PANAMA: In-Network Aggregation for Shared Machine Learning Clusters
Explore PANAMA, a novel in-network aggregation framework for distributed ML training, enhancing performance and resource sharing in shared clusters.
The Quest for an Autonomous Storage Fabric - Building Self-Driving NVMe Storage Infrastructure
Explore the evolution of autonomous storage fabrics through advanced technologies like NVMe/FC, FPINs, and congestion control, focusing on self-driving capabilities and automated infrastructure management.
SIGCOMM 2020: 1RMA: Re-envisioning Remote Memory Access for Multi-Tenant Datacenters
Explore innovative remote memory access for multi-tenant datacenters, addressing RDMA challenges through connection-free operations, finite resource pools, and enhanced security features.
Designing Fast and Programmable Routers
Explore cutting-edge router design combining high performance and programmability. Learn about innovative primitives enabling line-rate programmable routers, including packet scheduling, stateful processing, and flow-based statistics measurement.
Network Protocols and TCP/IP - Lecture 6
Explore network protocols, focusing on TCP/IP, sliding windows, congestion control, and denial of service attacks. Includes practical demonstrations and packet tracing exercises.
Revisiting Congestion Control for Lossless Ethernet
Explores ACK-driven congestion control for lossless Ethernet, improving flow completion time and addressing issues like congestion spreading and head-of-line blocking through packet conservation principles.
Large-Scale Deployment of Data Center TCP (DCTCP) - Challenges and Experiences
Deployment experiences of DCTCP in large-scale data centers, highlighting performance improvements, unexpected challenges, and open research questions in network congestion control.
Precise Data Center Traffic Engineering with Constrained Hardware Resources
Optimizing data center traffic routing with limited hardware resources, adapting allocation based on volume and precision impact, resulting in improved load balancing and reduced congestion.
Pudica: Near-Zero Queuing Delay in Congestion Control for Cloud Gaming
Innovative congestion control algorithm for cloud gaming, reducing frame delay and stall rates while increasing bitrate, enhancing player experience through near-zero queuing and high link utilization.
Towards Provably Performant Congestion Control
Explores framework for designing congestion control algorithms with provable performance across diverse network scenarios, synthesizing algorithms and uncovering fundamental impossibility results.
Harmony: A Congestion-free Datacenter Architecture
Explore a novel datacenter architecture that guarantees congestion-free message delivery, offering bounded queueing and delays without sacrificing network utilization or performance.
SMUFF: Towards Line Rate Wi-Fi Direct Transport with Orchestrated On-device Buffer Management
Innovative Wi-Fi direct transport solution achieving near-Gbps speeds for efficient peer-to-peer file transfers, optimizing buffer management and reducing resource consumption on smartphones.
Circulation Control for Faster Minimum Cost Flow in Unit-Capacity Graphs
Explores advanced techniques for optimizing minimum cost flow in unit-capacity graphs, focusing on circulation control methods to enhance efficiency and reduce computational complexity.
Unit Capacity Maxflow in Almost m - 4/3 Time
Explore advanced algorithms for solving unit capacity maxflow problems, featuring innovative techniques and improved time complexity.
Shoal - A Network Architecture for Disaggregated Racks
Innovative network architecture for disaggregated racks using fast circuit switches, offering high density and performance with lower power consumption. Addresses unique challenges in connecting dense clusters of compute, memory, and storage blades.
Backpressure Flow Control
Explores Backpressure Flow Control, a novel congestion control architecture for data center networks, offering improved throughput and latency compared to end-to-end schemes.
Datacenter RPCs Can Be General and Fast
Efficient RPC library for datacenter networks offering high performance without sacrificing generality. Handles packet loss and congestion while achieving low latency and high bandwidth on commodity hardware.
The nanoPU - A Nanosecond Network Stack for Datacenters
Innovative NIC-CPU co-design accelerating datacenter applications with μs-scale RPCs. Features fast network-to-CPU path, hardware-supported transport and load balancing, achieving 69ns wire-to-wire response time and improved performance.
Adapting TCP for Reconfigurable Datacenter Networks
Explores TCP adaptation for modern reconfigurable datacenter networks, proposing dynamic queue resizing and explicit circuit state feedback to improve circuit utilization and performance.
TCP - Architecture, Enhancements, and Tuning
Comprehensive exploration of TCP architecture, enhancements, and tuning, covering core concepts, connection establishment, flow control, congestion avoidance, and performance optimization techniques for network engineers.
Bolt - Sub-RTT Congestion Control for Ultra-Low Latency
Innovative congestion control system for ultra-fast data center networks, reducing latency and improving flow completion time while maintaining high utilization at speeds up to 400Gbps.