.png){kind=icon}
.png){kind=icon}
Return
Delay-aware task scheduling and resource management for DAG-based services
.png){kind=icon}
OA
AI
Y
Y
T
H
DOI：10.1186/s13677-026-00952-z.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
With the rapid development of the Internet of Things (IoT), a growing number of latency-sensitive and computation-intensive applications have emerged, placing increasingly stringent demands on computing capabilities and network resources. Mobile Edge Computing (MEC), which extends cloud computing capabilities to the network edge, has become an effective solution for alleviating network congestion and reducing service latency. To satisfy the quality-of-service (QoS) requirements of diverse applications, joint optimization of task scheduling and resource management is essential. However, the dynamic nature of network environments, the heterogeneity of computing resources, and the diversity of task requirements make this joint optimization problem highly challenging. Furthermore, many practical applications consist of multiple interrelated subtasks with complex dependency relationships, which further increases the complexity of task scheduling and resource allocation. To address these challenges, this paper investigates the low-latency task scheduling and resource allocation problem for Directed Acyclic Graph (DAG)-based services. First, a cloud–edge collaborative architecture is constructed to enable coordinated management of heterogeneous computing resources. Then, the workload, communication, and computation models for DAG-based tasks are developed to accurately characterize task execution latency. Based on these models, a joint optimization problem is formulated with the objective of minimizing overall service latency by comprehensively considering network conditions, resource utilization, and task workload characteristics. To efficiently solve this problem, the task scheduling process is modeled as a Markov Decision Process (MDP). A Deep Reinforcement Learning (DRL) algorithm based on Proximal Policy Optimization (PPO) is proposed to jointly optimize task offloading decisions and resource allocation strategies. Finally, extensive simulation experiments are conducted to compare the proposed approach with four representative baseline schemes. The results demonstrate that the proposed scheme achieves faster convergence and lower service latency.
Keywords:
Edge computing
DAG-based services
Task scheduling
Resource allocation
Journal
J
IF:
4.3
Papers:
724
Citations:
2.2K