Embedded Files
One of our recent works entitled “Roracle: Enabling Lookahead Routing for Scalable Traffic Engineering with Supervised Learning” was accepted by The 31st IEEE International Conference on Network Protocols (ICNP 2023) on 7/29/2023. Congratulations!
In this work, we developed a Supervised Learning (SL)-based Graph Neural Network (GNN) framework called Roracle to greatly accelerate Traffic Engineering (TE) operations in today’s large-scale networks. Instead of solving a time-consuming routing optimization problem in real time, Roracle can bypass routing optimization and quickly infer lookahead routing decisions to improve network load balancing performance with good scalability. This work was presented by my PhD advisor, Prof. H. Jonathan Chao, at the ICNP conference in Reykjavik, Iceland, on 10/13/2023.
Project Overview:
(1) Limitations of traditional TE: Poor scalability in today’s large-scale wide-area networks
(2) Proposed TE solution: Roracle and its advantages
(3) GNN framework of Roracle
(4) Evaluation results in real-world large networks and traffic traces
Abstract:
Traditional Traffic Engineering (TE) usually balances the load on network links by formulating and solving a routing optimization problem based on measured Traffic Matrices (TMs). Given that traffic demands could change unexpectedly and significantly in realistic scenarios, routing strategies optimized based on currently measured TMs might not work well in future traffic scenarios. To compensate for the mismatch between stale routing decisions and future TMs, network operators may perform routing updates more frequently, which could introduce significant network disturbance and service disruption. Moreover, given the high routing computation overhead of TE optimization in today’s large-scale networks, routing updates could experience severe delay and thus cannot accommodate future traffic changes in time.
To address these challenges, we propose Roracle, a scalable learning-based TE that quickly predicts a good routing strategy for a long sequence of future TMs, while the learning process is guided by the optimal solutions of Linear Programming (LP) problems using Supervised Learning (SL). We design a scalable Graph Neural Network (GNN) architecture that greatly facilitates training and inference processes to accelerate TE in large networks.
Extensive simulation results on real-world network topologies and traffic traces show that Roracle outperforms existing TE solutions by up to 36% in terms of worst-case performance under future unknown traffic scenarios. Additionally, Roracle achieves good scalability by providing at least 71× speedup over the most efficient baseline method in large-scale networks.
Publications:
[ICNP ’23] Minghao Ye, Junjie Zhang, Zehua Guo, and H. Jonathan Chao, “Roracle: Enabling Lookahead Routing for Scalable Traffic Engineering with Supervised Learning,” The 31st IEEE International Conference on Network Protocols (ICNP), 2023. (Acceptance rate: 18.8%, 34/181) [Paper URL] [PDF]
Report abuse