Privacy-Preserving Remaining Ready Queue Processing Time Prediction Using Federated Learning and Lottery Scheduling

Sudden system failure, cyber-attacks, discrepancies in communication while process execution are been observed very frequently in cloud-edge infrastructure been operated by distributed multiprocessor systems. The purpose is to estimate the ready queue processing remaining time for recovery from any type of disaster, resource allocation backup, and management of service level agreement, in the distributed multiprocessor’s environment. The traditional approach is having some loopholes such as the privacy concerns arises due to centralized data collection and the lottery scheduling approach which was the very basic classical way of dealing in a recent distributed multiprocessors system. The proposed research provides a novelty by presenting privacy preserving remaining ready queue processing time prediction (PPRRQPTP) framework which consists of the federated learning with support of lottery scheduling for the distributed multiprocessor environments. The model proposes a federated sampled ready queue framework by which independently the local nodes estimate the remaining processes times using statistics of the sampled queue, i.e without sharing the raw scheduled data estimates the partially processed jobs and imputed blocked processes. The federated server will be communicated about the aggregated learning updates and the encrypted parameters of the proposed model. The framework includes the Privacy-preserving distributed estimation, Federated Lottery Scheduling (FLS), Adaptive confidence interval prediction, Federated imputation of blocked jobs, and Failure-aware queue analytics. The proposed novel federated estimation strategy for predicting the remaining processing time after system breakdown while minimizing estimation variance and preserving node-level scheduling privacy, which is expected to outperform centralized scheduling estimation methods in terms of scalability, privacy preservation, estimation accuracy, and communication efficiency.