.png){kind=icon}
.png){kind=icon}
Return
ROCDSG: A Routing Optimization Framework for DCN
.png){kind=icon}
PRE
AI
DOI：10.1016/j.comnet.2026.112207.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
In recent years, the deep integration of Software-Defined Networking (SDN) and data center networks (DCNs) has provided a programmable foundation for global resource scheduling and path optimization. However, the rapid growth of data-intensive applications has posed a fundamental trade-off in DCNs among Quality of Service (QoS) assurance, load balancing, energy consumption control, and routing computational complexity under dynamic traffic fluctuations: enforcing low latency and strong balancing often incurs substantial global solving overhead, whereas lightweight routing reduces computational cost but tends to cause traffic concentration and degrade energy control performance, making it difficult to consistently guarantee QoS under hotspot and bursty traffic. To address this challenge, this paper proposes ROCDSG, a routing optimization framework based on the collaboration of dynamic subnet partitioning and gateway deployment, aiming to achieve coordinated improvements in QoS, load, and energy objectives with controllable computational complexity. Specifically, a Louvain-based dynamic subnet partitioning mechanism is first introduced to adaptively aggregate network nodes under traffic, load, and energy constraints, thereby shrinking the search space via structural dimensionality reduction. Next, a Mixed-Integer Nonlinear Programming (MINLP) model is formulated for gateway deployment to optimize border-gateway configuration for inter-subnet forwarding and is efficiently solved using the proposed Solution Space-Optimized Firefly Algorithm (SOFA). Finally, a latency-aware hierarchical routing algorithm is designed based on the resulting subnet–gateway skeleton to construct low-latency and stable paths for both intra- and inter-subnet traffic. Simulation results show that, compared with the next-best algorithm, ROCDSG reduces the end-to-end average latency by 25.2%, decreases the average hop count and execution time by 14.4% and 25.6%, respectively, while effectively maintaining balanced load distribution and energy consumption.
Journal
IF:
4.6
Papers:
1.5K
Citations:
1.6W