Pith. sign in

REVIEW 4 cited by

STrack: A Reliable Multipath Transport for AI/ML Clusters

Not yet reviewed by Pith; the record is open.

Re-run · record.json Download PDF Read on arXiv ↗

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2407.15266 v2 pith:GX6CZII6 submitted 2024-07-21 cs.NI

Yanfang Le , Rong Pan , Peter Newman , Jeremias Blendin , Abdul Kabbani , Vipin Jain , Raghava Sivaramu , Francis Matus This is my paper · ORCID
classification cs.NI
keywords strackworkloadscongestionrocev2transportbalancingcollectiveeven
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

Emerging artificial intelligence (AI) and machine learning (ML) workloads present new challenges of managing the collective communication used in distributed training across hundreds or even thousands of GPUs. This paper presents STrack, a novel hardware-offloaded reliable transport protocol aimed at improving the performance of AI /ML workloads by rethinking key aspects of the transport layer. STrack optimizes congestion control and load balancing in tandem: it incorporates an adaptive load balancing algorithm leveraging ECN, while adopts RTT as multi-bit congestion indicators for precise congestion window adjustment. Additionally, STrack facilitates out-of-order delivery, selective retransmission, and swift loss recovery in hardware for multipath environment. The extensive simulation comparing STrack and RoCEv2 demonstrates that STrack outperforms RoCEv2 by up to 6X with synthetic workloads and by 27.4% with collective workloads, even with the optimized RoCEv2 system setup.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 4 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Incast-Free MoE Rate-Based Scheduling

    cs.NI 2026-07 conditional novelty 5.0 of 10

    Round-robin scheduling in MoE dispatch causes the arrival rate at the hottest receiver to grow exponentially, and a demand-normalized rate allocation prevents this by construction.

  2. Unveiling orbital optical chirality through multipolar chiral light-matter interaction

    physics.optics 2025-08 unverdicted novelty 5.0 of 10

    Using focused optical vortex beams on a twisted gold nanorod dimer, the authors measure chiral dichroism that persists where spin-based chirality vanishes, attributing it to an OAM-driven quadrupole interaction.

  3. PRIME: Pseudo-Random Integrated Multi-Part Entropy for Adaptive Packet Spraying in AI/ML Data centers

    cs.NI 2025-07 conditional novelty 5.0 of 10

    A pseudo-random round-robin packet spraying scheme with congestion-aware path penalties improves flow completion time in simulated AI/ML data center networks.

  4. Congestion-Aware Path Selection for Load Balancing in AI Clusters

    cs.NI 2025-06 conditional novelty 5.0 of 10

    Hopper is a host-level, RTT-guided load balancer for RDMA in AI clusters that switches single-path flows to less congested paths when the current path is slow.

Pith tools