NDTAEP: Design of a novel deadline-aware task scheduling model using augmented ensemble pattern analysis

A wide variety of scheduling models have been proposed by researchers over the years, and each of them has varying performance in terms of deadline hit ratio, scheduling effort, efficiency of task mapping, etc. However, these models are highly context-sensitive and cannot be scaled to heterogeneous task types due to their internal mapping characteristics. To improve task scalability, this work proposes a design of a novel deadline-aware task scheduling model that uses augmented ensemble pattern analysis for task clustering. The pattern analysis module uses a combination of K-means, hierarchical, and Fuzzy C Means (FCM) clustering to effectively segregate tasks depending on their completion and deadline parameters. These tasks are given to a modified deadline-aware League Championship Algorithm (LCA) optimizer, which assists in mapping the clustered tasks with worker threads. The modified LCA model uses a combination of task priority, task deadline, and worker capacity for scheduling. The model maps tasks that require higher execution effort with moderately performing worker nodes, while tasks with nearer deadlines are allotted to higher-performance workers. Due to the use of an ensemble augmented pattern analyzer with a modified LCA optimizer, the proposed model can improve execution speed by 8%, deadline hit ratio by 1.5%, and scheduling efficiency by 6.5% when compared with various state-of-the-art scheduling approaches. The proposed model was evaluated and showcased a deadline hit ratio of 99.95%, computational efficiency of 96.26%, and average task-scheduling delay of less than 0.1 ms, which makes it highly useful for a wide variety of task scheduling application scenarios.