Aug 2026· Conference on Applications, Technologies, Architectures, and Protocols for Computer Communication· 0 citations· 50 references
Computer Science
TL;DR
InfiniFlow is presented, a credit-based hop-by-hop flow control method that supports massive VCs with a limited buffer budget via per-port buffer sharing, and introduces a paradigm shift in buffer management: Upstream Allocates Buffer for Downstream (UABD).
Abstract
Modern datacenters increasingly rely on Remote Direct Memory Access (RDMA) for high throughput and low latency. To ensure lossless transmission, RDMA fabrics employ hop-by-hop flow control at the granularity of Virtual Channels (VCs). However, multiplexing multiple flows within the same VC couples their backpressure behavior, leading to severe congestion spreading, degrading throughput and tail latency significantly. Existing methods try to improve isolation by increasing the number of VCs but require per-VC buffer reservation, causing buffer requirements to scale linearly with the number of VCs and fundamentally limiting VC scalability. We present InfiniFlow, a credit-based hop-by-hop flow control method that supports massive VCs with a limited buffer budget via per-port buffer sharing. InfiniFlow introduces a paradigm shift in buffer management: Upstream Allocates Buffer for Downstream (UABD), where, for each link, the upstream scheduler maintains a per-port shared credit pool and dynamically allocates credits to VCs. Building on UABD, InfiniFlow further incorporates a Buffer Usage Control Protocol that enforces per-VC dynamic credit thresholds using real-time backlog feedback to bound credit consumption and prevent buffer monopolization. With InfiniFlow, the required buffer size per port becomes independent of the number of VCs and is instead bounded by a small budget proportional to per-hop Bandwidth-Delay Product (BDP) (approximately 3× BDP). We implement InfiniFlow on a Xilinx Alveo U280 FPGA and demonstrate support for over 16k VCs using only 401 KB of on-chip memory per port while sustaining 100 Gbps line-rate transmission, achieving 512× and 1024× higher VC scalability than CBFC and PFC, respectively. Extensive NS-3 simulations show that InfiniFlow achieves near-optimal flow completion times under realistic workloads without relying on slow end-to-end rate adaptation, outperforming prior hop-by-hop flow control methods and state-of-the-art congestion control schemes.
This work introduces CSIG, a protocol that delivers precise, multi-bit bottleneck congestion signals via a fixed-length Ethernet header, and proposes Fast Ramp-Up, a congestion control primitive that leverages these bottleneck signals to reduce median RPC latency by 20% and unclaimed bandwidth by 60% in production.
Abhiram Ravi, Nandita Dukkipati, Weiwu Pang et al.· Conference on Applications,...· 0 citations
ProLet is a load balancing scheme that enables proactive probing and reroutes elephant flows at flowlet granularity in lossless RDMA networks and reduces average and tail flow completion time slowdowns by 69% and 79%, respectively, compared to state-of-the-art load balancing schemes.
Hong Wang, Jin-Hao Luo, J. Tan et al.· Asia-Pacific Workshop on Net...· 0 citations
Datacenter performance is often limited by network-centric congestion controls relying on low-level metrics (e.g., packet loss, latency) that misinterpret applications needs. This work argues that applications should participate in congestion control decisions and introduces the starvation ratio (SR), a metric that detects when applications are truly limited by the network. Experimental evaluations within an 11-flow bottleneck scenario on a Linux-based prototype show that asynchronous applications can absorb network variations within a newly identified “silence zone” without degradation, proving conventional controls are overly restrictive. By deploying proactive and reactive mechanisms, our approach consistently reduces Flow Completion Time (FCT) for network-sensitive workloads. Notably, the proactive configuration eliminates micro-recovery delays, keeping the starvation ratio close to zero and reinforcing the baseline protocol through stable congestion window regulation. We conclude that shifting to application-driven signaling aligns network transmission with the receiver’s processing pace, preventing computational underutilization.
Anderson Henrique da Silva Marcondes, Enzo B. Boscatto, Guilherme Piêgas Koslovski· Anais do LIII Seminário Inte...· 0 citations
Initial evaluations demonstrate that PFC-S can reduce the average flow completion time and effectively prevent congestion spreading, and experimental results show that PFC-S provides better protection for victim flows compared to standard PFC, BFC, and HPCC methods.
Weimin Gao, Jiawei Huang, Qile Wang et al.· Journal of High Speed Networ...· 0 citations
Driven by the requirements of machine learning, cloud storage, and other network-intensive applications, remote direct memory access (RDMA) has been widely adopted in high-speed networks and is gradually being applied to geographically distributed datacenters. However, in cross-datacenter scenarios, long control loop latency and mixed traffic prevent existing RDMA congestion control schemes from perceiving and reacting to congestion in a timely and fair manner; this can lead to severe performance degradation and unfairness. To address these issues, we propose cdcPIM, a proactive congestion control scheme extended from datacenter parallel iterative matching (dcPIM) for cross-datacenter networks, which restructures the end-to-end control loop by introducing switch-coordinated control points, effectively transforming long-haul, RTT-bound feedback into localized control. Specifically, cdcPIM deploys a local control point by moving the token generation from the receiver to the sender side cross-datacenter switch, constraining the congestion control loop for inter-datacenter traffic within a single datacenter. Furthermore, cdcPIM introduces a remote control point to perform admission control for inter-datacenter traffic entering the receiver’s datacenter, thus avoiding intra-datacenter congestion caused by traffic bursts. Simulations demonstrate that when cdcPIM manages inter-datacenter traffic while cooperating with datacenter quantized congestion notification (DCQCN) for intra-datacenter traffic, long-haul congestion is effectively mitigated. Under mixed cross-datacenter workloads, DCQCN + cdcPIM reduces the overall average flow completion time (FCT) slowdown by up to 25.7% and the P99 FCT slowdown of intra-DC flows by up to 65.0% compared with the baseline scheme Themis.
Wenqiang Deng, Junyan Chen, Xuefeng Huang et al.· Journal of King Saud Univers...· 0 citations
To better utilize the switch memory for packet buffering, the packet buffer in datacenter switches is shared across all switch ports. As a result, the buffer sharing algorithm is extremely critical to the performance of datacenter switches and networks. Previous work has shown that push-out algorithms for buffer sharing achieve much higher throughput than drop-tail algorithms. However, switches today still implement drop-tail algorithms, as it is extremely challenging to implement push-out operations at the line rate of datacenter switches. In this paper, we present OBM, a new system for managing shared packet buffers in network switches based on the best-known online push-out algorithm for buffer sharing called the Longest Queue Drop (LQD). OBM makes two key contributions. First, OBM extends the classic LQD algorithm by making it priority-aware for packet admission and push-out, without sacrificing LQD's throughput guarantee. Second, OBM makes LQD-based push-out practical to implement on modern switches, by proposing a novel hardware pipeline and tree-based switching interconnect that can perform push-out operations at line rate with low latency and low hardware resource usage. We synthesize the OBM's design on both an FPGA and an ASIC compiler. Our prototype of the OBM switch is both high performance and consumes nominal hardware resources. Using large-scale network simulations, we show that OBM not only significantly outperforms state-of-the-art drop-tail and push-out buffer management schemes, but also matches the performance of ideal albeit impractical LQD for a single priority class while significantly outperforming it for multiple priority classes.
Dan Mani Binu, Jason Lei, Vishal Shrivastav· Proceedings of the ACM SIGCO...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
