Bursty workload-adaptive scheduling with data protections for industrial cloud–edge collaboration

Abstract

Industrial Internet applications impose stringent latency and security requirements on cloud–edge collaborative computing. Existing task-scheduling schemes, however, suffer from two critical shortcomings: (i) sensitive attribute data of computing nodes may be leaked or tampered with during transmission or scheduling; (ii) they lack efficient resource–task matching when bursty workloads coincide with limited node capacities. To overcome these challenges, we present the bursty workload-adaptive scheduling mechanism with data protection for industrial cloud–edge collaboration (BWAS-DP). First, we introduce the Zero-Knowledge random range proof based scheduling data protection algorithm (ZKRRP-SDPA), which ensures confidentiality, integrity, and authenticity. It employs an enhanced zero-knowledge range proof for real-valued ranges together with digital signatures, guaranteeing the security of node attribute data against any adversary except the node itself. Second, we propose the congestion prediction and adaptive strategies based task scheduling optimization algorithm (CPAS-TSOA) that combines congestion-aware dynamic resource reservation based on Bayesian-optimized LSTM networks with multi-objective task-scheduling optimization employing task/node-adaptive strategies to enhance decision efficiency. We have implemented a prototype system; extensive experiments on both customized and public industrial datasets show that BWAS-DP reduces average latency by 12.3% under anticipated workloads and by 18.8% under bursty workloads, improves the security metric by approximately 11.1%, and increases decision-making efficiency by roughly 7.2%, thereby delivering lower latency, greater stability, and stronger security.