Research on Resource Optimization and Scheduling Strategy of E-Commerce Poverty Alleviation Platform Under Multi-Cloud Collaborative Architecture

Aiming at the resource optimization problem of e-commerce poverty alleviation platform under multi-cloud architecture, this paper proposes a dynamic priority resource scheduling algorithm (DPRS). The algorithm constructs a hierarchical task classification model, determines the task priority weight through AHP, and introduces Q-learning to realize dynamic allocation of cross-cloud resources. The experiment is based on the CloudSim simulation platform, using a real data set of e-commerce poverty alleviation in a certain province (100,000 tasks). In the cross-cloud bandwidth 5Gbps scenario, the average task completion time of DPRS is 427ms, which is 37.5% shorter than that of genetic algorithm (GA); the SLA violation rate of high-priority tasks is only 0.3%, which is better than FCFS's 7.2%. Cost optimization experiments show that DPRS reduces the cost of cross-cloud migration by 28.3% and improves resource utilization by 26%. In the scalability test, the system supports 50,000 tasks per minute, and the throughput is 41% higher than that of traditional solutions. The results show that DPRS effectively balances the real-time, cost and fairness requirements of the e-commerce poverty alleviation platform, and provides a new solution for the resource management of the smart agriculture system.