REVIEW 3 cited by
STrack: A Reliable Multipath Transport for AI/ML Clusters
Not yet reviewed by Pith; the record is open.
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
STrack: A Reliable Multipath Transport for AI/ML Clusters
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.
Forward citations
Cited by 3 Pith papers
-
OpenURMA: A Clean-Room Open Implementation of the Unified Bus Protocol
OpenURMA is the first clean-room open implementation of the Unified Bus transport and transaction layers, showing ~500 ns end-to-end latency for 64-byte remote loads versus 2186 ns for RoCEv2 RC.
-
A Protocol-Independent Transport Architecture
PITA is a protocol-independent transport architecture that enables full data-path programmability at line rate by structuring the datapath around a uniform abstraction and reworking scheduling, packet generation, and ...
-
Incast-Free MoE Rate-Based Scheduling
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.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.