
Innovative checkpointing system for large ML models, reducing bandwidth and storage needs through differential checkpointing and quantization techniques without compromising training accuracy.

Explore PCAT, a system addressing evolving network telemetry challenges at Facebook through change tracking, intent-based layering, and adaptive design for large-scale production environments.

Explore NetHint, a white-box networking approach for multi-tenant data centers. Learn how cloud providers and tenants can collaborate to enhance application performance through adaptive data transfer scheduling.

Explore RDMA's Turing completeness and its potential for complex offloads in distributed systems, enhancing performance and reducing CPU load through innovative programming techniques.

Innovative multicast architecture for datacenter networks, reducing switch state and communication overhead while improving scalability and efficiency for large-scale group communications.

Explore FlexTOE, a high-performance TCP offload engine for SmartNICs. Learn about its flexible design, fine-grained parallelism, and performance benefits in various network scenarios and hardware architectures.

Explore a data-driven approach to optimize CDN performance by dynamically configuring web server parameters based on network and device characteristics, improving tail web performance by 32-67%.

Explores improving application performance by deliberately delaying and reordering network packets to increase traffic locality, enhancing cache efficiency in high-speed network interfaces.

Edge-queued datagram service for datacenters moves queuing to sending hosts, enabling multiple protocols to coexist, improving performance, and facilitating future enhancements without modifying higher-layer stacks.

Scalable hardware acceleration architecture for stateful layer-4 load balancing, supporting high traffic rates, large concurrent flows, and numerous new connections without traffic pattern assumptions.

Explore the design of ultra-low latency wide-area networks using free-space microwave wireless connectivity, addressing practical challenges and demonstrating economic viability.

Innovative approach to scale Open vSwitch using NuevoMatch algorithm, improving packet classification performance and rule update speed for enhanced network virtualization efficiency.

Innovative virtual switch solution preventing packet loss and improving network performance through dynamic queuing, doorbell queues, and overlay network techniques for enhanced application responsiveness.

Explores Backpressure Flow Control, a novel congestion control architecture for data center networks, offering improved throughput and latency compared to end-to-end schemes.

Explore how Microsoft Teams adapted to enable remote education for millions during COVID-19, discussing challenges, solutions, and engineering strategies for large-scale online learning.

Explores DeepSketch, a machine learning-based technique for post-deduplication delta compression in storage systems, improving data reduction efficiency through enhanced reference search accuracy.

Explores a novel solution to remove double-logging in LSM-tree based relational databases, improving performance and efficiency while maintaining data reliability and fast recovery.

Resilient remote memory system with low latency, high availability, and efficient erasure coding. Improves performance-efficiency tradeoff and reduces data loss probability under correlated failures.

Explores FusionFS, a firmware-level filesystem leveraging near-storage computation to reduce I/O bottlenecks through CISCOps, offering improved performance and faster recovery for applications.

Explore InfiniFS, an efficient metadata service for large-scale distributed filesystems, addressing challenges like load balancing, path resolution, and near-root hotspots.

Explores innovative techniques to enhance BFS-like algorithm performance on semi-external graph systems, focusing on I/O efficiency improvements through strategic graph data pre-processing.