Elastic cloud platform resource dynamic scheduling and multi-mode disaster recovery method for smart scenic spot

Abstract

The invention discloses an elastic cloud platform resource dynamic scheduling and multi-mode disaster recovery method for a smart scenic spot. The method comprises the following steps: 1) performing multi-dimensional load prediction and elastic scaling; 2) a cross-regional multi-activity disaster recovery architecture; and 3) unified access and management of the Internet of Things equipment. According to the method, multi-source information such as historical passenger flow data, Internet of Things sensor real-time flow and meteorological data is deeply fused, and the business load fluctuation in the next 24 hours is accurately predicted. And dynamic adjustment of the number of virtual machine instances and the container resource quota is realized by means of a Kubernetes cluster. Meanwhile, a cross-regional multi-activity disaster recovery system is built by adopting a double-center and edge node architecture, database transaction-level real-time synchronization is realized by applying an OGG technology, a robus load balancing layer is built by combining Nginx and Keepalived, and seamless switching of services within less than or equal to 3 seconds when a main center fails is ensured. According to the method, the service continuity, the resource elasticity and the operation and maintenance efficiency of the scenic spot are remarkably improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of the integration of cloud computing and the Internet of Things, and specifically relates to a method for dynamic resource scheduling and multimodal disaster recovery of elastic cloud platforms for smart scenic spots. The method is particularly suitable for dynamic resource scheduling and disaster recovery management in response to holiday traffic peaks, multi-source device access, and sudden failure scenarios in scenic spots. Background Art

[0002] Traditional scenic spot information systems face numerous bottlenecks, severely hindering their efficient operation and development. On the one hand, the static resource allocation model struggles to adapt to cyclical fluctuations in visitor traffic, leading to resource overload and slow system response during peak seasons, while resources remain largely idle during off-seasons, resulting in wasted resources. On the other hand, the single-data center architecture's disaster recovery capabilities are extremely weak. Hardware failures or network outages can paralyze the entire system, resulting in lengthy service recovery times, with average recovery times exceeding 30 minutes. Furthermore, scenic spots contain a wide variety of IoT devices, including heterogeneous terminals such as environmental monitoring, ticket gates, and security cameras. Due to protocol differences, these devices are expensive to access and complex to operate and maintain, significantly increasing the difficulty and cost of scenic spot management. Faced with these thorny issues, an innovative solution is urgently needed that enables dynamic resource scheduling, offers robust disaster recovery capabilities, and supports unified management of multi-protocol devices. Summary of the Invention

[0003] This paper proposes a method for dynamic resource scheduling and multimodal disaster recovery on an elastic cloud platform for smart scenic spots. This method aims to address the problems faced by traditional scenic spot cloud platforms, such as low resource utilization, high risk of single points of failure, and complex device access. This method first constructs a multidimensional load prediction system based on an ARIMA-LSTM hybrid model, integrating historical passenger flow data, real-time traffic from IoT sensors, and meteorological data to accurately predict business load fluctuations over the next 24 hours. Based on the prediction results, the Kubernetes cluster dynamically adjusts the number of virtual machine instances and container resource quotas to optimize resource allocation and improve resource utilization. Secondly, a "dual-center + edge node" architecture is used to build a cross-regional active-active disaster recovery system. GoldenGate technology is used to achieve real-time database transaction-level synchronization. A load balancing layer is constructed using Nginx and Keepalived, ensuring seamless service failover within 3 seconds in the event of a primary center failure. Furthermore, an intelligent gateway supporting multiple protocol conversions is developed. Based on the Spring Cloud microservices framework, full lifecycle management of IoT devices is implemented, including automatic registration, status monitoring, firmware OTA upgrades, and encrypted data transmission. By integrating passenger flow prediction, elastic scaling, intelligent disaster recovery and heterogeneous equipment management, this invention creates a highly elastic and highly available technical foundation for scenic spots, comprehensively improves the service continuity, resource elasticity and operation and maintenance efficiency of scenic spots, and promotes the digital transformation and intelligent upgrading of scenic spots.

[0004] The technical solution for achieving the purpose of the present invention is: A method for dynamic resource scheduling and multimodal disaster recovery of elastic cloud platforms for smart scenic spots includes the following steps: 1) Multi-dimensional load prediction and elastic scaling: Based on the ARIMA-LSTM hybrid model, historical passenger flow data, real-time traffic flow from IoT sensors, and meteorological data are integrated to predict business load fluctuations over the next 24 hours. The number of virtual machine instances and container resource quotas are dynamically adjusted through the Kubernetes cluster. 2) Cross-regional active-active disaster recovery architecture: This architecture utilizes a dual-center + edge node architecture to build primary and backup data centers. GoldenGate (OGG) enables real-time database transaction-level synchronization, and a load balancing layer is built using Nginx + Keepalived. In the event of a primary center failure, service failover is completed in ≤3 seconds using heartbeat detection and a dynamic weight allocation algorithm. 3) Unified access and management of IoT devices: Using an intelligent gateway that supports Zigbee, LoRa, and NB-IoT multi-protocol conversion, and based on the Spring Cloud microservices framework, we implement full device lifecycle management, including automatic registration, status monitoring, firmware OTA upgrades, and encrypted data transmission.

[0005] The multi-dimensional load prediction and elastic scaling steps in step 1) specifically include: 1.1 Data collection and preprocessing: Integrate daily passenger flow data from the ticketing system, parking lot traffic volume, and temperature and humidity data from the meteorological station, and form a multi-dimensional training set after cleaning and standardization; 1.2 Model Training and Optimization: We use the ARIMA model to extract the periodic features of the data in the multi-dimensional training set. We use the LSTM neural network to capture the nonlinear relationships of these features. Using the grid search method, we optimize the hyperparameters of these features to generate an accurate 24-hour load forecast curve. 1.3 Resource Scheduling and Adjustment: When load overcapacity is predicted, Kubernetes HPA is triggered to automatically scale out containers and call the OpenStack API to dynamically add or delete virtual machine nodes to ensure that resources are allocated on demand.

[0006] The steps of the cross-region active-active disaster recovery architecture in step 2) specifically include: 2.1 Data Center Deployment: Build a primary data center and a backup data center, deploy core business systems and databases, and ensure high-speed network connectivity between the primary and backup data centers; 2.2 Data synchronization: GoldenGate technology is used to establish a two-way real-time synchronization link between primary and backup data, ensuring data latency is less than 1 second; 2.3 Traffic Scheduling: Deploy an Nginx Plus global load balancer in the primary data center to intelligently route traffic based on geographic location and implement dynamic load balancing in conjunction with Keepalived. 2.4 Fault Detection and Switching: Prometheus is used to monitor the health status of edge nodes in the primary and backup data centers in real time. When the service response time exceeds 2 seconds or the CPU utilization exceeds 90%, a seamless service switchover is completed within 3 seconds based on heartbeat detection and a weighted algorithm.

[0007] The steps for unified access and management of IoT devices in step 3) specifically include: 3.1 Protocol Conversion: Using intelligent gateways compatible with Zigbee, LoRa, and NB-IoT, with built-in Modbus, MQTT, and CoAP protocol stacks, and using a dynamic plug-in mechanism to adapt to new and old devices to achieve unified JSON / XML data encapsulation; 3.2 Device Access and Registration: A management platform is built based on the Spring Cloud framework, which supports automatic device registration and strengthens identity authentication and authorization. 3.3 Equipment Monitoring and Operation and Maintenance: Build a 3D model of the equipment on the management platform, map and visualize the equipment status in real time, and use the rule engine to set abnormal data alarms, provide remote OTA upgrade functions, and use AI algorithms to provide early warning of equipment failures.

[0008] This technical solution addresses the pain points of smart scenic spot cloud platforms by innovatively integrating elastic computing, intelligent disaster recovery, and IoT technologies. It uses the ARIMA-LSTM model to accurately predict load and integrates Kubernetes to achieve dynamic resource scaling, breaking the limitations of traditional static allocation. It employs a "dual-center + edge node" architecture with GoldenGate data synchronization, combined with Nginx and Keepalived for rapid failover and high service availability. It also develops a multi-protocol intelligent gateway for unified management of IoT devices, reducing access complexity. This approach builds a highly elastic and available cloud platform for scenic spots, effectively promoting their digital transformation and intelligent upgrades. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1 is a flow chart of an embodiment; Figure 2 This is a flowchart of multi-dimensional load prediction and elastic scaling in the embodiment; Figure 3 This is a flow chart of a cross-region multi-active disaster recovery architecture in an embodiment; Figure 4 This is a flow chart of unified access and management of IoT devices in an embodiment. DETAILED DESCRIPTION

[0010] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. Example

[0011] Reference Figure 1 A method for dynamic resource scheduling and multimodal disaster recovery of elastic cloud platforms for smart scenic spots includes the following steps: 1) Multi-dimensional load prediction and elastic scaling: Based on the ARIMA-LSTM hybrid model, historical passenger flow data, real-time traffic flow from IoT sensors, and meteorological data are integrated to predict business load fluctuations over the next 24 hours. The number of virtual machine instances and container resource quotas are dynamically adjusted through the Kubernetes cluster. 2) Cross-regional active-active disaster recovery architecture: This architecture utilizes a dual-center + edge node architecture to build primary and backup data centers. GoldenGate (OGG) enables real-time database transaction-level synchronization, and a load balancing layer is built using Nginx + Keepalived. In the event of a primary center failure, service failover is completed in ≤3 seconds using heartbeat detection and a dynamic weight allocation algorithm. 3) Unified access and management of IoT devices: Using an intelligent gateway that supports Zigbee, LoRa, and NB-IoT multi-protocol conversion, and based on the Spring Cloud microservices framework, we implement full device lifecycle management, including automatic registration, status monitoring, firmware OTA upgrades, and encrypted data transmission.

[0012] like Figure 2 The multi-dimensional load prediction and elastic scaling steps in step 1) specifically include: 1.1 Data collection and preprocessing: Integrate daily passenger flow data from the ticketing system, parking lot traffic volume, and temperature and humidity data from the meteorological station, and form a multi-dimensional training set after cleaning and standardization; 1.2 Model Training and Optimization: We use the ARIMA model to extract the periodic features of the data in the multi-dimensional training set. We use the LSTM neural network to capture the nonlinear relationships of these features. Using the grid search method, we optimize the hyperparameters of these features to generate an accurate 24-hour load forecast curve. 1.3 Resource Scheduling and Adjustment: When load overcapacity is predicted, Kubernetes HPA is triggered to automatically scale out containers and call the OpenStack API to dynamically add or delete virtual machine nodes to ensure that resources are allocated on demand.

[0013] like Figure 3 The steps of the cross-regional active-active disaster recovery architecture in step 2) specifically include: 2.1 Data Center Deployment: Build a primary data center and a backup data center, deploy core business systems and databases, and ensure high-speed network connectivity between the primary and backup data centers; 2.2 Data synchronization: GoldenGate technology is used to establish a two-way real-time synchronization link between primary and backup data, ensuring data latency is less than 1 second; 2.3 Traffic Scheduling: Deploy an Nginx Plus global load balancer in the primary data center to intelligently route traffic based on geographic location and implement dynamic load balancing in conjunction with Keepalived. 2.4 Fault Detection and Switching: Prometheus is used to monitor the health status of edge nodes in the primary and backup data centers in real time. When the service response time exceeds 2 seconds or the CPU utilization exceeds 90%, a seamless service switchover is completed within 3 seconds based on heartbeat detection and a weighted algorithm.

[0014] like Figure 4 The steps for unified access and management of IoT devices in step 3) specifically include: 3.1 Protocol Conversion: Using intelligent gateways compatible with Zigbee, LoRa, and NB-IoT, with built-in Modbus, MQTT, and CoAP protocol stacks, and using a dynamic plug-in mechanism to adapt to new and old devices to achieve unified JSON / XML data encapsulation; 3.2 Device Access and Registration: A management platform is built based on the Spring Cloud framework, which supports automatic device registration and strengthens identity authentication and authorization. 3.3 Equipment Monitoring and Operation and Maintenance: Build a 3D model of the equipment on the management platform to map and visualize the equipment status in real time. The management platform uses a rule engine to set abnormal data alarms, provide remote OTA upgrade functions, and use AI algorithms to provide early warning of equipment failures.

Claims

1. A method for dynamic resource scheduling and multimodal disaster recovery of elastic cloud platforms for smart scenic spots, characterized by: The following steps are involved: 1) Multi-dimensional load prediction and elastic scaling: Based on the ARIMA-LSTM hybrid model, historical passenger flow data, real-time traffic flow from IoT sensors, and meteorological data are integrated to predict business load fluctuations over the next 24 hours. The number of virtual machine instances and container resource quotas are dynamically adjusted through the Kubernetes cluster. 2) Cross-regional active-active disaster recovery architecture: This architecture utilizes a dual-center + edge node architecture to build primary and standby data centers. GoldenGate is used to achieve real-time database transaction-level synchronization. A load balancing layer is built using Nginx + Keepalived. In the event of a primary center failure, service failover is completed in ≤3 seconds based on heartbeat detection and a dynamic weight allocation algorithm. 3) Unified access and management of IoT devices: Using an intelligent gateway that supports Zigbee, LoRa, and NB-IoT multi-protocol conversion, and based on the Spring Cloud microservices framework, we implement full device lifecycle management, including automatic registration, status monitoring, firmware OTA upgrades, and encrypted data transmission.

2. The method for dynamic resource scheduling and multimodal disaster recovery of an elastic cloud platform for smart scenic spots according to claim 1 is characterized in that: The multi-dimensional load prediction and elastic scaling steps in step 1) specifically include: 1.1 Data collection and preprocessing: Integrate daily passenger flow data from the ticketing system, parking lot traffic volume, and temperature and humidity data from the meteorological station, and form a multi-dimensional training set after cleaning and standardization; 1.2 Model Training and Optimization: We use the ARIMA model to extract the periodic features of the data in the multi-dimensional training set. We use the LSTM neural network to capture the nonlinear relationships of these features. Using the grid search method, we optimize the hyperparameters of these features to generate an accurate 24-hour load forecast curve. 1.3 Resource Scheduling and Adjustment: When load overcapacity is predicted, Kubernetes HPA is triggered to automatically scale out containers and call the OpenStack API to dynamically add or delete virtual machine nodes to ensure that resources are allocated on demand.

3. The method for dynamic resource scheduling and multimodal disaster recovery of elastic cloud platforms for smart scenic spots according to claim 1 is characterized in that: The steps of the cross-region active-active disaster recovery architecture in step 2) specifically include: 2.1 Data Center Deployment: Build a primary data center and a backup data center, deploy core business systems and databases, and ensure high-speed network connectivity between the primary and backup data centers; 2.2 Data synchronization: GoldenGate technology is used to establish a two-way real-time synchronization link between primary and backup data, ensuring data latency is less than 1 second; 2.3 Traffic Scheduling: Deploy an Nginx Plus global load balancer in the primary data center to intelligently route traffic based on geographic location and implement dynamic load balancing in conjunction with Keepalived. 2.4 Fault Detection and Switching: Prometheus is used to monitor the health status of edge nodes in the primary and backup data centers in real time. When the service response time exceeds 2 seconds or the CPU utilization exceeds 90%, a seamless service switchover is completed within 3 seconds based on heartbeat detection and a weighted algorithm.

4. The method for dynamic resource scheduling and multimodal disaster recovery of elastic cloud platforms for smart scenic spots according to claim 1 is characterized in that: The steps for unified access and management of IoT devices in step 3) specifically include: 3.1 Protocol Conversion: Using intelligent gateways compatible with Zigbee, LoRa, and NB-IoT, with built-in Modbus, MQTT, and CoAP protocol stacks, and using a dynamic plug-in mechanism to adapt to new and old devices to achieve unified JSON / XML data encapsulation; 3.2 Device Access and Registration: A management platform is built based on the Spring Cloud framework, which supports automatic device registration and strengthens identity authentication and authorization. 3.3 Equipment Monitoring and Operation and Maintenance: Build a 3D model of the equipment on the management platform to map and visualize the equipment status in real time. The management platform uses a rule engine to set abnormal data alarms, provide remote OTA upgrade functions, and use AI algorithms to provide early warning of equipment failures.

Images