A Weighted Bi-objective Strategy for Executing Scientific Workflows in Containerized Environments

Scientific workflows support the execution of complex simulation-based experiments across heterogeneous computing environments. Containerization technologies, such as Docker, improve portability by encapsulating tasks together with their dependencies. However, they also introduce challenges in resource management, as containers incur additional memory and CPU overhead and may execute concurrently on the same virtual or physical machine. These challenges are particularly critical in memory-constrained environments, where inefficient scheduling can lead to performance degradation or even task failures. To address this issue, we propose a weighted bi-objective scheduling strategy that considers memory consumption and execution time, allowing users to prioritize one objective or achieve a balance between the two. Experimental evaluations with both synthetic and real-world workflows demonstrate that our approach enhances performance and resource utilization.