
Efficient federated learning system that improves model training and testing by intelligently selecting participants based on data utility and device capabilities, enhancing accuracy and performance.

Optimizing tensor programs with partially equivalent transformations, improving DNN efficiency by up to 2.5×. Introduces PET framework for automated corrections and theoretical foundations for equivalence examination.

Innovative hardware-software design for scalable, dynamic secure memory in enclaves, featuring Guarded Page Tables and Mountable Merkle Trees to enhance protection, performance, and resource utilization.

MAGE: An execution engine for secure computation, enabling efficient processing of large-scale encrypted data by leveraging oblivious memory access patterns to optimize virtual memory management.

Explore three approaches to accurately calculate latency SLOs using histograms for correct quantiles. Learn to frame SLOs effectively for improved service level management and monitoring.

Explore Twitter's Aperture algorithm for client-side load balancing, addressing scalability challenges in large service clusters and improving RPC framework efficiency.

Strategies for resilient data pipelines: observability, immutable inputs, declarative pipelines, and data validation. Reduce operational complexity, update risks, and accuracy concerns in complex systems.

Explore scaling Kafka with limited resources, covering challenges, failures, and tactical approaches for managing complex distributed systems in a fast-growing business environment.

Explore best practices for implementing and improving code review processes, focusing on organizational culture, author communication, and reviewer strategies for high-quality software development.

Interactive workshop on designing distributed systems, covering scaling, failure handling, reliability, and consistency. Participants apply concepts to real-world scenarios using cloud components.

Explore categories of unforeseen, catastrophic system failures and strategies to fortify against them, including capacity testing, incident management, and communication.

Explore how software engineering can learn from NASA's Challenger incident, drawing parallels to modern reliability challenges and navigating complexity in system architecture.

Explore cognitive science techniques to enhance SRE learning, incident response, and team collaboration. Optimize human observability for better system understanding and problem-solving.

Intel's DMP storage architecture employs Ethernet-based fabric, disaggregated storage, and Kubernetes management. It offers NVMe-oF remote volumes and local Optane DCPM volumes, optimized for different I/O patterns and managed by LINBIT's storage orchest…

Comprehensive study on accelerating erasure coding for data storage, exploring optimization techniques, bitmatrix design, and vectorization to improve computation efficiency and performance in various scenarios.

Explores efficient graph processing on emerging storage devices, proposing a new partitioning framework that leverages improved random access capabilities for 2X better performance than state-of-the-art solutions.

ScaleCheck: A novel approach for finding scalability bugs in large distributed systems using program analysis and colocation techniques, enabling testing at scale on a single commodity PC.

Explores design tradeoffs in SSD reliability techniques, proposing a holistic management scheme for balancing performance, reliability, and lifetime across various workloads and wear states.

Innovative distributed caching mechanism providing provable load balancing for large-scale storage systems. Partitions hot objects and adaptively routes queries, enabling linear cache throughput increase with node count.

Optimizing hash-based indexing for persistent memory: CCEH reduces overhead, guarantees constant lookup time, and ensures failure-atomicity without explicit logging, improving query latency compared to existing techniques.

Optimizing data structures and memory allocation to reduce bit flipping in byte-addressable non-volatile memory, improving power efficiency and memory longevity without compromising performance.