Rain: RDMA-assisted In-Network Scheduling for Microsecond-scale Workloads

2026-06-03T07:23:55Za292f7ebecf6927d2e49dde87c8da38ee5ee9524f98643a46c7950cb51031227
5G802.11bqAUGUSTEBBRIDSKV-cacheLLM-inferenceQUICRDMARPLURLLCcongestion-controlfoundation-modelsin-network-schedulinglive-streamingmicrosecond-latencymmWavemodel-routingmulti-armed-banditnetwork-aware-schedulingone-sided-RDMA-WRITEprogrammable-switchesqueueingschedulingvLLM-router","BigDipper","data-availability","BFT","payment-ch渠道

What happened

Collection of recent systems and networking research: Rain proposes an RDMA-assisted, programmable-switch in-network scheduler that buffers tasks and worker tokens on-switch and uses one-sided WRITE multicasts to pre-write large tasks, improving throughput and tail latency for microsecond-scale workloads (1.75x vs SOTA on RocksDB). Papers include a combinatorial MAB approach for multi-AP IEEE P802.11bq throughput tuning, AUGUSTE — an online-learning MAC scheduler that predicts uplink arrivals to deliver URLLC-like latency with far lower overhead in 5G testbeds, and BBR-Copilot — a QUIC-sideaux

Why it matters

A reviewed impact interpretation has not been published for this record.

Evidence and limitations

Source ID
arxiv_cs_ni
Record identifier
a292f7ebecf6927d2e49dde87c8da38ee5ee9524f98643a46c7950cb51031227
Enrichment time
2026-06-03T07:23:55Z
AI-assisted enrichment
Yes

This record may overlap with other records. Its enrichment can be incomplete or wrong, and machine assistance was used. Validate consequential decisions against the linked source and your own environment.