Journal
IEEE TRANSACTIONS ON GREEN COMMUNICATIONS AND NETWORKING
Volume 7, Issue 2, Pages 928-939Publisher
IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
DOI: 10.1109/TGCN.2022.3187463
Keywords
Multi-mode green IoT; resource scheduling and route management; multi-timescale optimization; collaborative learning; backpressure awareness
Categories
Ask authors/readers for more resources
This paper proposes a multi-timescale VNF Embedding and floW Scheduling algorithm named NEWS to maximize throughput while reducing VNF embedding cost and energy consumption. The joint optimization problem is transformed into three subproblems, including large-timescale VNF embedding, small-timescale admission control, and small-timescale route selection and computation resource allocation. Simulations demonstrate that NEWS performs superior in terms of throughput, embedding cost, and energy consumption.
The multi-mode green Internet of things (IoT) provides a communication support for social assets of smart park connecting to power grid for low-carbon operation. Software defined networking (SDN) and network function virtualization (NFV) can flexibly integrate heterogeneous communication modes through network resource scheduling and route management. However, the joint optimization of virtual network functions (VNF) embedding and flow scheduling faces several challenges of differentiated QoS guarantee, coupling and externality of VNF embedding, and route selection conflicts. In this work, a multi-timescale VNF Embedding and floW Scheduling algorithm named NEWS is proposed to maximize throughput while reducing VNF embedding cost and energy consumption. Specifically, the joint optimization problem is transformed into three subproblems, i.e., large-timescale VNF embedding, small-timescale admission control, small-timescale route selection and computation resource allocation. A swap matching-based low-cost VNF embedding algorithm is proposed for the first subproblem. Then, a queue backlog threshold-based admission control strategy is proposed for the second subproblem. Next, the third subproblem is decomposed into two stages, where a collaborative Q-learning-based backpressure-aware algorithm is presented in the first stage, and a greedy-based computation resource allocation algorithm is given in the second stage. Simulations demonstrate that NEWS performs superior in throughput, embedding cost, and energy consumption.
Authors
I am an author on this paper
Click your name to claim this paper and add it to your profile.
Reviews
Recommended
No Data Available