To comply with privacy regulations and self-protection from cloud outages, increasingly more users are leveraging multiple cloud providers for service deployment. However, due to the isolation of highly Non-Identically and Independently Distributed (Non-IID) workload datasets and deficiencies of existing methods as reflected in our experimental studies, no cloud providers can single-handedly capture the workload patterns of such users for accurate workload prediction. Accordingly, we proposeP3Forecast, aPersonalizedPrivacy-Preserving Cloud workload prediction framework based on Federated Generative Adversarial Networks (GANs), which allows cloud providers with Non-IID workload data to collaboratively train workload prediction models as preferred while protecting privacy. We first design a data synthesis quality assessment method based on Dynamic Time Warping (calledpattern-aware DTW), which is insusceptible to time series length and reliable for the comparison of temporal patterns. By usingpattern-aware DTWas the model aggregation weights, we adopt the Federated Learning (FL) of a GAN model for the augmentation of IID workload training datasets per cloud provider. Then, we further design a post-training method of local workload prediction models, which consists of a query mechanism based on comprehensive evaluation of data synthesis informativeness and an adaptive learning rate adjustment strategy for stable convergence. Extensive experiments driven by real-world workloads demonstrate that compared with the state-of-the-art,P3Forecastimproves workload prediction accuracy by 23.7%-64.5% on average over all cloud providers, while ensuring the fastest convergence in Federated GAN training.

The workload data generated in large-scale cloud environments is becoming increasingly complex, making collaborative training a promising approach for developing more efficient task schedulers. Considering privacy security and transfer costs, Federated Reinforcement Learning (FRL) emerges as a promising solution. However, our exploratory experiments demonstrate that the environmental heterogeneity contributes to performance degradation in FRL, which makes the above issue challenging. Accordingly, we propose a Personalized FRL method based on Dual-critic networks and Multi-head attention aggregator (PFRL-DM), which achieves the optimal scheduling policies by collaborative training on diverse workload data in heterogeneous environments without exposing private data. We initially introduce a novel Reinforcement Learning (RL) environment modeling, serving as a foundation for the collaborative training of the cloud scheduling agents. Then, we implement a dual-critic network Proximal Policy Optimization (PPO) algorithm for each client, effectively balancing the influence between global and local models on the agents. Furthermore, we integrate multi-head attention weights into the server-side aggregator to implement personalization. Extensive experiments on various real-world workloads have demonstrated that, compared to state-of-the-art algorithm MFPO, the proposed algorithm exhibits faster convergence, shorter response and completion times, and achieves the highest resource utilization. Additionally, the PFRL-DM algorithm constructs personalized models for each client, enabling greater adaptability in heterogeneous and hybrid workload environments. The codes for this paper can be found at https://github.com/liyan2015/PFRL-DM.