HybridDep: An elastic hybrid resources allocation strategy for I/O-intensive applications
Published in IET Communications, 2025
Along with the rapid development of B5G/6G, the number of applications grows rapidly and the data amount explodes exponentially, putting a massive burden on the resource-limited edge servers. To fully utilize the limited resources, virtualization technology is introduced to provide elastic deployment for applications in edge servers. But for I/O-intensive applications, allocating elastic resources is not as easy as for compute-intensive ones, because the amount of required I/O resources is unknown due to the request uncertainty. In this paper, we propose an elastic hybrid deployment strategy HybridDep, which consists of two phases: coarse-grained scheduling and fine-grained tuning. Coarse-grained minimizes deployment costs based on complementary peak and valley patterns in time and space dimensions by integer linear programming. Fine-grained dynamically adjusts each application’s resources allocated to meet the time-varying application resource requirements. Results show that the deployment cost is 3.2% below the Optimal’s while ensuring high quality of service (QoS).
Recommended citation: Pengmiao Li, Yuchao Zhang, Shaoxuan Yun, Fucai Yu, and Aizhi Wu. (2025). "HybridDep: An elastic hybrid resources allocation strategy for I/O-intensive applications." IET Communications.
Download Paper
