weight and a periodic weight comprises:
performing, in response to the target prediction model not converging based on the new initial periodic weight, the following fourth adjustment step based on the initial trend weight: increasing the initial trend weight to obtain a new initial trend weight, and determining whether the target prediction model converges based on the new initial trend weight; determining, in response to the target prediction model converging based on the new initial trend weight, whether the fourth precision of the target prediction model is smaller than the preset second precision threshold; ending the adjustment in response to the fourth precision of the target prediction model being smaller than the preset second precision threshold; and
performing, in response to the fourth precision of the target prediction model being not smaller than the preset second precision threshold, the fourth adjustment step with the new trend weight as an initial trend weight.

8. The method according to claim 7 , wherein the adjusting, with the target prediction model converges as a goal, the initial trend weight and the initial periodic weight using the target historical data set of the target cloud service, to generate a trend weight and a periodic weight also comprises:
adjusting the preset second precision threshold in response to the target prediction model not converging based on the new initial trend weight, and continuing to perform the third adjustment step based on the adjusted preset second precision threshold.

9. The method according to claim 1 , wherein the basic model comprises a burst factor model, wherein an input of the burst factor model is a specified burst demanded amount, and a third ou

Assignees

• Beijing Baidu Netcom Sci & Tech Co Ltd

Inventors

• Chen Xiaoxu
• Liu Feng
• Yang Tao
• Gao Xiang
• Li Geng
• Zhang Xianghui

Classifications

• H04L41/0803

  Configuration setting · CPC title

• G06F30/27

  using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model · CPC title

• G06F9/5077

  Logical partitioning of resources; Management or configuration of virtualized resources (specific details on emulation or internal functioning of virtual machines G06F9/455) · CPC title

• G06F2209/5011

  Pool · CPC title

• H04L41/145

  involving simulating, designing, planning or modelling of a network · CPC title