Insights and
Performance Research
Field-tested results improving throughput, reliability, and user experience.
BitRipple Technical Report: LT3™ - Efficient Transport for AI Workloads (Intel, CoreWeave, UC Berkeley, BitRipple)
Modern AI training clusters, which may include thousands of GPUs, often run below their potential because of inefficiencies in data movement between nodes. These inefficiencies show up as communication stalls that leave GPUs idle for up to 30% of the time instead of computing. The problem grows worse as models and datasets scale up, or training begins to expand across data centers, because today’s networking methods struggle to balance traffic evenly or recover quickly from packet loss.
Case Study ARA - Iowa State University: Measurement of Dynamics and Liquid Data Transport in OneWeb LEO Satellite Networks
LEO satellite networks like SpaceX, AST SpaceMobile, Amazon LEO, and OneWeb deliver broadband to enterprise and remote environments, but packet loss and inter-satellite handovers can severely degrade TCP performance. This Iowa State University measurement study characterizes OneWeb's real-world latency, throughput, and handover dynamics against its 100 Mbps downlink SLA. Researchers then deployed the BitRipple Tunnel over the live OneWeb link, showing that liquid data transport shields TCP from packet loss, holding congestion windows steady at target rates and cutting retransmissions from 1,767 to just 81. A must-read for enterprises evaluating LEO satellite connectivity. Published October, 2025: IEEE Military Communications Conference (MILCOM).
BitRipple Technical Report: LT3™ FLUID - Fountain LiqUId Delivery
In this paper we introduce BitRipple's FLUID (Fountain LiqUId Delivery) protocol, which combines fountain coding with receiver feedback to deliver data blocks over lossy networks with substantially lower latency than traditional ARQ. Instead of insisting that every packet in a block be received, FLUID introduces a controlled slack parameter ε that lets delivery finish as soon as enough encoded packets have accumulated. Under the Loss-Product Rule, recovery completes once the product of per-round loss fractions falls below ε, so FLUID finishes in a small number of rounds even when every round experiences packet loss. The slack parameter ε directly controls the gap between FLUID and the bandwidth-optimal ARQ baseline, giving operators a tunable trade-off between latency and overhead.
BitRipple Technical Report: LT3™ ARC - Jitter Reduction and Flow Smoothing Innovations
In this paper we introduce BitRipple's Adaptive Release Control (ARC), a lightweight receiver-side scheduling protocol that restores sender-side timing for applications such as cloud gaming, video streaming, telemetry, ML inference, and data transfer. Rather than letting network jitter and recovery dynamics distort delivery, ARC releases recovered data in a way that follows the sender's original timing and ordering, producing smooth delivery with minimal added latency. It runs entirely on the receiver clock and requires no feedback, synchronization, or changes to the underlying transport. The paper shows that integrating ARC into LT3™ removes virtually all large jitter excursions on a cloud-gaming workload and improves perceptual smoothness for transports carried over LT3™, including TCP, QUIC, WebRTC, UDP, and RTP.
Case Study ESnet - US Department of Energy Scientific Network: Optimized Starlink Uplink for Scientific Data Collection
This paper shows how BitRipple Tunnel can dramatically improve Starlink uplink performance for demanding scientific data workflows. In initial ESnet testing, BitRipple more than doubled average upload throughput, raising transfer speeds from a 5.2 Mbps baseline to as high as 14.4 Mbps while cutting upload times from 1,541 seconds to as little as 536 seconds. The results highlight BitRipple’s ability to overcome packet loss, fluctuating bandwidth, and satellite handover instability with a smarter, more resilient approach to transport optimization. For organizations moving critical data from remote environments, this paper offers a compelling look at why BitRipple could redefine uplink performance.
