MARS: Multipath Adaptive Reliable Service

2026-08-06

AI Summary

Existing multipath transport schemes struggle to balance deployability, path visibility, and congestion feedback latency. This work proposes a receiver-driven, forwarder-assisted multipath transport mechanism built upon an incrementally deployable UDP overlay network, integrating hierarchical synchronized path discovery, ICN-style receiver-driven forwarding, and sender-receiver cooperative congestion control. The design circumvents end-to-end path limitations without modifying IP routing, thereby enhancing path diversity and robustness while supporting flexible deployment across clients, servers, relays, and CDNs. Experimental results demonstrate that, compared to an ECMP baseline, the proposed approach reduces the 95th percentile flow completion time by up to 81.5%. Even when using identical path sets, it significantly outperforms conventional end-to-end schemes, exhibiting superior performance in tail latency, packet loss resilience, and recovery from link failures.

Abstract

Multipath transport is increasingly important for Internet/WAN services that move large data volumes across heterogeneous paths, including geo-distributed analytics, content distribution, and cloud-service pipelines. Existing solutions, however, face a practical trade-off: end-to-end transports such as MPTCP and MPQUIC are deployable but limited by endpoint-visible paths and delayed congestion feedback, while routing-or forwarder-assisted approaches often require infrastructure support or lack safe coordination across forwarding choices. This paper presents MARS, a receiver-driven, forwarder-assisted multipath transport for Internet/WAN environments. MARS combines tier-synchronized overlay path discovery with coupled consumer/forwarder congestion control, enabling it to safely expand usable forwarding opportunities and react near bottlenecks. It runs as an incrementally deployable UDP overlay at clients, servers, relays, or CDN-like nodes. We implement MARS in simulation and as a prototype, and evaluate it through large-scale simulation and Mininet emulation under different deployment scales, loss rates, and failure scenarios. The results show that MARS provides deployment-dependent benefits: with endpoint-only deployment, it remains competitive with end-to-end multipath baselines; with cooperating overlay forwarders, it exposes richer usable path diversity and reduces max p95 flow completion time by up to 81.5\% over ECMP-limited baselines. Even against path-expanded end-to-end baselines given the same path set, MARS achieves lower worst-case p95 FCT and stronger robustness under packet loss, while also recovering quickly from transient link failures. These results demonstrate that ICN-style receiver-driven forwarding can serve as a deployable overlay transport substrate for WAN multipath, providing benefits beyond purely end-to-end designs without requiring changes to IP routing.

Problem

• Multipath transport
• Congestion control
• Forwarder-assisted
• Deployment trade-off
• Path diversity

Innovation

• multipath transport
• receiver-driven forwarding
• forwarder-assisted congestion control
• overlay network
• incremental deployability